Data Analytics For Accounting, 2nd Edition Vernon Richardson PDF

page i
Data Analytics for Accounting

SECOND EDITION
Vernon J. Richardson
University of Arkansas,
Baruch College
Ryan A. Teeter
University of Pittsburgh
Katie L. Terrell
University of Arkansas
page ii
DATA ANALYTICS FOR ACCOUNTING
Published by McGraw-Hill Education, 2 Penn Plaza, New

York, NY 10121. Copyright © 2021 by McGraw-Hill
Education. All rights reserved. Printed in the United States of
America. Previous editions © 2019. No part of this
publication may be reproduced or distributed in any form or
by any means, or stored in a database or retrieval system,
without the prior written consent of McGraw-Hill Education,
including, but not limited to, in any network or other electronic
storage or transmission, or broadcast for distance learning.
Some ancillaries, including electronic and print components,

may not be available to customers outside the United States.
This book is printed on acid-free paper.
1 2 3 4 5 6 7 8 9 LWI 24 23 22 21 20
ISBN 978-1-260-57109-7
MHID 1-260-57109-2
Cover Image: Rawpixel.com/Shutterstock

All credits appearing on page or at the end of the book are
considered to be an extension of the copyright page.
The Internet addresses listed in the text were accurate at the

time of publication. The inclusion of a website does not
indicate an endorsement by the authors or McGraw-Hill
Education, and McGraw-Hill Education does not guarantee
the accuracy of the information presented at these sites.
mheducation.com/highered
page iii
Dedications
My wonderful daughter, Melissa, for your

constant love, encouragement and support.
—Vern Richardson
My wife, Erin, and children, Sylvia and

Theodore.
—Ryan Teeter
To my co-author, friend, and colleague,

Vernon Richardson. Thank you for inviting
me to be on this textbook journey. And thank
you for your guidance and patience—I’m
thrilled to be a part of your team!
—Katie Terrell
page iv
Preface
Data Analytics is changing the business world—data simply
surrounds us! So much data is available to businesses about each of
us—how we shop, what we read, what we buy, what music we listen
to, where we travel, whom we trust, where we invest our time and
money, etc. Accountants can create value by addressing
fundamental business and accounting questions using data
analytics.
All accountants must develop data analytic skills to address the
needs of the profession in the future. Data Analytics for Accounting,
2e recognizes that accountants don’t need to become data scientists
—they may never need to build a data repository or do the real hard-
core Data Analytics or learn how to program a computer to do
machine learning. However, there are seven skills that analytic-
minded accountants must have to be prepared for a data-filled world,
including:
1. An analytics mindset—recognize when and how Data Analytics can
address accounting questions.
2. Data scrubbing and data preparation—comprehend the process
needed to extract (query), clean and prepare the data before
analysis.
3. Data quality—recognize what is meant by data quality, be it
completeness, reliability, or validity.
4. Descriptive data analysis—perform basic analysis to understand the
quality of the underlying data and their ability to address the
business question.
5. Data analysis through data manipulation—demonstrate ability to
sort, rearrange, merge, and reconfigure data in a manner that
allows enhanced analysis.
6. Problem solving through statistical data analysis—identify and
implement an approach that will use statistical data analysis to draw
conclusions and make recommendations on a timely basis.
7. Data visualization and data reporting—report results of analysis in
an accessible way to each varied decision maker and his or her
specific needs.
Consistent with these skills, it’s important to recognize that Data
Analytics is a process. The process begins by identifying business
questions that can be addressed with data, extracting and testing the
data, refining our testing, and finally, communicating those findings
to management. Data Analytics for Accounting, 2e describes this
process by relying on an established data analytics model called the
IMPACT cycle1
1. Identify the question.
2. Master the data.
3. Perform test plan.
4. Address and refine results.
5. Communicate insights.
6. Track outcomes
page v
Adapted from Win with Advanced Business Analytics: Creating
Business Value from Your Data, by Jean Paul Isson and Jesse S.
Harriott.
The IMPACT cycle is described in the first four chapters and then
the process is illustrated in audit, managerial accounting, financial
accounting and tax in Chapters 5-9, adding an all-new tax chapter to
Data Analytics for Accounting, 2e. In response to instructor
feedback, Data Analytics for Accounting, 2e now also includes two
new project chapters, giving students a chance to practice the full
IMPACT model with multiple labs that build on each other.
Data Analytics for Accounting, 2e emphasizes hands-on practice.
Students are provided with hands-on instruction (e.g., click-by-click
instructions, screenshots, etc.) on datasets within the chapter; within
the end-of-chapter materials; and in the labs at the end of each
chapter. Throughout the text, students identify questions, extract and
download data, perform testing, and then communicate the results of
that testing.
The use of real-world data is highlighted by using data from
LendingClub, College Scorecard, Dillard’s, the State of
Oklahoma, as well as other data from our labs. In particular, we
emphasize the rich data from Dillard’s sales transactions that we
use in more than fifteen of the labs throughout the text (including
Chapter 11).
Data Analytics for Accounting, 2e also emphasizes the various
data analysis tools students will use throughout the rest of their
career—Microsoft Excel, Microsoft Access (including SQL), Tableau
(free student license), IDEA (free student license), and Weka (free
student license). Using multiple tools allows students to learn which
tool is best suited for the necessary data analysis, data visualization,
and communication of the insights gained—for example, which tool
is easiest for internal controls testing, which is best for analysis or
querying (using SQL) big datasets, which is best for data
visualizations, and so on.
1Jean Paul Isson and Jesse S. Harriott, Win with Advanced Business Analytics: Creating
Business Value from Your Data (Hoboken, NJ: Wiley, 2013).
page vi
About the Authors
Vernon J. Richardson
Vernon J. Richardson is a Distinguished Professor of Accounting

and the G. William Glezen Chair in the Sam M. Walton College of
Business at the University of Arkansas and a Visiting Professor at
Baruch College. He received his BS, Master of Accountancy, and
MBA from Brigham Young University and a PhD in accounting from
the University of Illinois at Urbana–Champaign. He has taught
students at the University of Arkansas, Baruch College, University of
Illinois, Brigham Young University, Aarhus University, and University
of Kansas and internationally at the China Europe International
Business School (Shanghai), Xi’an Jiaotong Liverpool University,
and the University of Technology Sydney.
Dr. Richardson is a member of the American Accounting
Association. He has served as president of the American Accounting
Association Information Systems section. He previously served as
an editor of The Accounting Review and is currently an editor at
Accounting Horizons. He has published articles in The Accounting
Review, Journal of Information Systems, Journal of Accounting and
Economics, Contemporary Accounting Research, MIS Quarterly,
International Journal of Accounting Information Systems, Journal of
Management Information Systems, Journal of Operations
Management, and Journal of Marketing. Dr. Richardson is also the
author of McGraw-Hill’s Accounting Information Systems textbook.
Ryan A. Teeter
Ryan A. Teeter is a Clinical Assistant Professor of Accounting in the

Katz Graduate School of Business at the University of Pittsburgh. He
teaches accounting information systems, auditing, and accounting
data analytics. Prior to receiving his PhD in accounting information
systems from Rutgers University, he worked at Google in Mountain
View, California. He has since worked with internal audit
organizations at Siemens, Procter & Gamble, Alcoa/Arconic, and
FedEx, helping to develop robotic process automation programs and
data analytic solutions.
Dr. Teeter is a member of the American Accounting Association
and has published articles in the Journal of Strategic Technologies in
Accounting and Issues in Accounting Education. He has received
grant funding for data analytics research from PwC.
Katie L. Terrell
Katie L. Terrell is an instructor in the Sam M. Walton College of
Business at the University of Arkansas. She received her BA
degrees in English literature and in the Spanish language from the
University of Central Arkansas and her MBA from the University of
Arkansas. She expects a doctoral degree by 2020. She has taught
students at the University of Arkansas; Soochow University (Suzhou,
China); the University College Dublin (Ireland); and Duoc UC, a
branch of the Catholic University of Chile (Vina del Mar, Chile).
She is a member of the American Accounting Association and
has published a Statement on Management Accounting for the
Institute of Management Accountants on managing organizational
change in operational change initiatives. Terrell was named the 2019
Business Professional of the Year (Education) by the national Beta
Alpha Psi organization. She has recently been recognized for her
innovative teaching by being the recipient of the Mark Chain/FSA
Teaching Award for innovative graduate-level accounting teaching
practices in 2016. She has worked with Tyson Foods, where she
held various information system roles, focusing on business
analysis, project management for ERP implementations and
upgrades, and organizational change management.
page vii
Acknowledgments
Our sincere thanks to all who helped us on this project.
Our biggest thanks to the awesome team at McGraw-Hill
Education, including Steve Schuetz, Tim Vertovec, Allie Kukla, Fran
Simon, Kevin Moran, and Sarah Wood.
Our thanks also to each of the following:
The Walton College Enterprise Team (Paul Cronan, Ron Freeze,
Michael Gibbs, Michael Martz, Tanya Russell) for their work helping
us get access to the Dillard’s data.
Shane Lunceford from LendingClub for helping gain access to
LendingClub data.
Marcia Watson, University of North Carolina–Charlotte; Ryan
Baxter, Boise State University; Antoinette Smith, Florida International
University; and Lorrie Metzger, University at Buffalo for their
accuracy check and review of the manuscript.
In addition, the following reviewers and classroom testers who
provided ideas and insights for this edition. We appreciate their
contributions.
Amelia Annette Baldwin

University of South Alabama
Dereck Barr-Pulliam
University of Wisconsin–Madison
Heather Carrasco
Texas Tech University
Elizabeth Felski
State University of New York at Geneseo
Chris C. Hsu
York College, City University of New York
Venkataraman Iyer
University of North Carolina at Greensboro
Andrea S. Kelton
Middle Tennessee State University
Brandon Lock
Baruch College, CUNY
Sharon M. Lightner
National University
Margarita Maria Lenk
Colorado State University
Partha Mohapatra
California State University, Sacramento
Uday Murthy
University of South Florida
Kalana Malimage
University of Wisconsin–Whitewater
Bonnie Morris
Duquesne University
Kathy Nesper
University at Buffalo
Ali Saeedi
University of Minnesota Crookston
Karen Schuele
John Carroll University
Drew Sellers
Kent State University
Joe Shangguan
Robert Morris University
Vincent J. Shea
St. John’s University
Marcia Watson
University of North Carolina at Charlotte
Liu Yang
Southeast Missouri State University
Zhongxia Ye
University of Texas, San Antonio
Qiongyao (Yao) Zhang
Robert Morris University
Vernon Richardson
Ryan Teeter
Katie Terrell
page viii
Key Features
Emphasis on Skills: Working through the IMPACT cycle
framework, students will learn problem assessment, data
preparation, data analysis, data visualization, control contesting,
and more.
Emphasis on Hands-On Practice: Students will be provided
hands-on learning (click-by-click instructions with screenshots) on
datasets within each chapter, within the end-of-chapter materials,
and in the labs and comprehensive cases.
Emphasis on Datasets: To illustrate data analysis techniques and
skills, multiple practice datasets (audit, financial, and managerial
data) will be used in every chapter. Students gain real-world
experience working with data from LendingClub, Dillard’s,
College Scorecard, the State of Oklahoma, as well as financial
statement data (via XBRL) from Fortune 100 companies.
Emphasis on Tools: Students will learn how to conduct data
analysis using Excel Access (including SQL), Tableau (free student
license), IDEA (free student license), and Weka (free student
license). Students will compare and contrast the different tools to
determine which are best suited for basic data analysis and data
visualization, which are easiest for internal controls testing, which
are best for SQL queries, and so on.
©Tableau Software, Inc. All rights reserved.
page ix
Main Text Features

Chapter Maps
These maps provide a guide of what we’re going to cover in the
chapter as well as a guide of what we’ve just learned and what’s
coming next.
Chapter-Opening Vignettes
Because companies are facing the new and exciting opportunities
with their use of Data Analytics to help with accounting and business
decisions, we detail what they’re doing and why in our chapter-
opening vignettes.
Learning Objectives
We feature learning objectives at the beginning of each chapter.
Having these learning objectives provides students with an overview
of the concepts to be taught in the chapter and the labs.
Progress Checks
Periodic progress check questions are posed to the students
throughout each chapter. These checks provoke the student to stop
and consider the concepts presented.
page x
End-of-Chapter Materials
Answers to Progress Checks
Allow students to evaluate if they are on track with their
understanding of the materials presented in the chapter.
Multiple Choice Questions

Quickly assess student’s knowledge of chapter content.
Discussion Questions
Provide questions for group discussion.
page xi
Problems
Challenge the student’s ability to see relationships in the learning
objectives by employing higher-level thinking and analytical skills.
Labs
Give students hands-on experience working with different types of
data and the tools used to analyze them. Students will conduct data
analysis using Excel, Access (including SQL), Tableau, IDEA, XBRL,
and Weka.
Comprehensive Cases
Use a real-life Big Data set based on Dillard’s actual company data.
This dataset allows students to build their skills and test their
conclusions across concepts covered in each chapter. The
Comprehensive Cases can be followed continuously from the first
chapter or picked up at any later point in the book; enough
information is provided to ensure students can get right to work.
page xii
Data Analytics for Accounting,

2e Content Updates
General Updates for the 2nd Edition
Added additional End-of-Chapter Multiple Choice Questions and
Problems throughout the text.
Significantly revised many End-of-Chapter Problems for availability
and auto-grading within Connect.
Revised and added many new Discussion Questions in most
chapters.
Chapter by Chapter Updates

Specific chapter changes for Data Analytics for Accounting, 2nd
Edition, are as follows:
Chapter 1
Updated the opening vignette and statistics on Alibaba sales and
use of e-commerce.
Updated the statistics and screenshots for Lending Club Analysis.
Revised Connect questions for problems and labs.
Chapter 2
Improved and clarified the discussion of relational databases,
including updated figures.
Expanded the discussion of different RDBMS (Access, SQLite, and
SQL Server).
Improved discussion of Excel and SQL. The brief introduction to
how to use SQL now has its own place in a dedicated appendix at
the end of the text, and it has been vastly expanded to teach
beginners how to write queries.
Expanded the discussion on data quality.
Added a brief discussion of ETL v. ELT.
Improved labs for clarity and a better learning experience,
particularly Labs 2-1, 2-2, and 2-4.
Chapter 3
Reorganized chapter structure to follow the descriptive, diagnostic,
predictive, and prescriptive approaches to Data Analytics.
New exhibits and examples to illustrate analytics approaches.
Removed previous edition flowchart for model selection.
Additional explanation and examples of each of the methods and
approaches.
Improved labs for clarity.
Chapter 4
Updated the opening vignette.
Improved the discussion on the differences between qualitative and
quantitative data and the discussion of the normal distribution.
Improved and clarified how to select a visualization based on the
four chart types (qualitative vs. quantitative and declarative vs.
exploratory).
page xiii
Updated the discussion on the Gartner Quadrant to take
into account Gartner’s January 2019 analysis of BI tools (focusing
on Excel and Tableau).
Extended the discussion on written and spoken communication.
Added a lab to work with visualizing data and creating dashboards
in Power BI to interactively compare the tool with Tableau.
Chapter 5
Expanded discussion on the modern data environment.
Included additional examples of the Audit Data Standard.
Improved and clarified content to match the focus on descriptive,
diagnostic, predictive, and prescriptive analytics.
New labs (5-1 and 5-2) that have students transform data using a
common data model.
Improved existing labs.
Chapter 6
Clarified chapter content to match the focus on descriptive,
diagnostic, predictive, and prescriptive analytics.
Improved labs.
Chapter 7
Clarified chapter content and provided additional new exhibits and
examples, such as variance analysis.
Improved labs.
Chapter 8
Reorganized chapter content to focus on financial statement
analysis using descriptive, diagnostic, predictive, and prescriptive
approaches.
Added new content on common size and ratio analysis.
Improved discussion of XBRL data.
Improved XBRL dataset (in Lab 8-4), accessible via Microsoft
Access and included options to do analysis in Excel.
Chapter 9
All-new chapter on tax analytics, including examples of tax data, tax
analysis, tax planning, and tax visualizations.
Chapter 10
All-new basic project chapter that explores the order-to-cash and
procure-to-pay cycles from different user perspectives.
Chapter 11
All-new advanced project chapter, estimating sales returns at
Dillard’s with three question sets highlighting descriptive and
exploratory analysis, hypothesis testing, and predictive analytics.
page xiv
Appendixes
Several all-new appendixes have been added to ease the lab
experience and introduce tools used or mentioned throughout the
text:
Appendix A: Basic Statistics Tutorial.
Appendix B: Accessing the Excel Data Analysis Toolpak.
Appendix C: Excel (Formatting, Sorting, Filtering, and PivotTables).
Appendix D: SQL Part 1. This tutorial introduces the SQL language
for extracting data and explains the following SQL syntax: SELECT,
FROM, INNER JOIN, ON, WHERE, GROUP BY, HAVING, ORDER
BY.
Appendix E: SQLite. We have added SQLite files as an option for
each lab that uses Microsoft Access. This lab explains how to
download SQLite and how to use the tool.
Appendix F: Power Query. This appendix contains a short tutorial
on transforming data using Power Query. How to access data files
on the University of Arkansas’ remote desktop is also discussed.
Appendix G: Tableau.
Appendix H: SQL Part 2: On the heels of learning Tableau, students
learn about more complex joins—LEFT and RIGHT.
Appendix I: Power BI.
Appendix J: Dillard’s ER Diagram.
Appendix K: Data Dictionaries.
page xv
Connect for Data Analytics for

Accounting
With Connect for Data Analytics in Accounting, your

students receive proven study tools and hands-on
assignment materials as well as an adaptive eBook. All of the
following assets are assignable in Connect.
SmartBook 2.0: A personalized and adaptive learning tool
used to maximize the learning experience by helping
students study more efficiently and effectively. Smartbook 2.0
highlights where in the chapter to focus, asks review
questions on the materials covered and tracks the most
challenging content for later review recharge. Smartbook 2.0
is available both online and offline.
Orientation Videos: Video-based tutorial assignments are
designed to train students via an overview video followed by
a quiz for each of the assignment types they will find in
Connect.
Multiple Choice Questions: The multiple choice questions
from the end-of-chapter materials are assignable in Connect,
providing students with instant feedback on their answers.
Problems: Select problems from the text are available for
assignment in Connect to ensure students are building an
analytical skill set.
page xvi
Labs: Select labs are assignable in Connect but will require

students to work outside of Connect to complete the lab.
Once completed, students go back into Connect to answer
questions designed to ensure they completed the lab and
understood the key skills and outcomes from their lab work.
Comprehensive Cases: Select comprehensive labs/cases
are assignable in Connect but will require students to work
outside of Connect to complete the lab using the Dillard’s
real-world Big Data set. Once students complete the
comprehensive lab, they will go back into Connect to answer
questions designed to ensure they completed the lab and
understood the key skills and outcomes from their lab work.
page xvii
Lab Walkthrough Videos: Get the help you need, when you
need it. These author-led videos will explain how to access
and use the tools needed to complete processes essential to
the labs.
Author Lecture Videos: Lecture Videos teach each
chapter’s core learning objectives and concepts through an
author-developed, hands-on presentation, bringing the text
content to life. The videos have the touch and feel of a live
lecture, rather than a canned presentation, so you can learn
at your own pace.
Test Bank: The test bank includes auto-graded multiple

choice and true/false assessment questions. It is available in
Connect and Test Builder.
page xviii
FOR INSTRUCTORS
You’re in the driver’s seat.

Want to build your own course? No problem. Prefer to use our
turnkey, prebuilt course? Easy. Want to make changes throughout
the semester? Sure. And you’ll save time with Connect’s auto-
grading too.
Less Time Grading
They’ll thank you for it.

Adaptive study resources like SmartBook® 2.0 help your students be
better prepared in less time. You can transform your class time from
dull definitions to dynamic debates. Find out more about the powerful
personalized learning experience available in SmartBook 2.0 at
www.mheducation.com/highered/connect/smartbook
Laptop: McGraw-Hill; Woman/dog: George Doyle/Getty Images
Make it simple, make it affordable.

Connect makes it easy with seamless integration using any of
the major Learning Management Systems—Blackboard®,
Canvas, and D2L, among others—to let you organize your
course in one convenient location. Give your students access to
digital materials at a discount with our inclusive access
program. Ask your McGraw-Hill representative for more
information.
Padlock: Jobalou/Getty Images
Solutions for your challenges.

A product isn’t a solution. Real solutions are affordable, reliable,
and come with training and ongoing support when you need it
and how you want it. Our Customer Experience Group can also
help you troubleshoot tech problems—although Connect’s 99%
uptime means you might not need to call them. See for yourself
at status.mheducation.com
Checkmark: Jobalou/Getty Images
page xix
FOR STUDENTS
Effective, efficient studying.

Connect helps you be more productive with your study time and get
better grades using tools like SmartBook 2.0, which highlights key
concepts and creates a personalized study plan. Connect sets you
up for success, so you walk into class with confidence and walk out
with better grades.
Study anytime, anywhere.

Download the free ReadAnywhere app and access your online
eBook or SmartBook 2.0 assignments when it’s convenient, even if
you’re offline. And since the app automatically syncs with your
eBook and SmartBook 2.0 assignments in Connect, all of your work
is available every time you open it. Find out more at
www.mheducation.com/readanywhere
“I really liked this app—it made it easy to

study when you don’t have your textbook
in front of you.”
- Jordan Cunningham, Eastern Washington University
No surprises.
The Connect Calendar and Reports tools keep you on track with the
work you need to get done and your assignment scores. Life gets
busy; Connect tools help you keep learning through it all.
Calendar: owattaphotos/Getty Images
Learning for everyone.

McGraw-Hill works directly with Accessibility Services Departments
and faculty to meet the learning needs of all students. Please contact
your Accessibility Services office and ask them to email
accessibility@mheducation.com, or visit
www.mheducation.com/about/accessibility for more information.
Top: Jenner Images/Getty Images, Left: Hero Images/Getty Images, Right: Hero
Images/Getty Images
page xx
Brief Table of Contents

Preface iv
About the Authors vi
Acknowledgments vii
Key Features viii
Main Text Features ix
End-of-Chapter Materials x
Data Analytics for Accounting, 2e Content Updates xii
Connect for Data Analytics for Accounting xv
Chapter 1 Data Analytics for Accounting and Identifying
the Questions 2
Chapter 2 Mastering the Data 38
Chapter 3 Performing the Test Plan and Analyzing the
Results 92
Chapter 4 Communicating Results and Visualizations
138
Chapter 5 The Modern Accounting Environment 200
Chapter 6 Audit Data Analytics 224
Chapter 7 Managerial Analytics 268
Chapter 8 Financial Statement Analytics 326
Chapter 9 Tax Analytics 360
Chapter 10 Project Chapter (Basic) 390
Chapter 11 Project Chapter (Advanced): Analyzing
Dillard’s Data to Predict Sales Returns 408
Appendix A Basic Statistics Tutorial 434
Appendix B Accessing the Excel Data Analysis Toolpak
440
Appendix C Excel (Formatting, Sorting, Filtering, and
PivotTables) 442
Appendix D SQL Part 1 453
Appendix E SQLite 466
Appendix F Power Query 470
Appendix G Tableau 476
Appendix H SQL Part 2 480
Appendix I Power BI 483
Appendix J Dillard’s ER Diagram 491
Appendix K Data Dictionaries 492
GLOSSARY 500
INDEX 504
page xxi
Detailed TOC
Chapter 1
Data Analytics for Accounting and Identifying the
Questions 2
Data Analytics 4
How Data Analytics Affects Business 4
How Data Analytics Affects Accounting 5
Auditing 5
Financial Reporting 6
Taxes 7
The Data Analytics Process Using the Impact Cycle 8
Step 1: Identify the Questions (Chapter 1) 8
Step 2: Master the Data (Chapter 2) 8
Step 3: Perform Test Plan (Chapter 3) 9
Step 4: Address and Refine Results (Chapter 3) 11
Steps 5 and 6: Communicate Insights and Track
Outcomes (Chapter 4 and each chapter thereafter) 11
Back to Step 1 12
Data Analytic Skills Needed by Analytic-Minded
Accountants 12
Hands-on Example of the Impact Model 13
Identify the Questions 13
Master the Data 13
Perform Test Plan 15
Address and Refine Results 17
Communicate Insights 19
Track Outcomes 19
Summary 20
Key Words 20
Answers to Progress Checks 21
Multiple Choice Questions 23
Discussion Questions 24
Problems 24
Lab 1-0 How to Complete Labs in This Text 27
Lab 1-1 Data Analytics in Financial Accounting 28
Lab 1-2 Data Analytics in Managerial Accounting 31
Lab 1-3 Data Analytics in Auditing 33
Lab 1-4 Comprehensive Case: Dillard’s Store Data 34
Chapter 2
Mastering the Data 38
How Data are Used and Stored in the Accounting Cycle 40
Data and Relationships in a Relational Database 42
Columns in a Table: Primary Keys, Foreign Keys, and
Descriptive Attributes 42
Data Dictionaries 44
Extract, Transform, and Load (ETL) the Data 45
Extract 46
Transform 49
Load 52
Summary 52
Key Words 53
Problems 57
Lab 2-1 Create a Request for Data Extraction 58
Lab 2-2 Use PivotTables to Denormalize and Analyze the
Data 60
Lab 2-3 Resolve Common Data Problems in Excel and
Access 67
Lab 2-4 Generate Summary Statistics in Excel 71
Lab 2-5 College Scorecard Extraction and Data
Preparation 73
Lab 2-6 Comprehensive Case: Dillard’s Store Data: How to
Create an Entity-Relationship Diagram 75
Lab 2-7 Comprehensive Case: Dillard’s Store Data: How to
Preview Data from Tables in a Query 78
Lab 2-8 Comprehensive Case: Dillard’s Store Data:
Connecting Excel to a SQL Database 80
Lab 2-9 Comprehensive Case: Dillard’s Store Data: Joining
Tables 90
Chapter 3
Performing the Test Plan and Analyzing the Results 92
Performing the Test Plan 94
Descriptive Analytics 97
Example of Data Reduction in Internal and External
Auditing 98
Examples of Data Reduction in Other Accounting
Areas 99
Diagnostic Analytics 100
Example of Profiling in Management Accounting 102
Example of Profiling in an Internal Audit 103
Example of Profiling in Auditing 103
Example of the Clustering Approach in Auditing 105
Predictive Analytics 106
page xxii
Examples of the Regression Approach in Managerial

Accounting 108
Examples of the Regression Approach in Auditing 108
Other Examples of the Regression in Accounting 108
Classification Terminology 109
Evaluating Classifiers 111
Overfitting 111
Prescriptive Analytics 112
Summary 114
Key Words 115
Problems 119
Lab 3-1 Data Reduction Using Fuzzy Matching 121
Lab 3-2 Regression in Excel 125
Lab 3-3 Classification 127
Lab 3-4 Comprehensive Case: Dillard’s Store Data: Data
Abstract (SQL) and Regression (Part I) 130
Lab 3-5 Comprehensive Case: Dillard’s Store Data: Data
Abstract (SQL) and Regression (Part II) 135
Chapter 4
Communicating Results and Visualizations 138
Determine the Purpose of Your Data Visualization 141
Quadrants 1 and 3 versus Quadrants 2 and 4:
Qualitative versus Quantitative 142
A Special Case of Quantitative Data: The Normal
Distribution 143
Quadrants 1 and 2 versus Quadrants 3 and 4:
Declarative versus Exploratory 144
Choosing the Right Chart 145
Charts Appropriate for Qualitative Data 145
Charts Appropriate for Quantitative Data 147
Tools to Help When Picking a Visual 149
Learning to Create a Good Chart by (Bad) Example
151
Further Refining Your Chart to Communicate Better 156
Data Scale and Increments 157
Color 157
Communication: More than Visuals—Using Words to Provide
Insights 158
Content and Organization 158
Audience and Tone 159
Revising 160
Summary 160
Key Words 161
Problems 164
Lab 4-1 Use PivotCharts to Visualize Declarative Data
166
Lab 4-2 Use Tableau to Perform Exploratory Analysis and
Create Dashboards 168
Lab 4-3 Comprehensive Case: Dillard’s Store Data: Create
Geographic Data Visualizations in Tableau and in
Power BI 177
Visualizing Regression in Tableau 196
Chapter 5
The Modern Accounting Environment 200
The Modern Data Environment 202
The Increasing Importance of the Internal Audit 203
Enterprise Data 203
Common Data Models 204
Automating Data Analytics 206
Continuous Monitoring Techniques 208
Alarms and Exceptions 208
Working Papers And Audit Workflow 209
Electronic Working Papers and Remote Audit Work
209
Summary 210
Key Words 210
Problems 213
Lab 5-1 Create a Common Data Model 215
Lab 5-2 Create a Dashboard Based on a Common Data
Model 217
Lab 5-3 Set Up a Cloud Folder 219
Lab 5-4 Review Changes to Working Papers 220
Lab 5-5 Identify Audit Data Requirements 221
Lab 5-6 Prepare Audit Plan 222
Chapter 6
Audit Data Analytics 224
When to Use Audit Data Analytics 226
Identify the Problem 226
Master the Data 226
Perform the Test Plan 228
page xxiii

Communicate Insights 230
Track Outcomes 230
Descriptive Analytics 230
Age Analysis 231
Sorting 232
Summary Statistics 233
Sampling 233
Diagnostic Analytics and Benford’s Law 235
Z-Score 235
Benford’s Law 236
Drill-Down 239
Exact and Fuzzy Matching 239
Sequence Check 241
Stratification and Clustering 242
Creating Advanced Predictive and Prescriptive Analytics
242
Regression 242
Classification 242
Probability 242
Sentiment Analysis 243
Applied Statistics 243
Artificial Intelligence 243
Additional Analyses 243
Summary 244
Key Words 244
Problems 247
Lab 6-1 Evaluate the Master Data for Interesting
Addresses 248
Lab 6-2 Perform Substantive Tests of Account Balances
250
Lab 6-3 Finding Duplicate Payments 256
Hypothesis Testing (Part I) 257
Hypothesis Testing (Part II—Data Visualization)
264
Chapter 7
Managerial Analytics 268
Identifying Management Accounting Questions 270
Relevant Costs 270
Key Performance Indicators and Variance Analysis 270
Cost Behavior 271
Balanced Scorecard and Key Performance Indicators 272
Master the Data and Perform the Test Plan 276
Summary 278
Key Words 279
Problems 282
Lab 7-1 Evaluate Management Requirements and Identify
Useful KPIs from a List 284
Lab 7-2 Create a Balanced Scorecard Dashboard in
Tableau 286
Creating KPIs in Excel (Part I) 295
Creating KPIs in Excel (Part II) 302
Creating KPIs in Excel (Part III) 309
Creating KPIs in Excel (Part IV—Putting It All
Together) 316
Advanced Models in Tableau 321
Chapter 8
Financial Statement Analytics 326
Financial Statement Analysis 328
Descriptive Financial Analytics 328
Vertical and Horizontal Analysis 328
Other Classes of Ratios 329
Diagnostic Financial Analytics 331
Predictive Financial Analytics 331
Visualizing Financial Data 333
Showing Trends 333
Relative Size of Accounts 333
Text Mining and Sentiment Analysis 334
XBRL and Financial Data Quality 336
XBRL Data Quality 338
XBRL, XBRL-GL, and Real-Time Financial Reporting
340
Examples of Financial Statement Analytics Using
XBRL 340
Summary 341
Key Words 341
Problems 344
Lab 8-1 Create a Horizontal and Vertical Analysis Using
XBRL Data 346
page xxiv
Lab 8-2 Create Dynamic Common Size Financial

Statements 349
Lab 8-3 Analyze Financial Statement Ratios 352
Lab 8-4 Use PivotTables to Analyze Data from an XBRL
Database 355
Chapter 9
Tax Analytics 360
Introduction to Tax Analytics 362
Mastering the Data through Tax Data Management 363
Tax Data Analytics Visualizations 364
Tax Data Analytics Visualizations and Tax Compliance
364
Evaluating Sales Tax Liability 365
Evaluating Income Tax Liability 365
Tax Data Analytics for Tax Planning 367
What-If Scenarios 368
What-If Scenarios for Potential Legislation, Deductions,
and Credits 369
Summary 370
Key Words 370
Problems 373
Lab 9-1 State Sales Taxes and Create a Data
Visualization 375
Lab 9-2 Comprehensive Case 1: Dillard’s Store Data:
Calculate Sales Tax for Dillard’s States 379
Calculate Sales Tax for Dillard’s States Part 2—
Compare Year over Year 381
Calculate Sales Tax for Dillard’s States Part 3—
Calculate City Tax and Compare Tax Owed Year
over Year 386
Lab 9-5 Comprehensive Case 4: Dillard’s Store Data: Does
a State’s Tax Rate Affect Dillard’s Decision to
Open Stores There? 387
Chapter 10
Project Chapter (Basic) 390
Evaluating Business Processes 392
Question Set 1: Order-to-Cash 392
Question 1.1: How Efficiently Are We Collecting Our
Cash? 392
Question 1.2: Is the Delivery Process Following the
Expected Procedure? 396
Question 1.3: What Is the Total Revenue and Balance in
Accounts Receivable? 398
Question 1.4: What Else Can You Determine about the
O2C Process? 400
Question Set 2: Procure-to-Pay 400
Question 2.1: How Long Are We Taking to Pay Our
Invoices? 400
Question 2.2: Are There Any Erroneous Payments?
404
Question 2.3: Are We Missing Out on Discounts by
Paying Late? 405
Question 2.4: What Else Can You Determine about the
P2P Process? 406
Chapter 11
Project Chapter (Advanced): Analyzing Dillard’s Data to
Predict Sales Returns 408
Estimating Sales Returns 410
Question Set 1: Descriptive and Exploratory Analysis 410
Question 1.1: Which Attributes Could Help Predict
Percentage of Returned Sales? 410
Question 1.2: How Can We Explore the Product
Hierarchy Through Data Visualization? 412
Question Set 2: Diagnostic Analytics—Hypothesis Testing
422
Question 2.1: Is the Percentage of Sales Returned
Significantly Higher in January After the Holiday
Season? 422
Question 2.2: Is the Percentage of Sales Returned
Significantly Different in Arkansas Than the Rest of the
Country? 426
Question Set 3: Predictive Analytics 427
Question 3.1: By Looking at Line Charts for 2014 and
2015, Does the Average Percentage of Sales Returned
in 2014 Seem to Be Predictive of Returns in 2015? 427
Question 3.2: Using Regression, What Can We Predict
for Returns as a Percentage of Sales Based on
Historical Transactions? 429
Appendix A
Basic Statistics Tutorial 434
Appendix B
Accessing the Excel Data Analysis Toolpak 440
Appendix C
Excel (Formatting, Sorting, Filtering, and PivotTables)
442
page xxv
Appendix D
SQL Part 1 453
Appendix E
SQLite 466
Appendix F
Power Query 470
Appendix G
Tableau 476
Appendix H
SQL Part 2 480
Appendix I
Power BI 483
Appendix J
Dillard’s ER Diagram 491
Appendix K
Data Dictionaries 492
GLOSSARY 500
INDEX 504
page 2
Chapter 1
Data Analytics for Accounting and
Identifying the Questions
A Look at This Chapter

Data Analytics is changing the business world. In this chapter, we
define it and explain its impact on business and the accounting
profession, noting that the value of Data Analytics is in the insights it
provides. We also describe how to develop an analytics mindset. We
describe the Data Analytics Process using the IMPACT cycle model
and explain how this process is used to address both business and
accounting questions. We specifically emphasize the importance of
identifying appropriate questions that Data Analytics might be able to
address.
A Look Ahead
Chapter 2 provides a description of how data are prepared and
scrubbed to be ready for analysis to answer business questions. We
explain how to extract, transform, and load data and then how to
validate and normalize the data. In addition, we explain how data
standards are used to facilitate the exchange of data between
senders and receivers.
page 3
Shutterstock
The Chinese e-commerce company Alibaba is perhaps the biggest online

commerce company in the world. Using its three main websites, Taobao.com,
Tmall.com, and Alibaba.com, it hosts millions of businesses and hundreds of millions
of users with $345 billion in 2018 sales last year (more than eBay and Amazon
combined!). With so many transactions and so many users, Alibaba has worked to
capture fraud signals directly from its extensive database of user behaviors and its
network. It then analyzes them in real time using machine learning to accurately sort
the potentially fraudulent users from the good ones. Alibaba has developed five
stages of fraud detection for each user: (1) account check, (2) device check, (3)
activity check, (4) risk strategy, and (5) manual review. These stages all combine to
develop a risk score for each user. This fraud risk prevention score is so valuable to
Alibaba and others that Alibaba shares and sells it to external customers. What will
Data Analytics do next?
Sources: J. Chen, Y. Tao, H. Wang, and T. Chen, “Big Data Based Fraud Risk
Management at Alibaba,” Journal of Finance and Data Science 1, no. 1 (2015), pp.
1–10; and K. Pal, “How to Combat Financial Fraud by Using Big Data,” 2016,
http://www.kdnuggets.com/2016/03/combat-financial-fraud-using-big-data.html.
OBJECTIVES
After reading this chapter, you should be able to:
LO 1-1 Define Data Analytics.

LO 1-2 Understand why Data Analytics matters to business.
LO 1-3 Explain why Data Analytics matters to accountants.
LO 1-4 Describe the Data Analytics Process using the
IMPACT cycle.
LO 1-5 Describe the skills needed by accountants.
LO 1-6 Explain how to translate common business
questions into fields and values.
page 4
DATA ANALYTICS
LO 1-
1
Define
Data
Analytics
.
Data surrounds us! By the year 2020, about 1.7 megabytes of new
information will be created every second for every human being on
the planet. In fact, more data have been created in the last 2 years
than in the entire previous history of the human race.1 With so much
data available about each of us (i.e., how we shop, what we read,
what we’ve bought, what music we listen to, where we travel, whom
we trust, etc.), arguably, there is the potential for analyzing those
data in a way that can answer fundamental business questions and
create value.
We define Data Analytics as the process of evaluating data with
the purpose of drawing conclusions to address business questions.
Indeed, effective Data Analytics provides a way to search through
large structured and unstructured data to discover unknown patterns
or relationships.2 In other words, Data Analytics often involves the
technologies, systems, practices, methodologies, databases,
statistics, and applications used to analyze diverse business data to
give organizations the information they need to make sound and
timely business decisions.3 That is, the process of Data Analytics
aims to transform raw data into knowledge to create value.
Big Data refers to datasets that are too large and complex for
businesses’ existing systems to handle utilizing their traditional
capabilities to capture, store, manage, and analyze these datasets.
Another way to describe Big Data is by use of 3 Vs: its volume (the
sheer size of the dataset), velocity (the speed of data processing),
and variety (the number of types of data). While sometimes Data
Analytics and Big Data are terms used interchangeably, we will use
the term Data Analytics throughout and focus on the ability to turn
data into knowledge and knowledge into value.
PROGRESS CHECK
1. How does having more data around us translate into value
for a company?
2. Banks know a lot about us, but they have traditionally used
externally generated credit scores to assess
creditworthiness when deciding whether to extend a loan.
How would you suggest a bank use Data Analytics to get a
more complete view of its customers’ creditworthiness?
Assume the bank has access to a customer’s loan history,
credit card transactions, deposit history, and direct deposit
registration. How could it assess whether a loan might be
repaid?
HOW DATA ANALYTICS AFFECTS

BUSINESS
LO 1-
2
Understa
nd why
Data
Analytics
matters
to
business
.
There is little question that the impact of data analytics on business

is overwhelming. In fact, in PwC’s 18th Annual Global CEO Survey,
86 percent of chief executive officers (CEOs) say they find it
important to champion digital technologies and emphasize a clear
vision of using technology for a competitive advantage, while 85
percent say they put a high value on Data Analytics. In fact, per
PwC’s 6th Annual Digital IQ survey of more than 1,400 page 5
leaders from digital businesses, the area of investment
that tops CEOs’ list of priorities is business analytics.4
A recent study from McKinsey Global Institute estimates that
Data Analytics could generate up to $3 trillion in value per year in
just a subset of the total possible industries affected.5 Data Analytics
could very much transform the manner in which companies run their
businesses in the near future because the real value of data comes
from Data Analytics. With a wealth of data on their hands,
companies use Data Analytics to discover the various buying
patterns of their customers, investigate anomalies that were not
anticipated, forecast future possibilities, and so on. For example,
with insight provided through Data Analytics, companies could
execute more directed marketing campaigns based on patterns
observed in their data, giving them a competitive advantage over
companies that do not use this information to improve their
marketing strategies. Patterns discovered from past archives also
enable businesses to identify opportunities and risks and better plan
for the future. In addition to producing more value externally, studies
show that Data Analytics affects internal processes, improving
productivity, utilization, and growth.6
PROGRESS CHECK
3. Let’s assume a brand manager at Samsung identifies that
an older demographic might be concerned with the use of a
Samsung Galaxy smartphone and the radiation impact it
might have on the brain. How might Samsung use Data
Analytics to assess if this is a problem?
4. How might Data Analytics assess the higher cost of paying
employees to work overtime? Consider how Data Analytics
might be helpful in reducing a company’s overtime direct
labor costs in a manufacturing setting.
HOW DATA ANALYTICS AFFECTS

ACCOUNTING
LO 1-
3
Explain
why
Data
Analytics
matters
to
accounta
nts.
Data Analytics is expected to have dramatic effects on auditing and

financial reporting as well as tax and managerial accounting. We
detail how we think this might happen in each of the following
sections.
Auditing
Data Analytics plays an increasingly critical role in the future of audit.
In a recent Forbes Insights/KPMG report, “Audit 2020: A Focus on
Change,” the vast majority of survey respondents believe both that:
1. Audit must better embrace technology.
2. Technology will enhance the quality, transparency, and accuracy of
the audit.
Indeed, “As the business landscape for most organizations
becomes increasingly complex and fast-paced, there is a movement
toward leveraging advanced business analytic techniques to refine
the focus on risk and derive deeper insights into an page 6
7
organization.” Many auditors believe that auditor data
analytics will, in fact, lead to deeper insights that will enhance audit
quality. This sentiment of the impact of Data Analytics on the audit
has been growing for several years now and has given many public
accounting firms incentives to invest in technology and personnel to
capture, organize, and analyze financial statement data to provide
enhanced audits, expanded services, and added value to their
clients. As a result, Data Analytics is expected to be the next
innovation in the evolution of the audit and professional accounting
industry.
Given the fact that operational data abound and are easier to
collect and manage, combined with CEOs’ desires to utilize these
data, the accounting firms may now approach their engagements
with a different mindset. No longer will they be simply checking for
errors, material misstatements, fraud, and risk in financial statements
or merely be reporting their findings at the end of the engagement.
Instead, audit professionals will now be collecting and analyzing the
company’s data similar to the way a business analyst would to help
management make better business decisions. This means that, in
many cases, external auditors will stay engaged with clients beyond
the audit. This is a significant paradigm shift. The audit process will
be changed from a traditional process toward a more automated
one, which will allow audit professionals to focus more on the logic
and rationale behind data queries and less on the gathering of the
actual data.8 As a result, audits will not only yield important findings
from a financial perspective, but also information that can help
companies refine processes, improve efficiency, and anticipate future
problems.
“It’s a massive leap to go from traditional audit approaches to one that fully integrates
big data and analytics in a seamless manner.”9
Data Analytics also expands auditors’ capabilities in services like

testing for fraudulent transactions and automating compliance-
monitoring activities (like filing financial reports to the SEC or to the
IRS). This is possible because Data Analytics enables auditors to
analyze the complete dataset, rather than the sampling of the
financial data done in a traditional audit. Data Analytics enables
auditors to improve its risk assessment in both its substantive and
detailed testing.
Financial Reporting
Data Analytics also potentially has an impact on financial reporting.
With the use of so many estimates and valuations in financial
accounting, some believe that employing Data Analytics may
substantially improve the quality of the estimates and valuations.
Data from within an enterprise system and external to the company
and system might be used to address many of the questions that
face financial reporting. Many financial statement accounts are just
estimates and so accountants often ask themselves questions like
this to evaluate those estimates:
1. How much of the accounts receivable balance will ultimately be
collected? What should the allowance for loan losses look like?
2. Is any of our inventory obsolete? Should our inventory be valued at
market or cost (applying the lower-of-cost-or-market rule)? When
will it be out of date? Do we need to offer a discount on it now to get
it sold?
page 7
3. Has our goodwill been impaired due to the reduction in
profitability from a recent merger? Will it regain value in the near
future?
4. How should we value contingent liabilities like warranty claims or
litigation? Do we have the right amount?
Data Analytics may also allow an accountant or auditor to assess
the probability of a goodwill write-down, warranty claims or the
collectability of bad debts based on what customers, investors, and
other stakeholders are saying about the company in blogs and in
social media (like Facebook and Twitter). This information might help
the firm determine both its optimal response to the situation and
appropriate adjustment to its financial reporting.
It may be possible to use Data Analytics to scan the environment
—that is, scan Google searches and social media (such as
Instagram and Facebook) to identify potential risks and opportunities
to the firm. For example, in a data analytics sense, it may allow a
firm to monitor its competitors and its customers to better understand
opportunities and threats around it. For example, are its competitors,
customers, or suppliers facing financial difficulty, etc., that might
affect the company’s interactions with them and/or open up new
opportunities that otherwise it wouldn’t have considered?
Tax
Traditionally, tax work dealt with compliance issues based on data
from transactions that have already taken place. Now, however, tax
executives must develop sophisticated tax planning capabilities that
assist the company with minimizing its taxes in such a way to avoid
or prepare for a potential audit. This shift in focus makes tax data
analytics valuable for its ability to help tax staffs to predict what will
happen rather than reacting to what just did happen. Arguably, one
of the things that Data Analytics does best is predictive analytics—
predicting the future! An example of how tax data analytics might be
used is the capability to predict the potential tax consequences of a
potential international transaction, R&D investment, or proposed
merger or acquisition.
One of the issues of performing predictive Data Analytics is the
efficient organization and use of data stored across multiple systems
on varying platforms that were not originally designed for the tax
department. Organizing tax data into a data warehouse to be able to
consistently model and query the data is an important step toward
developing the capability to perform tax data analytics. This issue is
exemplified by the 29 percent of tax departments that find the
biggest challenge in executing an analytics strategy is integration
with the IT department and available technology tools.10
PROGRESS CHECK
5. How could the use of internal audit data analytics find the
pattern that one accountant enters the majority of the
journal entries each quarter? How might this data be used
to check if segregation of duties was appropriately
maintained? Why might this be an issue that would need
addressing?
6. How specifically will Data Analytics change the way a tax
staff does its taxes?
page 8
THE DATA ANALYTICS PROCESS

USING THE IMPACT CYCLE
LO 1-
4
Describe
the Data
Analytics
Process
using the
IMPACT
cycle.
Data Analytics is a process to identify business questions and

problems that can be addressed with data. We start to describe our
Data Analytics Process by using an established Data Analytics
model called the IMPACT cycle by Isson and Harriott (as shown in
Exhibit 1-1).
EXHIBIT 1-1
The IMPACT Cycle
Source: Isson, J. P., and J. S. Harriott. Win with Advanced Business Analytics: Creating
Business Value from Your Data. Hoboken, NJ: Wiley, 2013.
We explain the full IMPACT cycle briefly here, but in more detail
in later in Chapters 2, 3, and 4. We use its approach throughout this
textbook.
Step 1: Identify the Questions (Chapter 1)
It all begins with understanding a business problem that needs
addressing. Questions can arise from many sources, including how
to better attract customers, how to price a product, or how to find
errors or fraud. Having a concrete, specific question that is
potentially answerable by Data Analytics is an important first step.
Accountants and auditors might be interested in questions like
the following:
Are employees circumventing internal controls over payments?
Are there any suspicious travel and entertainment expenses?
How can we increase the amount of add-on sales of additional
goods to our customers?
Are our customers paying us in a timely manner?
How can we predict the allowance for loan losses for our bank
loans?
How can we find transactions that are risky in terms of accounting
issues?
Who authorizes checks above $100,000?
How can errors be identified?
Step 2: Master the Data (Chapter 2)
Mastering the data requires one to know what data are available and
whether those data might be able to help address the business
problem. We need to know everything about the data, including how
to access, availability, reliability (if there are errors), and what time
periods are covered to make sure the data coincide with the timing of
our business problem, etc.
page 9
In addition, to give us some idea of the data questions, we may

want to consider the following:
Review data availability in a firm’s internal systems (including those
in the financial reporting system or enterprise systems that might
occur in its accounting processes—financial, procure-to-pay,
production, order-to-cash, human resources).
Review data availability in a firm’s external network, including those
that might already be housed in an existing data warehouse.
Data dictionaries and other contextual data—to provide details
about the data.
Extraction, transformation, and loading.
Data validation and completeness—to provide a sense of the
reliability of the data.
Data normalization—to reduce data redundancy and improve data
integrity.
Data preparation and scrubbing—Data Analytics professionals
estimate that they spend between 50 and 90 percent of their time
cleaning data so the data can be analyzed.11
Step 3: Perform Test Plan (Chapter 3)
After mastering the data and after the data are ready (in step 2), we
are prepared for analysis. With the data ready for analysis, we need
to think of the right approach to the data to be able to answer the
question.
In Data Analytics, we work to extract knowledge from the data to
address questions and problems. Using all available data, we see if
we can identify a relationship between the response or dependent
variables and those items that affect the response (also called
predictor, explanatory, or independent variables). To do so, we’ll
generally make a model, or a simplified representation of reality, to
address this purpose.
An example might be helpful here. Let’s say we are trying to
predict each of your classmates’ performance on their next
intermediate accounting exam. The response or dependent variable
will be the score on the next exam. What helps predict the
performance of each exam will be our predictor, explanatory, or
independent variables. Variables such as study time, score on last
exam, IQ, and standardized test scores (ACT, SAT, etc.), as well as
student enjoyment of accounting, might all be considered. Perhaps
given your experience you can name other predictor variables to
include in our model predicting exam performance.
The research question, the model, the data availability, and the
expected statistical inference may all suggest the use of different
data approaches. Provost and Fawcett12 detail eight different
approaches to data analytics depending on the question. We will
discuss the most applicable ones to accounting more formally in
Chapter 3 and highlight accounting questions that they might
address. The eight different approaches include the following:
Classification—An attempt to assign each unit (or individual) in a
population into a few categories. An example classification might
be, of all the loans this bank has offered, which are most likely to
default? Or which loan applications are expected to be approved?
Or which transactions would a credit card company flag as
potentially being fraudulent and deny payment?
Regression—A data approach used to predict a specific dependent
variable value based on independent variable inputs using a
statistical model. An example regression analysis might be, given a
balance of total accounts receivable held by a firm, what is the
appropriate level of allowance for doubtful accounts for bad debts?
page 10
Similarity matching—An attempt to identify similar
individuals based on data known about them. The opening vignette
mentioned Alibaba and its attempt to identify seller and customer
fraud based on various characteristics known about them to see if
they were similar to known fraud cases.
Clustering—An attempt to divide individuals (like customers) into
groups (or clusters) in a useful or meaningful way. In other words,
identifying groups of similar data elements and the underlying
drivers of those groups. For example, clustering might be used to
segment a customer into a small number of groups for additional
analysis and marketing activities.
Co-occurrence grouping—An attempt to discover associations
between individuals based on transactions involving them. Amazon
might use this to sell another item to you by knowing what items are
“frequently bought together” or “Customers who bought this item
also bought . . .” as shown in Exhibit 1-2.
EXHIBIT 1-2
Example of Co-occurrence Grouping on Amazon.com
Source: Amazon Inc.
Profiling—An attempt to characterize the “typical” behavior of an
individual, group, or population by generating summary statistics
about the data (including mean, standard deviations, etc.). By
understanding the typical behavior, we’ll be able to more easily
identify abnormal behavior. When behavior departs from that typical
behavior—which we’ll call an anomaly—then further investigation is
warranted. Profiling might be used in accounting to identify fraud or
just those transactions that might warrant some additional
investigation (e.g., travel expenses that are three standard
deviations above the norm).
Link prediction—An attempt to predict a relationship between two
data items. This might be used in social media. For example,
because an individual might have 22 mutual Facebook friends with
me and we both attended Brigham Young University, is there a
chance we would like to be Facebook friends as well? Exhibit 1-3
provides an example of this used in Facebook. Link prediction in an
accounting setting might work to use social media to look for
relationships between related parties that are not otherwise
disclosed.
EXHIBIT 1-3
Example of Link Prediction on Facebook
Source: Facebook Inc.; Exactostock/SuperStock (Ana); Michael
DeLeon/Getty Images (Diego); Daniel Ernst/Getty Images (Leslie);
Sam Edwards/Glow Images (Daniel)
Data reduction —A data approach that attempts to reduce the

amount of information that needs to be considered to focus on the
most critical items (i.e., highest cost, highest risk, largest page 11
impact, etc.). It does this by taking a large set of data
(perhaps the population) and reducing it with a smaller set that has
the vast majority of the critical information of the larger set. An
example might include the potential to use these techniques in
auditing. While auditing has employed various random and stratified
sampling over the years, Data Analytics suggests new ways to
highlight which transactions do not need the same level of vetting
as other transactions.
Step 4: Address and Refine Results (Chapter
3)
After the data have been analyzed (in step 3 of the IMPACT cycle),
the fourth step is to address and refine results. Data analysis is
iterative. We slice and dice the data, find correlations, ask ourselves
further questions, ask colleagues what they think, and revise and
rerun the analysis. But once that is complete, we have the results
ready to communicate to interested stakeholders.
Steps 5 and 6: Communicate Insights and
Track Outcomes (Chapter 4 and each chapter
thereafter)
Once the results have been determined (in step 4 of the IMPACT
cycle), insights are formed by decision makers and are
communicated (the “C” in the IMPACT cycle) and some outcomes
will be continuously tracked (the “T” in the IMPACT cycle).
Chapter 4 discusses ways to communicate results, including the
use of static reports, digital dashboards, and data visualizations.
Data Analytics is especially interested in reporting results that help
decision makers see the data in an all-new way to develop insights
that help answer business questions. Digital dashboards and data
visualizations are particularly helpful in communicating results.
page 12
Back to Step 1
Of course, the IMPACT cycle is iterative, so once insights are gained
and outcomes are tracked, new questions emerge and the IMPACT
cycle begins anew.
PROGRESS CHECK
7. Let’s say we are trying to predict how much money college
students spend on fast food each week. What would be the
response, or dependent, variable? What would be
examples of independent variables?
8. How might a data reduction approach be used in auditing to
spend time and effort on the most important items?
DATA ANALYTIC SKILLS NEEDED BY
ANALYTIC-MINDED ACCOUNTANTS
LO 1-
5
Describe
the skills
needed
by
accounta
nts.
While we don’t believe that accountants need to become data

scientists—they may never need to build a data repository or do the
real, hardcore Data Analytics—they must know how to do the
following:
Clearly articulate the business problem the company is facing.
Communicate with the data scientists about specific data needs and
understand the underlying quality of the data.
Draw appropriate conclusions to the business problem based on the
data and make recommendations on a timely basis.
Present their results to individual members of management (CEOs,
audit managers, etc.) in an accessible manner to each member.
Consistent with that, in this text, we emphasize seven skills that
analytic-minded accountants should have:
1. Develop an analytics mindset—recognize when and how data
analytics can address business questions.
2. Data scrubbing and data preparation—comprehend the process
needed to clean and prepare the data before analysis.
3. Data quality—recognize what is meant by data quality, be it
completeness, reliability, or validity.
4. Descriptive data analysis—perform basic analysis to understand the
quality of the underlying data and its ability to address the business
question.
5. Data analysis through data manipulation—demonstrate ability to
sort, rearrange, merge and reconfigure data in a manner that allows
enhanced analysis.
6. Define and address problems through statistical data analysis—
identify and implement an approach that will use statistical data
analysis to draw conclusions and make recommendations on a
timely basis.
7. Data visualization and data reporting—report results of analysis in
an accessible way to each varied decision maker and his or her
specific needs.
We address these seven skills throughout the first four chapters
in the text in hopes that the analytic-minded accountant will develop
and practice these skills to be ready to address business questions.
We then demonstrate these skills in the labs and hands-on analysis
throughout the rest of the book.
page 13
HANDS-ON EXAMPLE OF THE

IMPACT MODEL
LO 1-
6
Explain
how to
translate
common
business
question
s into
fields
and
values.
Here we provide a complete, hands-on example of the IMPACT

model to show how it could be implemented for a specific situation.
Let’s suppose I am trying to get a loan to pay off some credit card
debt and my friend has told me about a new source of funds that
doesn’t involve a bank. In recent years, facilitated by the Internet,
peer-to-peer lenders allow individuals to both borrow and lend
money to each other. While there are other peer-to-peer lenders, in
this case, we will specifically consider the LendingClub.
My question is whether I will be able to get a loan, given my prior
loan history (poor), credit score, and the like. According to our
approaches mentioned above, this would be an example of a
classification approach because we are attempting to predict
whether a person applying for a loan will be approved and funded or
whether she will be denied a loan.
Identify the Questions

Stated specifically, our question is, “Given my borrower profile, can I
expect the LendingClub to extend a loan to me?”
Master the Data

LendingClub is a U.S.-based, peer-to-peer lending company,
headquartered in San Francisco, California. LendingClub facilitates
both borrowing and lending by providing a platform for unsecured
personal loans between $1,000 and $35,000. The loan period is for
either 3 or 5 years. There is information available that allows
potential investors to search and browse the loan listings on the
LendingClub website and select loans in which they would like to
invest. The available information includes information supplied about
the borrower, amount of the loan, loan grade (and related loan
interest rate), and loan purpose. Investors invest in the loans and
make money from interest. LendingClub makes money by charging
borrowers an origination fee and investors a service fee. Since 2007,
hundreds of thousands of borrowers have obtained more than $44
billion in loans via LendingClub.13
Some basic lending statistics are included on the LendingClub
Statistics website (Exhibit 1-4). Each bar represents the volume of
loans each quarter during its respective year.
EXHIBIT 1-4
LendingClub Statistics
Source: Accessed March, 2019. https://www.lendingclub.com/info/statistics.action.
page 14
Borrowers borrow money for a variety of reasons, including

refinancing other debt and paying off credit cards, as well as
borrowing for other purposes (Exhibit 1-5).
EXHIBIT 1-5
LendingClub Statistics by Reported Loan Purpose
68.18% of LendingClub borrowers report using their loans to
refinance existing loans or pay off their credit cards as of 12/31/18.
Source: Accessed March, 2019. https://www.lendingclub.com/info/statistics.action.
LendingClub actually provides datasets: data on the loans they
approved and funded as well as data for the loans that were
declined. In this chapter, we will emphasize the rejected loans and
the reasons why they were rejected.
As we learn about the data, it is important to know what is
available to us. To that end, there is a data dictionary that provides
descriptions for all of the data attributes of the dataset. A cut-out of
the data dictionary for the rejected stats file (i.e., the statistics about
those loans rejected) is shown in Exhibit 1-6.
EXHIBIT 1-6
2007–2012 LendingClub Data Dictionary for Declined Loan Data
Source: Accessed March, 2019.
RejectStats
Description
File
Amount Total requested loan amount

Requested
Application Date of borrower application

Date
RejectStats
Description
File
Loan Title Loan title
Risk_Score Borrower risk (FICO) score
Dept-To- Ratio of borrower total monthly debt payments

Income Ratio divided by monthly income.
Zip Code The first 3 numbers of the borrower zip code

provided from loan application.
State Two digit State Abbreviation provided from loan

application.
Employment Employment length in years, where 0 is less than

Length 1 and 10 is greater than 10.
Policy Code policy_code= 1 if publicly available.

policy_code=2 if not publicly available
We could also take a look at the data files available for the
funded loan data. However, for our analysis in the rest of this
chapter, we use the Excel file “RejectStatsA Ready,” page 15
which has rejected loan statistics from 2007 to 2012.
It is a cleaned-up, transformed file ready for analysis. We’ll learn
more about data scrubbing in Chapter 2.
Exhibit 1-7 provides a cut-out of the 2007–2012 “Approved Loan”
dataset provided.
EXHIBIT 1-7
2007–2012 Declined Loan Applications (RejectStatsA) Dataset
Source: Microsoft Excel, 2016.
Amount Application Risk

Loan Title
Requested Date Score
2175 12/19/2012 major_purchase 850
35000 8/13/2012 other 850
10000 9/19/2012 major_purchase 850
10000 11/9/2012 car 850
3000 11/27/2012 vacation 850
5000 5/20/2012 Lower Rate 850
20000 9/8/2012 Home loan 850
Loan is for new
8000 10/22/2012 850
kitch
18500 7/19/2012 bussiness loan 850
10000 7/11/2012 car 850
25000 10/6/2010 debt_consolidation 849
1000 9/9/2012 Hospital expenses 849
35000 5/26/2012 smalljbusiness 849
6800 7/13/2012 Be my Own Boss 849
25000 8/10/2012 home_improvement 849
35000 12/2/2012 debt_consolidation 848
1500 11/13/2012 other 848
Perform Test Plan

Considering our question, “Will I receive a loan from LendingClub?”
and the available data, we will do three analyses to predict whether
we will receive a loan, including:
1. The debt-to-income ratios and number of rejected loans.
2. The length of employment and number of rejected loans.
3. The credit (or risk) score and number of rejected loans.
Because LendingClub collects this information, we believe it will
give LendingClub an idea if the borrower will be able to pay back
the loan and give us an idea if our loan will be approved or rejected.
The first analysis we perform considers the debt-to-income ratio.
That is, how big is the debt compared to the size of the annual
income earned?
To consider the debt-to-income ratio in our analysis, three
buckets (labeled DTI bucket) are constructed for each grouping of
the debt-to-income ratio. These three buckets include the following:
1. High (debt is greater than 20 percent of income).
2. Medium (“mid”) (debt is between 10 and 20 percent of income).
3. Low (debt is less than 10 percent of income).
Once those buckets are constructed, we are ready to analyze the
breakdown of rejected loan applications by the debt-to-income ratio.
The Excel PivotTable is an easy way to make comparisons
between the different levels of DTI. When we run a PivotTable
analysis, we highlight the loans, which count the number of loans
applied for and rejected, and the DTI bucket (see Exhibit 1-8). The
PivotTable counts the number of loan applications in each of the
three DTI buckets: high, medium (mid), and low. This suggests that
because the high DTI bucket has the highest number of loan
applications, perhaps the applicant asked for a loan page 16
that was too big given his or her income.
LendingClub might have seen that as too big of a risk and chosen
to not extend the loan to the borrower using the debt-to-income ratio
as an indicator.
EXHIBIT 1-8
LendingClub Declined Loan Applications by DTI (Debt-to-Income)
DTI bucket includes high (debt > 20 percent of income), medium
(“mid”) (debt between 10 and 20 percent of income), and low (debt <
10 percent of income). (PivotTable shown here required manually
sorting rows to get in proper order.)
The second analysis was on the length of employment and its

relationship with rejected loans (see Exhibit 1-9). Arguably, the
longer the employment, the more stable of a job and income stream
you will have to ultimately repay the loan. LendingClub reports the
number of years for each of the rejected applications. The PivotTable
analysis lists the number of loans by the length of employment.
Almost 77 percent (495,109 out of 645,414) out of the total rejected
loans had worked at a job for less than 1 year, suggesting potentially
an important reason for rejecting the requested loan. Perhaps some
had worked a week, or just a month, and still want a big loan?
EXHIBIT 1-9
LendingClub Declined Loan Applications by Employment Length (Years of
Experience)

page 17
The third analysis we perform is to consider the credit or risk

score of the applicant. As noted in Exhibit 1-10, risk scores are
typically classified in this way with those in the excellent and very
good category receiving the lowest possible interest rates and best
terms with a credit score above 750. On the other end of the
spectrum are those with very bad credit (with a credit score less than
600).
EXHIBIT 1-10
Breakdown of Customer Credit Scores (or Risk Scores)
Source: Cafecredit.com.
Another predictor of loan repayment is the credit score that the
borrower has. We classify the sample according to this breakdown
into excellent, very good, good, fair, poor, and very bad credit
according to their credit score noted in Exhibit 1-10.
Address and Refine Results

After performing a PivotTable analysis (as seen in Exhibit 1-11), we
count the number of rejected loan applications by credit (risk) score.
We’ll note in the rejected loans that nearly 82 percent [(167,379 +
151,716 + 207,234)/645,414] of the applicants have either very bad,
poor, or fair credit ratings, suggesting this might be a good reason for
a loan rejection. We also note that only 0.3 percent (2,494/645,414)
of those rejected loan applications had excellent credit.
page 18
EXHIBIT 1-11
The Count of LendingClub Rejected Loan Applications by Credit or Risk Score
(PivotTable shown here required
Classification Using PivotTable Analysis
manually sorting rows to get in proper order.)
So, if these are the applications that were all rejected, the
question is how many of these that might apply for a loan not only
had excellent credit, but also had worked more than 10 years and
had asked for a loan that was less than 10 percent of their income
(in the low DTI bucket)? Use of a PivotTable (as shown in Exhibit 1-
12) allows us to consider this three-way interaction and provides an
answer of 365 out of 645,414 (0.057 percent of the total). This might
suggest that the use of these three metrics is reasonable at
predicting loan rejection because the number who have excellent
credit, worked more than 10 years, and requested a loan that was
less than 10 percent of their income was such a small percentage of
the total.
EXHIBIT 1-12
The Count of LendingClub Declined Loan Applications by Credit Score, Debt-to-
Income, and Employment Length Using PivotTable Analysis (highlighting added)
page 19
Perhaps those with excellent credit just asked for too big of a
loan given their existing debt and that is why they were rejected.
Exhibit 1-13 shows the PivotTable analysis. The analysis shows
those with excellent credit asked for a larger loan (16.2 percent of
income) given the debt they already had as compared to any of the
others, suggesting a reason even those potential borrowers with
excellent credit were rejected.
EXHIBIT 1-13
The Average Debt-to-Income Ratio (shown as a percentage) by Credit (Risk) Score for
LendingClub Declined Loan Applications Using PivotTable Analysis
Communicate Insights
Certainly further and more sophisticated analysis could be
performed, but at this point we have a pretty good idea of what
LendingClub uses to decide whether to extend a loan. We can
communicate these insights either by showing the PivotTables or
stating what three of the determinants are.
Track Outcomes
There are a wide variety of outcomes that could be tracked. But in
this case, it might be best to see if we could predict future outcomes.
For example, the data we analyzed was from 2007–2012. We could
make our predictions for subsequent years based on what we had
found in the past and then test and see how accurate we are with
those predictions. We could also change our prediction model when
we learn new insights and additional data become available.
In this chapter, we discussed how businesses and accountants
derive value from Data Analytics. We gave some specific examples
of how Data Analytics is used in business, auditing, managerial
accounting, financial accounting, and tax accounting.
We introduced the IMPACT model and explained how it is used.
And then we talked specifically about the importance of identifying
the question. We walked through the first few steps of the IMPACT
model and introduced eight data approaches. We also discussed the
data analytic skills needed by analytic-minded accountants.
We followed this up by looking at the case of why LendingClub
rejected loans for a set of its customers using the IMPACT model.
We performed this analysis using various filtering and PivotTable
tasks.
page 20
PROGRESS CHECK
9. Doing your own analysis, download the rejected loans
dataset titled “RejectStatsA Ready” and perform an Excel
PivotTable analysis by state and figure out the number of
rejected applications for the state of California. That is,
count the loans by state and see what percentage of the
rejected loans came from California. How close is that to
the relative proportion of the population of California as
compared to that of the United States?
10. Doing your own analysis, download the rejected loans
dataset titled “RejectStatsA Ready” and run an Excel
PivotTable by risk (or credit) score classification and DTI
bucket to determine the number of rejected loans requested
by those rated as having an excellent credit score.
Summary
With data all around us, businesses and accountants
are looking at Data Analytics to extract the value that the
data might possess.
Data Analytics is changing the audit and the way that
accountants look for risk. Now, auditors can consider
100 percent of the transactions in their audit testing. It is
also helpful in finding anomalous or unusual
transactions. Data Analytics is also changing the way
financial accounting, managerial accounting, and taxes
are done at a company.
The IMPACT cycle is a means of doing Data Analytics
that goes all the way from identifying the question, to
mastering the data, to performing data analyses and
communicating results. It is recursive in nature,
suggesting that as questions are addressed, new
important questions may emerge that can be addressed
in a similar way.
Eight data approaches address different ways of testing
the data: classification, regression, similarity matching,
clustering, co-occurrence grouping, profiling, link
prediction, and data reduction. These are explained in
more detail in Chapter 3.
Data analytic skills needed by analytic-minded
accountants are specified and are consistent with the
IMPACT cycle, including the following:
Develop an analytics mindset.

Data scrubbing and data preparation.
Data quality.
Descriptive data analysis.
Data analysis through data manipulation.
Define and address problems through statistical data
analysis.
Data visualization and data reporting.
Key Words
Big Data (4) Datasets that are too large and complex
for businesses’ existing systems to handle utilizing their
traditional capabilities to capture, store, manage, and
analyze these datasets.
classification (9) A data approach that attempts to

assign each unit in a population into a few categories
potentially to help with predictions.
clustering (10) A data approach that attempts to divide

individuals (like customers) into groups (or clusters) in a
useful or meaningful way.
page 21
co-occurrence grouping (10) A data approach that

attempts to discover associations between individuals
based on transactions involving them.
Data Analytics (4) The process of evaluating data with

the purpose of drawing conclusions to address business
questions. Indeed, effective Data Analytics provides a way
to search through large structured and unstructured data
to identify unknown patterns or relationships.
data dictionary (14) Centralized repository of

descriptions for all of the data attributes of the dataset.
data reduction (10) A data approach that attempts to
reduce the amount of information that needs to be
considered to focus on the most critical items (i.e., highest
cost, highest risk, largest impact, etc.)
link prediction (10) A data approach that attempts to

predict a relationship between two data items.
profiling (10) A data approach that attempts to

characterize the “typical” behavior of an individual, group
or population by generating summary statistics about the
data (including mean, standard deviations, etc.).
predictor (or independent or explanatory) variable

(9) A variable that predicts or explains another variable.
response (or dependent) variable (9) A variable that

responds to, or is dependent, on another.
regression (9) A data approach used to predict a

specific dependent variable value based on independent
variable inputs using a statistical model.
similarity matching (10) A data approach that

attempts to identify similar individuals based on data
known about them.
ANSWERS TO PROGRESS
CHECKS
1. The plethora of data alone does not necessarily
translate into value. However, if we carefully use the
data to help address critical business problems and
questions, the data may create value.
2. Banks could certainly use credit scores from companies
like Experian, TransUnion, and Equifax, but if they
have access to all of the banking information of their
clients, arguably they could make more informed
decisions. Banks would know how much money they
have and how they spend it. Banks would know if they
had prior loans and if they were paid in a timely
manner. Banks would know where they work and their
monthly income via the direct deposits. All of these
combined, in addition to a credit score, might be used
to assess creditworthiness to gain a better evaluation of
customers’ creditworthiness when they would like a
loan. It might also give us needed information for a
marketing campaign to target potential creditworthy
customers.
3. The brand manager at Samsung might use Data
Analytics to see what is being said about Samsung’s
phones on social media websites (e.g., Snapchat,
Instagram, and Facebook), particularly those that
attract an older demographic. This will help the
manager assess if there is a problem with the
perceptions of its phones.
4. Data Analytics might be used to collect information on
the amount of overtime. Who worked overtime? What
were they working on? Do we actually need more full-
time employees to reduce the level of overtime (and its
related costs to the company and to the employees)?
All of these questions could be addressed by looking at
recent records explaining the use of overtime records.
5. Data Analytics could tabulate the number of journal
entries by an accountant to see who entered the most
journal entries. This might be an issue if there was a
perception of a problem in risk, such as segregation of
duties in having one person enter so many journal
entries or just how the accounting workload is
distributed across accounting staff.
page 22
6. The tax staff would become much more
adept at efficiently organizing data from multiple
systems across an organization and performing Data
Analytics to help with tax planning to structure
transactions in a way that might minimize taxes.
7. The dependent variable could be the amount of money
spent on fast food. Independent variables could be
proximity of the fast food, ability to cook own food,
discretionary income, socioeconomic status, etc.
8. The data reduction approach might help auditors spend
more time and effort on the riskiest transactions or on
those that might be anomalous in nature. This will help
them more efficiently spend their time on items that may
well be of highest importance.
9. An analysis of the rejected loans suggests that 85,793
of the total 645,414 rejected loans were from the state
of California. That represents 13.29 percent of the total
rejected loans. This is greater than the relative
population of California to the United States as of the
2010 census, of 12.1 percent
(37,253,956/308,745,538).
10. A PivotTable analysis of the rejected loans suggests

that more than 30.6 percent (762/2,494) of those in the
excellent risk/credit score range asked for a loan with a
debt-to-income ratio of more than 20 percent.
page 23
1. Big Data is often described by the three Vs, or
a. volume, velocity, and variability.
b. volume, velocity, and variety.
c. volume, volatility, and variability.
d. variability, velocity, and variety.
2. Which approach to Data Analytics attempts to assign
each unit in a population into a small set of classes (or
groups) where the unit best fits?
a. Regression
b. Similarity matching
c. Co-occurrence grouping
d. Classification
3. Which approach to Data Analytics attempts to identify
similar individuals based on data known about them?
a. Classification
b. Regression
c. Similarity matching
d. Data reduction
4. Which approach to Data Analytics attempts to predict
relationship between two data items?
a. Profiling
b. Classification
c. Link prediction
d. Regression
5. Which of these terms is defined as being a central
repository of descriptions for all of the data attributes of
the dataset?
a. Big Data
b. Data warehouse
c. Data dictionary
d. Data Analytics
6. Which skills were not emphasized that analytic-minded
accountants should have?
a. Develop an analytics mindset
b. Data scrubbing and data preparation
c. Classification of test approaches
d. Define and address problems through statistical data
analysis
7. Which skills were not emphasized that analytic-minded
accountants should have?
a. Data quality
b. Descriptive data analysis
c. Data visualization
d. Data and systems analysis and design
8. The IMPACT cycle includes all except the following
process:
a. perform test plan.
b. visualize the data.
c. master the data.
d. track outcomes.
page 24
9. The IMPACT cycle includes all except the
following process:
a. data preparation.
b. communicate insights.
c. address and refine results.
d. perform test plan.
10. By the year 2020, about 1.7 megabytes of new
information will be created every:
a. week.
b. second.
c. minute.
d. day.
1. Define Data Analytics and explain how a university
might use its techniques to recruit and attract potential
students.
2. Give an example of how Data Analytics creates value
for businesses.
3. Give an example of how Data Analytics creates value
for accounting.
4. How might Data Analytics be used in financial
reporting? And how might it be used in doing tax
planning?
5. Describe the IMPACT cycle. Why does its order of the
processes and its recursive nature make sense?
6. Why is identifying the question such a critical first step
in the IMPACT process cycle?
7. What is included in mastering the data as part of the
IMPACT cycle described in the chapter?
8. In the chapter, we mentioned eight different data
approaches. Which data approach was used by
Alibaba, as mentioned in the chapter-opening vignette?
9. What data approach mentioned in the chapter might be
used by Facebook to find friends?
10. Auditors will frequently use the data reduction approach
when considering potentially risky transactions. Provide
an example of why focusing on a portion of the total
number of transactions might be important for auditors
to assess risk.
11. Which data approach might be used to assess the
appropriate level of the allowance for doubtful
accounts?
12. Why might the debt-to-income attribute included in the
declined loans dataset considered in the chapter be a
predictor of declined loans? How about the credit (risk)
score?
13. To address the question “Will I receive a loan from
LendingClub?” we had available data to assess the
relationship among (1) the debt-to-income ratios and
number of rejected loans, (2) the length of employment
and number of rejected loans, and (3) the credit (or risk)
score and number of rejected loans. What additional
data would you recommend to further assess whether a
loan would be offered? Why would it be helpful?
Problems
1. Navigate to the Additional Student Resources page on
Connect. Under Chapter 1 Data Files, download and
consider the LendingClub data dictionary file
“LCDataDictionary” specifically the LoanStats tab. This
represents the data dictionary for the loans that were
funded. Choosing some of the data attributes listed
there, which attributes do you think might predict which
loans will go delinquent and which will ultimately be fully
repaid? How could we test that?
page 25
2. Download and consider the rejected loans
dataset of LendingClub data titled “RejectStatsA
Ready.” Given the analysis performed in the chapter,
what three items do you believe would be most useful
in predicting loan acceptance or rejection? What
additional data do you think could be solicited either
internally or externally that would help you predict loan
acceptance or rejection?
3. Download the rejected loans dataset of LendingClub
data titled “RejectStatsA Ready” from the Connect
website and do an Excel PivotTable by state; then
figure out the number of rejected applications for the
state of Arkansas. That is, count the loans by state and
compute the percentage of the total rejected loans in
the USA that came from Arkansas. How close is that to
the relative proportion of the population of Arkansas as
compared to the overall U.S. population (per 2010
census)?
4. Download the rejected loans dataset of LendingClub
data titled “RejectStatsA Ready” from the Connect
website and do an Excel PivotTable by state; then
figure out the number of rejected applications for each
state. Reorder these and make a graph ordering the
states and the number of rejected loans from highest to
lowest. Is there a lot of variability among states?
For Problems 5, 6, and 7, we will be cleaning a data
file in preparation for subsequent analysis.
The analysis performed on LendingClub data in the
chapter was for the years 2007–2012. For this and
subsequent problems, please download the declined
loans table for 2013–2014 from the Connect website.
5. Consider the 2013 declined loan data from
LendingClub titled “RejectStatsB2013” from the
Connect website. Similar to the analysis done in the
chapter, let’s scrub the risk score data. First, because
our analysis requires risk scores, debt-to-income data,
and employment length, we need to make sure each of
them has valid data.
a. Open the file in Excel.
b. Sort the file based on risk score and remove those
observations (the complete row or record) that have a
missing score or a score of zero, if needed.
c. Assign each risk score to a risk score bucket similar to
the chapter. That is, classify the sample according to
this breakdown into excellent, very good, good, fair,
poor, and very bad credit according to their credit
score noted in Exhibit 1-10. Classify those with a
score greater than 850 as “Excellent.” Consider using
nested if–then statements to complete this. Or sort by
risk score and manually input into appropriate risk
score buckets.
d. Run a PivotTable analysis that shows the number of
loans in each risk score bucket. Which group had the
most rejected loans (biggest count)? Which group had
the least rejected loans (smallest count)? This is the
deliverable. Is it similar to Exhibit 1-11 performed on
years 2007–2012?
LendingClub titled “RejectStatsB2013.” Similar to the
analysis done in the chapter, let’s scrub the debt-to-
income data. Because our analysis requires risk scores,
debt-to-income data, and employment length, we need
to make sure each of them has valid data.
a. Sort the file based on debt-to-income and remove
those observations (the complete row or record) that
have a missing score, a score of zero, or a negative
score.
b. Assign each valid debt-to-income ratio into three
buckets (labeled DTI bucket) by classifying each debt-
to-income ratio into high (>20.0 percent), medium
(10.0–20.0 percent), and low (<10.0 percent) buckets.
Consider using nested if-then statements to complete
this. Or sort the row and manually input.
c. Run a PivotTable analysis that shows the number of
loans in each DTI bucket. Any interpretation of why
these loans were declined based on debt-to-income
ratios?
LendingClub titled “RejectStatsB2013.” Similar to the
analysis done in the chapter, let’s scrub the
employment length. Because our analysis requires risk
scores, debt-to-income data, and employment length,
we need to make sure each of them has valid data.
page 26
a. Sort the file based on employment length
and remove those observations (the complete row or
record) that have a missing score (“NA”). Note that we
are including the employment lengths of zero, different
than the analysis in the chapter text.
b. Sort the file based on debt-to-income and remove
those observations (the complete row or record) that
have a missing score, a score of zero, or a negative
score, similar to that done in Problem 1-6.
c. Sort the file based on risk score and remove those
observations (the complete row or record) that have a
missing score or a score of zero, similar to that done
in Problem 1-5.
d. There should now be 669,993 observations. Any
thoughts on what biases are imposed when we
remove observations? Is there another way to do this?
e. Run a PivotTable analysis to show the number of
Excellent Risk Scores but High DTI Bucket loans in
each Employment year bucket. Any interpretation of
why these loans were declined?
page 27
Lab 1-0 How to Complete Labs in

This Text
The labs in this book will provide valuable hands-on
experience in generating and analyzing accounting
problems. Each lab will provide a company summary with
relevant facts, techniques that you will use to complete
your analysis, software that you’ll need, and an overview
of the lab steps.
When you’ve completed your lab, you will submit a lab
report showing your thought process with written
responses and validating that you’ve completed specific
checkpoints by taking screenshots along the way. This lab
will demonstrate how to use basic lab tools.
In this lab, you will:
Part 1: Create a Word document on OneDrive.

Part 2: Take a screenshot of your document.
Part 3: Add another screenshot and submit your
document.
Submit two screenshots.
Part 1: Create a New Word Document

on OneDrive
On Office.com
1. Open your web browser and go to www.office.com.
2. Click All Microsoft and then click OneDrive and log in
using your university or personal email address and
password.
3. Click + New > OneDrive Word document. A new
window will open with a new blank document.
4. Type “Lab 1-0 Data Analytics Lab Overview [Your
name] [Your university email address]” in the first line
(e.g., Lab 1-0 Data Analytics Lab Overview Ryan Teeter
rteeter@pitt.edu).
5. Click File> Save As > Save As and name the
document “Lab 1-0 Data Analytics Lab Overview Ryan
Teeter rteeter@pitt.edu.” (You can also click the
document name in the title bar (e.g., Document2) and
change it there.
6. Because your document is in the cloud, changes are
saved automatically and you won’t lose your document
when you log out of a lab computer.
7. Keep your document open and go to the next part of
the lab.
Part 2: Take a Screenshot of Your

Document
In Windows
1. Click the Start button and Search for “Snipping Tool.”
2. Click New (Rectangular Snip) and draw a rectangle
across your screen that includes your entire window.
3. A preview window with your screenshot will appear.
4. Press Ctrl + C to copy your screenshot.
5. Go to your Word document and press Ctrl + V to paste
the screenshot into your document.
6. Keep your document open and go the next part of the
lab.
On a Mac
1. Press Cmd + Shift + 4 and draw a rectangle across
your screen that includes your entire window.
2. Your screenshot will be saved in your Desktop folder.
3. Drag the screenshot file into your Word document.
4. Keep your document open and go the next part of the
lab.
page 28
Part 3: Add Another Screenshot and

Submit Your Document
1. Open a new web browser window and go to
mhhm.com.
2. Take a screenshot of your results (label it 1-0A) of the
page and paste it into your lab document.
3. Save your document and submit it to your instructor. To
download your document for OneDrive, click File >
Save As > Download a Copy.
End of Lab
Lab 1-1 Data Analytics in Financial

Accounting
Let’s see how we might perform some simple Data
Analytics. The purpose of this lab is to help you identify
relevant questions that may be answered using Data
Analytics.
Company summary
You were just hired as an analyst for a credit rating
agency that evaluates publicly listed companies in the
United States. The company already has some Data
Analytics tools that it uses to evaluate financial
statements and determine which companies have higher
risk and which companies are growing quickly. The
company uses these analytics to provide ratings that will
allow lenders to set interest rates and determine whether
to lend money in the first place. As a new analyst, you’re
determined to make a good first impression.
Technique
Some experience with spreadsheets and basic formulas
is helpful here.
Software needed
Word processor
Web browser
Screen capture tool (Windows: Snipping Tool; Mac: Cmd
+ Shift + 4)

Part 1: Identify appropriate questions, and develop a
hypothesis for each question.
Part 2: Master the data.
Part 3: Perform a simple analysis.
Part 1: Identify the Questions

Think about ways that you might analyze data from a
financial statement. You could use a horizontal analysis to
view trends over time, a vertical analysis to show account
proportions, or ratios to analyze relationships.
1. Create a new word processing document and name the
file “Lab 1-1 Data Analytics in Financial Accounting
Lab—[Your name] [Your email address].”
2. Use what you know about financial statement analysis
(or search the web if you need a refresher) to generate
three different metrics for evaluating financial
performance. For example, if you wanted to page 29
evaluate a company’s profit margin from
one year to the next your question might be, “Has
[Company X’s] gross margin increased in the last three
years?” Type your three questions in your
document.
3. Next to each question generate a hypothetical
answer to the question to help you identify what your
expected output would be. You may use some insight or
intuition or search for industry averages to inform your
hypothesis. For example: “Hypothesis: Apple Inc’s
gross margin has increased slightly in the past 3 years.”
4. Save your document.
Part 2: Master the Data

To answer your questions, you’ll need to evaluate specific
account values or financial statement paragraphs. As an
analyst, you have access to the Security and Exchange
Commission’s (SEC’s) EDGAR database of XBRL
financial statements as well as a list of XBRL tags from
the Financial Accounting Standards Board (FASB). XBRL
stands for eXtensible Business Reporting Language and
is used to make the data in financial statements machine-
readable. Public companies have been preparing XBRL
reports since 2008. While there are some issues with
XBRL data, such data have become a useful means for
comparing and analyzing financial statements. Every
value, date, and paragraph is “tagged” with a label that
identifies what each specific value represents, similar to
assigning attributes in a database. Because companies
tag their financial statements with XBRL tags, you can
use those tags to identify specific data that you need to
answer your questions from Part 1.
Analyze your questions:
5. Evaluate each question from Part 1. There are
specific data attributes that will help you find the answer
you’re looking for. For example, if your question was
“Has [Company X’s] gross margin increased in the last
3 years?” and the expected answer is “Apple Inc’s
gross margin has increased slightly in the past 3 years,”
this tells you what attributes (or fields) to look for:
company name, gross margin (sales revenues – cost of
goods sold), year.
6. For each of your questions, identify the account or
data attribute you need to answer your question. Then
use FASB’s XBRL taxonomy (see next section for
instructions) to identify the specific XBRL tags that
represent those accounts. For example:
Company name =
EntitySectorIndustryClassificationPrimary
Gross margin = GrossProfit
Sales revenues = SalesRevenueNet
Cost of goods sold = CostOfGoodsAndServicessold
Year = DocumentPeriodEndDate
Identify XBRL tags from the FASB’s taxonomy:
8. Open a web browser, and go to xbrlview.fasb.org.
9. Click the + next to US GAAP (2019-01-31).
10. Click the ALL (Main/Entire) option, and then click
Open to load the taxonomy.
11. Navigate through the financial statements to determine
which accounts you need to answer your questions
from Part 1. The name of the XBRL tag is found in the
properties pane next to “Name.” For example, the tag
for Total Assets can be found by clicking + Statement of
Financial Position [Abstract], + Statement [Table], +
Statement [Line Items], + Assets [Abstract], + Assets,
Total, as shown in Lab Exhibit 1-1A. You may also use
the search function.
LAB EXHIBIT 1-1A

Browse the XBRL Taxonomy for Financial Fact Tags
Needed for Your Analysis
Source: Google.
page 30
Note: Be careful when you use the
search function. The tag you see in the results may
appear in the wrong statement. Double-click the tag to
expand the tree and show where the account appears.
Part 3: Perform the Analysis

Now that you’ve identified your questions and the data
sources, you can build your model and perform your
analysis. Because XBRL data are dynamic, we’ll use a
tool that pulls live data based on your inputs.
12. In your web browser, click on the eBook via Connect to
locate the table of contents where you will find
Additional Student Resources > Financial
Statement Analysis.
13. Log into your Google Account.
14. Click File > Make a Copy. . .
15. In your new document, add your tags from Part 2
under the Financial Facts header, similar to Exhibit 1-
1B.
page 31
LAB EXHIBIT 1-1B
Add Your Tags to Perform a Simple Analysis Using
XBRL Data
Source: Google.
16. Under the Analysis header, use formulas to create

your analysis from Part 1. You may enhance your
output by using conditional formatting or other
visualizations that will be covered in Chapter 4.
17. Take a screenshot (label it 1-1A) of your analysis and
paste it into your lab document.
18. Save your document and submit it to your instructor.
End of Lab
Lab 1-2 Data Analytics in Managerial
Accounting
Let’s see how we might use customer data to understand
some simple data analytics. The purpose of this lab is to
help you identify relevant questions that may be answered
using data analytics.
Company summary
LendingClub is a U.S.-based, peer-to-peer lending
company, headquartered in San Francisco, California.
LendingClub facilitates both borrowing and lending by
providing a platform for unsecured personal loans
between $1,000 and $35,000. The loan period is for either
3 or 5 years. You have been brought in to help managers
improve their loan application process.
Technique
Some critical and creative thinking is helpful here.
Software needed
Word processor
Part 1: Identify appropriate questions and develop a

Part 2: Identify fields and values in a database that are
relevant to your questions.
page 32

LendingClub currently assigns a risk score to all loan
applicants. This risk score is used to determine (1)
whether a loan is accepted and (2) what interest rate
approved loans will receive. The risk score has been used
for the past 5 years, but LendingClub thinks there may
be better ways to evaluate this given that the number of
defaulted loans has increased in the past 2 years. It would
like you to propose a model that would help it potentially
assign a risk score to loan applicants.
file “Lab 1-2 Data Analytics in Managerial
Accounting Lab – [Your name] [Your email
address].”
2. Use what you know about loan risk (or search the web
if you need a refresher) to identify three different
questions that might influence risk. For example, if you
suspect risky customers live in a certain location, your
question might be “Where do the customers live?” Type
your three questions in your document.
3. Next to each question, generate a hypothetical
answer to each question to help you identify what
your expected output would be. You may use some
insight or intuition or search the Internet for ideas on
how to inform your hypothesis. For example:
“Hypothesis: Risky customers likely live in coastal
towns.”
4. Finally, identify the data that you would need to answer
each of your questions. For example, to determine
customer location, you might need the city, state, and
zip code. Additionally, if you hypothesize a specific
region, you’d need to know which cities, state, and/or
zip codes belong to that region. Add your required
data sources to each question in your document.
To answer your questions, you’ll need to evaluate specific
data that LendingClub collects. It has provided a listing
of fields that it collects in Table 1-2A.
LAB TABLE 1-2A

Names and Descriptions of Selected Data Attributes Collected by
LendingClub
Attribute Description
id Loan identification number
member_id Membership id
loan_amnt Requested loan amount
emp_length Employment length
issue_d Date of loan issue
loan_status Fully paid or charged off
pymnt_plan Payment plan: yes or no
purpose Loan purpose: e.g., wedding, medical,
debt_consolidation, car
zip_code Zip code
addr_state State
dti Debt-to-income ratio
delinq_2y Late payments within the past two
years
earliest_cr_line Oldest credit account
inq_last_6mnths Credit inquiries in the past 6 months
open_acc Number of open credit accounts
revol_bal Total balance of all credit accounts
revol_util Percentage of available credit in use
total_acc Total number of credit accounts
application_type Individual or joint application
page 33
6. Evaluate each question from Part 1. Do the data you

identified in your questions exist in the table provided?
Write the applicable fields next to each question in
your document.
7. Are there data values you identified that don’t exist in
the table? Write where else you might look to collect
the missing data or how you might suggest
collecting those it.
8. Save your document and submit to your instructor.
End of Lab
Lab 1-3 Data Analytics in Auditing

The purpose of this lab is to help you identify relevant
questions that may be answered using data analytics in
auditing. Let’s evaluate how we might use master and
transaction data from an enterprise resource planning
system to perform some simple data analytics to assist
the financial statement audit.
Company summary
ABC Company is a large retailer that collects its order-to-
cash data in a large ERP system that was recently
updated to comply with the AICPA’s audit data standards.
ABC Company currently collects all relevant data in the
ERP system and digitizes any contracts, orders, or
receipts that are completed on paper. The credit
department reviews customers who request credit. Sales
orders are approved by managers before being sent to
the warehouse for preparation and shipment. Cash
receipts are collected by a cashier and applied to a
customer’s outstanding balance by an accounts
receivable clerk.
You have been assigned to the audit team that will
perform the internal controls audit of ABC Company.
Technique
Familiarity with database structure and primary-foreign
key relationships may be helpful.
Software needed
Word processor
Web browser
+ Shift + 4)

Part 2: Translate questions into target fields and value
in a database and perform a simple analysis.

Your audit team has been tasked with identifying potential
internal control weaknesses within the order-to-cash
process.
page 34

file “Lab 1-3 Data Analytics in Auditing Lab – [Your
name] [Your email address].”
2. Use what you know about internal controls over the
order-to-cash process (or search the web if you need a
refresher) to identify three different questions that might
indicate internal control weakness. For example, if you
suspect that a manager may be delaying approval of
shipments sent to customers, your question might be
“Are any shipping managers approving shipments more
than 2 days after they are received?” Type your three
questions in your document.
3. Next to each question generate a hypothetical
answer to each question to help you identify what
your expected output would be. You may use some
insight or intuition or search the Internet for ideas on
how to inform your hypothesis. For example:
“Hypothesis: Only 1 or 2 shipping managers are
approving shipments more than 2 days after they are
received.”
4. Finally, identify the data that you would need to answer
each of your questions. For example, to determine the
timing of approval and who is involved, you might need
the approver id, the order date, and the approval date.
Add your required data sources to each question in
your document.

To answer your questions, you’ll need to evaluate the data
that are available in the audit data standards.
6. Open your web browser and search for “Audit data
standards order to cash.” Follow the link to the “Audit
Data Standards Library—AICPA,” then look for the
“Audit Data Standard—Order to Cash Subledger
Standard” PDF document.
7. Quickly scan through the document for fields that relate
to each question you identified in Part 1. For example, if
you’re looking for the shipment timing and approval
data, you would need the
Shipments_Made_YYYYMMDD_YYYYMMDD table
and Approved_By, Entered_Date, and Approved_Date
fields. List the tables and fields from the audit data
standard in your document needed for each
question.
8. Identify any data that don’t appear in the audit data
standard that might also be relevant to your questions.
End of Lab
Lab 1-4 Comprehensive Case:

Dillard’s Store Data
The purpose of this lab is to help you identify relevant
questions for Dillard’s Inc. based on its data.
Company summary
Dillard’sis a department store with approximately 330
stores in 29 states. Its headquarters is in Little Rock,
Arkansas. You can learn more about Dillard’s by looking
at finance.yahoo.com (ticker symbol = DDS) and the
Wikipedia site for DDS. You’ll quickly note that William T.
Dillard II is an accounting grad of the University of
Arkansas and the Walton College of Business, which may
be why he shared transaction data with us to make
available for this lab and labs throughout this text.
page 35
Technique
The data for this lab and all other Dillard’s labs must be
accessed through the University of Arkansas Remote
Desktop. Directions for accessing the Remote Desktop
can be found at www.mhhe.com/richardsondaa2e. See
your instructor for login credentials. From the Walton
College website, we note the following:
The Dillard’s Department Store Database contains retail sales

information gathered from store sales transactions. The sale process
begins when a customer brings items intended for purchase (clothing,
jewelry, home décor, etc.) to any store register. A Dillard’s sales
associate scans the individual items to be purchased with a barcode
reader. This populates the transaction table (TRANSACT), which will
later be used to generate a sales receipt listing the item, department,
and cost information (related price, sale price, etc.) for the customer.
When the customer provides payment for the items, payment details
are recorded in the transaction table, the receipt is printed, and the
transaction is complete. Other tables are used to store information
about stores, products, and departments.
Source: Accessed July 15, 2019.

http://walton.uark.edu/enterprise/dillardshome.php.
This retail sales information, UA_DILLARDS, was

provided to the Walton College of Business by Dillard’s
Stores Inc. The information consists of five tables with
more than 128 million rows already populated and ready
for use.
This is a gifted dataset that is based on real
operational data. Like any real database, integrity
problems may be noted. This can provide a unique
opportunity not only to expose students to real data, but
also to illustrate the effects of data integrity problems.
Software needed
Word processor
Web browser
+ Shift + 4)
Access to the dataset is available at
http://walton.uark.edu/enterprise/dillardshome.php. If you
plan on doing additional labs on Dillard’s data, you must
receive permission from the Walton College to access
the data before use. Additional access instructions are
available from your instructor or on the Connect website.
Part 1: Identify appropriate questions for a retailer.

Part 2: Translate questions into target tables, fields,
and values in the Dillard’s database.

file “Lab 1-4 Comprehensive Case – Dillard’s Data
[Your name] [Your email address].”
2. Assume that Dillard’s wants to improve profitability.
Name three questions that could be asked to assess
current profitability levels for each product and how
profitability could be improved in the near future.
page 36
3. Assume that Dillard’s wishes to improve its
online sales and profitability on those sales. What three
questions could be asked to see where Dillard’s stands
on its online sales?

To answer your questions and related questions, access
the Data Dictionary and Entity Relationship Diagram for
the Dillard’s Entity Relationship data at
http://walton.uark.edu/enterprise/dillardshome.phpor
consider the information in Exhibits 1-4A and 1-4B.
LAB EXHIBIT 1-4A

LAB EXHIBIT 1-4B
Source: Accessed September 25, 2017.
http://walton.uark.edu/enterprise/dillardshome.php.
Metadata
Attribute Description Values
AMT Total amount of the 26.25, 44.00, . . .
transaction charge to
the customer
BRAND The brand name of TOMMY HI, MARK
the stock item ECK, . . .
CITY City where the store ST. LOUIS,
is located TAMPA, . . .
CLASSID Stock Item 5305, 4505, 8306, .
Classification ..
COLOR The color of the stock BLACK, KHAKI, . .
item .
COST The cost of the stock 9.00, 15.00, . . .
item
DEPT Department where 800, 801, 1100, . . .
the stock item belong
DEPTDESC Description of the CLINIQUE,
department LESLIE, . . .
INTERID Internal ID 265005802,
671901998, . . .
MIC Master Item Code 862, 689, . . .
ORGPRICE Original price of the 75.00, 44.00, . . .
item stock
PACKSIZE The quantity of item 1, 3, . . .
per pack
Metadata
QUANTITY Item quantity of the 1, 2, 3, . . .
transaction
REGISTER Register Number of 580, 30, 460, . . .
the current
transaction
RETAIL The retail price of the 19.75, 34.00, . . .
stock item
SALEDATE Sale price of the item 2005-01-20, 2005-
stock 06-02, . . .
SEQ Sequence number 298100028,
213500030, . . .
SIZE The size of the stock L, 070N, 22, . . .
item
SKU Stock Keeping Unit 4757355, 2128748,
number of the stock . . .
item
SPRICE Sale price of the item 26.25, 65.00, . . .
stock
STATE State where the store FL, MO, AR, . . .
is located
STORE Store Number 2, 3, 4, 100, . . .
STYLE The specific style of 51 MERU08, 9
the stock item 126NAO, . . .
STYPE Type of the P, R
transaction (Return
or Purchase)
TRANNUM Transaction Code 09700, 01800, . . .
Metadata
UPC Universal Product 000400004087945,
Code for the stock . . .
item
VENDOR The vendor number 5511283, 2726341,
of the stock item ...
ZIP ZIP Code 33710, 63126, . . .
5. You’re trying to learn about where Dillard’s stores are

located to identify locations for the next additional store.
Consider the STORE table. What questions could be
asked about store location given data availability?
6. What questions would you have regarding data fields in
the SKU table that could be used to help address the
cost of shipping? What additional information would be
helpful to address this question?
7. What table and fields could address the question of the
profit margin (sales price less cost) on each product
(SKU) available for sale?
page 37
8. If you’re interested in learning which product
is sold most often at each store, what tables and fields
would you consider?
End of Lab
1 http://www.forbes.com/sites/bernardmarr/2015/09/30/big-data-20-mind-boggling-facts-
everyone-must-read/#2a3289006c1d (accessed March 2019).
2 Roger S. Debreceny and Glen L. Gray, “IT Governance and Process Maturity: A
Multinational Field Study,” Journal of Information Systems 27, no. 1 (Spring 2013), pp.
157–88.
3 H. Chen, R. H. L. Chiang, and V. C. Storey, “Business Intelligence Research,” MIS
Quarterly 34, no. 1 (2010), pp. 201–3.
4 “Data Driven: What Students Need to Succeed in a Rapidly Changing Business World,”
PwC, https://www.pwc.com/us/en/faculty-resource/assets/pwc-data-driven-paper-
feb2015.pdf, February 2015 (accessed March 20, 2019).
5 “Open Data: Unlocking Innovation and Performance with Liquid Information,” McKinsey
Global Institute,
http://www.mckinsey.com/insights/business_technology/open_data_unlocking_innovation_
and_performance_with_liquid_information, October 2013 (accessed September 7, 2015).
6 Joseph Kennedy, “Big Data’s Economic Impact,” https://www.ced.org/blog/entry/big-datas-
economic-impact, December 3, 2014 (accessed January 9, 2016).
7 Deloitte, “Adding Insight to Audit: Transforming Internal Audit Through Data Analytics.”
Accessed January 10, 2016.
http://www2.deloitte.com/content/dam/Deloitte/ca/Documents/audit/ca-en-audit-adding-
insight-to-audit.pdf.
8 PwC, “Data Driven: What Students Need to Succeed in a Rapidly Changing Business
World,” http://www.pwc.com/us/en/faculty-resource/assets/PwC-Data-driven-paper-
Feb2015.pdf, February 2015 (accessed January 9, 2016).
9 EY, “How Big Data and Analytics Are Transforming the Audit.” Accessed January 27,
2016. https://eyo-iis-pd.ey.com/ARC/documents/EY-reporting-ssue-9.pdf,posted April
2015.
10 Deloitte, “The Power of Tax Data Analytics,”
http://www2.deloitte.com/us/en/pages/tax/articles/top-ten-things-about-tax-data-
analytics.html (accessed October 12, 2016).
11 “One-Third of BI Pros Spend Up to 90% of Time Cleaning Data,”
http://www.eweek.com/database/one-third-of-bi-pros-spend-up-to-90-of-time-cleaning-
data.html, posted June 2015 (accessed March 15, 2016).
12 Foster Provost and Tom Fawcett, Data Science for Business: What You Need to Know
about Data Mining and Data-Analytic Thinking (Sebastopol, CA: O’Reilly Media, 2013).
13 https://www.lendingclub.com/ (accessed September 29, 2016).
page 38
Chapter 2
Mastering the Data

This chapter provides an overview of the types of data that are used
in the accounting cycle and common data that are stored in a
relational database. The second step of the IMPACT cycle is
“mastering the data.” We will describe how data are requested and
extracted to answer business questions and how to transform data
for use via data preparation, validation, and cleaning. We conclude
with an explanation of how to load data into the appropriate tool in
preparation for analyzing data to make decisions.
A Look Back
Chapter 1 defined Data Analytics and explained that the value of
Data Analytics is in the insights it provides. We described the Data
Analytics Process using the IMPACT cycle model and explained how
this process is used to address both business and accounting
questions. We specifically emphasized the importance of identifying
appropriate questions that data analytics might be able to address.
A Look Ahead
Chapter 3 describes how to go from defining business problems to
analyzing data, answering questions, and addressing business
problems. We identify four types of data analytics and describe
various approaches and techniques that are most relevant to
analyzing accounting data.
page 39
Wichy/Shutterstock
We are lucky to live in a world in which data are abundant. However, even with rich
sources of data, when it comes to being able to analyze data and turn them into
useful information and insights, very rarely can an analyst hop right into a dataset
and begin analyzing. Datasets almost always need to be cleaned and validated
before they can be used. Not knowing how to clean and validate data can, at best,
lead to frustration and poor insights and, at worst, lead to horrible security violations.
While this text takes advantage of open source datasets, these datasets have all
been scrubbed not only for accuracy, but also to protect the security and privacy of
any individual or company whose details were in the original dataset.
In 2015, a pair of researchers named Emil Kirkegaard and Julius Daugbejerg
Bjerrekaer scraped data from OkCupid, a free dating website, and provided the data
onto the “Open Science Framework,” a platform researchers use to obtain and share
raw data. While the aim of the Open Science Framework is to increase transparency,
the researchers in this instance took that a step too far—and a step into illegal
territory. Kirkegaard and Bjerrekaer did not obtain permission from OkCupid or from
the 70,000 OkCupid users whose identities, ages, genders, religions, personality
traits, and other personal details maintained by the dating site were provided to the
public without any work being done to anonymize or sanitize the data. If the
researchers had taken the time to not just validate that the data were complete, but
also to sanitize them to protect the individuals’ identities, this would not have been a
threat or a news story. On May 13, 2015, the Open Science Framework removed the
OkCupid data from the platform, but the damage of the privacy breach had already
been done.1
OBJECTIVES
LO 2-1 Understand how data are organized in an accounting

information system.
LO 2-2 Understand how data are stored in a relational
database.
LO 2-3 Explain and apply extraction, transformation, and
loading (ETL) techniques.
page 40
As you learned in Chapter 1, Data Analytics is a process, and we

follow an established data analytics model called the IMPACT cycle
as introduced in Chapter 1.2 The IMPACT cycle begins with
identifying business questions and problems that can be, at least
partially, addressed with data (the “I” in the IMPACT model). Once
the opportunity or problem has been identified, the next step is
mastering the data (the “M” in the IMPACT model), which requires
you to identify and obtain the data needed for solving the problem.
Mastering the data requires a firm understanding of what data are
available to you and where they are stored, as well as being skilled
in the process of extracting, transforming, and loading (ETL) the data
in preparation for data analysis. While the extraction piece of the
ETL process may often be completed by the information systems
team or the database administrator, it is also possible that you will
have access to raw data that you will need to extract out of the
source database. Both methods of requesting data for extraction and
of extracting data yourself are covered in this chapter. The mastering
the data step can be described via the ETL process. The ETL
process is made up of the following five steps:
Step 1 Determine the purpose and scope of the data

request (extract).
Step 2 Obtain the data (extract).
Step 3 Validate the data for completeness and integrity
(transform).
Step 4 Sanitize the data (transform).
Step 5 Load the data in preparation for data analysis
(load).
This chapter will provide details for each of these five steps.
HOW DATA ARE USED AND STORED

IN THE ACCOUNTING CYCLE
LO 2-
1
Understa
nd how
data are
organize
d in an
accounti
ng
informati
on
system.
Before you can identify and obtain the data, you must have a
comfortable grasp on what data are available to you and where such
data are stored. A basic understanding of accounting processes and
its associated data, how those data are organized, and why the data
was captured, can help you request the right data and facilitate that
request so that you know exactly where each piece of data is held.
Accounting data can be presented in many formats. When you
think of accounting data, you might think of financial statements,
budgets, and managerial reports regarding point-of-sale, cost, supply
chain, or customer relationship management data. All of these
reports are the result of compiling data into a structured format that
is ready for consumption. Financial statements get placed online or
mailed out to investors, budgets get approved and managed, and
managerial reports are sent to decision-makers to monitor progress.
However, each of these documents originated with data in its rawest
form. Sometimes you will be required to make decisions or perform
analysis based on these formalized documents, but when you are
performing data analysis, it is always preferable to receive data in its
rawest form (not aggregated or organized with subheadings) so that
you have more flexibility with your analysis and so there is no
question about what the data represents. Because of that
preference, in this chapter, we will focus on how raw data is stored
so that you can learn how to access that data for your data analysis
projects.
Even with the focus on raw data and where it is stored, there is
variety in how data can be stored. Most commonly, data is stored in
either flat files or a database. For many of our examples and hands-
on activities in this text, we will transform our data that is stored in a
database into a flat file. The most common example of a flat file that
you are likely used to is a range of data in an Excel page 41
spreadsheet. Put simply, a flat file is a means of
maintaining all of the data you need in one place. We can do a lot of
incredible data analysis and number crunching in flat files in Excel,
but as far as storing our data, it is generally inefficient to store all of
the data that you need for a given business process all in one place.
Instead, a relational database is frequently used for data storage
because it is more capable of ensuring data integrity and maintaining
“one version of the truth” across multiple processes. There are a
variety of applications that support relational databases (these are
referred to as Relational Database Management Systems or
RDBMS). We cover three different RDBMS in this textbook:
Microsoft Access, SQLite, and Microsoft SQL Server.
Microsoft Access: for any user of Microsoft products (Word, Excel,
PowerPoint, etc.) the navigation of Microsoft Access is familiar, so it
is a relatively easy entry point for working with relational databases.
It is a great entry tool to learn how tables are related via primary
and foreign keys because entire databases can be built via a
graphical user interface instead of having to use SQL statements to
create tables and relationships. The usage of Microsoft Access
throughout the textbook is meant to provide an entry-level view into
data management and the structure of a relational database, as well
as a good starting point for writing SQL queries to extract data.
SQLite: SQLite is an open-source solution to data management.
For a user that is at least somewhat familiar with relational database
management, it is a friendly tool, and presents an intuitive interface
for writing SQL statements. The usage of SQLite throughout the
textbook is meant to provide an open-source alternative to Microsoft
Access. It also provides a good starting point for writing SQL
queries to extract data.
Microsoft SQL Server: Microsoft SQL Server can support
enterprise-level data in ways that smaller RDBMS programs, such
as Access and SQLite, cannot. While both Microsoft Access and
SQLite can be (and are) used in professional settings, the usage of
SQL Server throughout the textbook is meant to provide experience
that replicates working with much larger and more complex datasets
that you will likely experience in the professional world.
There are many other examples of relational database
management systems, including Teradata, MySql, Oracle RDBMS,
IBM DB2, Amazon RDS, and PostGreSQL.
Regardless of the DBMS, relational databases have principles
that guide how they are modeled.
Exhibit 2-1, a simplified version of a Unified Modeling Language
(UML) class diagram, is an illustration or a drawing of the tables and
their relationships to each other (i.e., a database schema). Relational
databases are discussed in greater depth in Learning Objective 2-2.
EXHIBIT 2-1
Procure-to-Pay Database Schema (Simplified)
page 42
DATA AND RELATIONSHIPS IN A

RELATIONAL DATABASE
LO 2-
2
Understa
nd how
data are
stored in
a
relational
databas
e.
In this text, we will work with data in a variety of forms, but

regardless of the tool we use to analyze data, structured data should
be stored in a normalized relational database. There are occasions
for working with data directly in the relational database, but many
times when we work with data analysis, we’ll prefer to export the
data from the relational database and view it in a more user-friendly
form. The benefit of storing data in a normalized database outweighs
the downside of having to export, validate, and sanitize the data
every time you need to analyze the information.
Storing data in a normalized, relational database instead of a flat
file ensures that data are complete, not redundant, and that
business rules and internal controls are enforced; it also aids
communication and integration across business processes. Each
one of these benefits is detailed here:
Completeness. Ensures that all data required for a business
process are included in the dataset.
No redundancy. Storing redundant data is to be avoided for several
reasons: It takes up unnecessary space (which is expensive), it
takes up unnecessary processing to run reports to ensure that there
aren’t multiple versions of the truth, and it increases the risk of data-
entry errors. Storing data in flat files yields a great deal of
redundancy, but normalized relational databases require there to be
one version of the truth and for each element of data to be stored in
only one place.
Business rules are enforced. As will become increasingly evident as
we progress through the material in this text, relational databases
can be designed to aid in the placement and enforcement of internal
controls and business rules in ways that flat files cannot.
Communication and integration of business processes. Relational
databases should be designed to support business processes
across the organization, which results in improved communication
across functional areas and more integrated business processes.3
It is valuable to spend some time basking in the benefits of
storing data in a relational database because it is not necessarily
easier to do so when it comes to building the data model or
understanding the structure. It is arguably more complex to
normalize your data than it is to throw redundant data without
business rules or internal controls into a spreadsheet.
Columns in a Table: Primary Keys, Foreign

Keys, and Descriptive Attributes
When requesting data, it is critical to understand how the tables in a
relational database are related. This is a brief overview of the
different types of attributes in a table and how these attributes
support the relationships between tables. It is certainly not a
comprehensive take on relational data modeling, but it should be
adequate in preparing you for creating data requests.
Every column in a table must be both unique and relevant to the
purpose of the table. There are three types of columns: primary
keys, foreign keys, and descriptive attributes.
Each table must have a primary key. The primary key is typically
made up of one column. The purpose of the primary key is to ensure
that each row in the table is unique, so it is often referred to as a
“unique identifier.” It is rarely truly descriptive; instead, a collection of
letters or simply sequential numbers are often used. As a student,
you are probably already very familiar with your page 43
unique identifier—your student ID number at the
university is the way you as a student are stored as a unique record
in the university’s data model! Other examples of unique identifiers
that you are familiar with would be check numbers and driver’s
license numbers.
One of the biggest differences between a flat file and a relational
database is simply how many tables there are—when you request
your data into a flat file, you’ll receive one big table with a lot of
redundancy. While this is often ideal for analyzing data, when the
data are stored in the database, each group of information is stored
in a separate table. Then, the tables that are related to one another
are identified (e.g., Supplier and Purchase Order are related; it’s
important to know which Supplier the Purchase Order is from). The
relationship is created by placing a foreign key in one of the two
tables that are related. The foreign key is another type of attribute,
and its function is to create the relationship between two tables.
Whenever two tables are related, one of those tables must contain a
foreign key to create the relationship.
The other columns in a table are descriptive attributes. For
example, Supplier Name is a critical piece of data when it comes to
understanding the business process, but it is not necessary to build
the data model. Primary and foreign keys facilitate the structure of a
relational database, and the descriptive attributes provide actual
business information.
Refer to Exhibit 2-1, the database schema for a typical procure-
to-pay process. Each table has an attribute with the letters “PK” next
to them—these are the primary keys for each table. The primary key
for the Materials Table is “Item_No.,” the primary key for the
Purchase Order Table is “PO_Number.,” and so on. Several of the
tables also have attributes with the letters “FK” next to them—these
are the foreign keys that create the relationship between pairs of
tables. For example, look at the relationship between the Supplier
Table and the Purchase Order Table. The primary key in the Supplier
Table is “SupplierID.” The line between the two tables links the
primary key to a foreign key in the Purchase Order Table, also
named “SupplierID.”
The Line Items Table in Table 2-1 has so much detail in it that it
requires two attributes to combine as a primary key. This is a special
case of a primary key often referred to as a composite primary key,
in which the two foreign keys from the tables that it is linking
combine to make up a unique identifier. The theory and details that
support the necessity of this linking table are beyond the scope of
this text—if you can identify the primary and foreign keys, you’ll be
able to identify the data that you need to request. Table 2-2 shows a
subset of the data that are represented by the Purchase Order table.
You can see that each of the attributes listed in the class diagram
appears as a column, and the data for each purchase order are
accounted for in the rows.
TABLE 2-1
Line Items Table: Purchase Order Detail Table
Purchase Order Detail

PO_Number Item_Number Quantity Purchased
1787 10 50
1787 25 50
1789 5 30
1790 5 100
TABLE 2-2
Purchase Order Table
page 44
PROGRESS CHECK
1. Referring to Exhibit 2-1, locate the relationship between the
Supplier and Purchase Order tables. What is the unique
identifier of each table? (The unique identifier attribute is
called the primary key—more on how it’s determined in the
next learning objective.) Which table contains the attribute
that creates the relationship? (This attribute is called the
foreign key—more on how it’s determined in the next
learning objective.)
2. Referring to Exhibit 2-1, review the attributes in the
Purchase Order table. There are two foreign keys listed in
this table that do not relate to any of the tables in the
diagram. Which tables do you think they are? What type of
data would be stored in those two tables?
3. Refer to the two tables that you identified in Progress
Check 2 above that would relate to the Purchase Order
table, but are not pictured in this diagram. Draw a sketch of
what the UML Class Diagram would look like if those tables
were included. Draw the two classes to represent the two
tables (i.e., rectangles), the relationships that should exist,
and identify the Primary Keys for the two new tables.
DATA DICTIONARIES
In the previous section, you learned about how data are stored by
focusing on the procure-to-pay database schema. Viewing schemas
and processes in isolation clarifies each individual process, but it can
also distort reality—these schemas typically do not represent their
own separate databases. Rather, each process-specific database
schema is a piece of a greater whole, all combining to form one
integrated database.
As you can imagine, once these processes come together to be
supported in one database, the amount of data can be massive.
Understanding the processes and the basics of how data are stored
is critical, but even with a sound foundation, it would be nearly
impossible for an individual to remember where each piece of data is
stored, or what each piece of data represents.
Creating and using a data dictionary is paramount in helping
database administrators maintain databases and analysts identify
the data they need to use. In Chapter 1, you were introduced to the
data dictionary for the LendingClub. The same cut-out of the
LendingClub data dictionary is provided in Exhibit 2-2 as a
reminder.
EXHIBIT 2-2
2007–2012 LendingClub Data Dictionary for Declined Load Data
Source: Accessed March, 2019. Available at https://www.lendingclub.com/info/download-

data.action
RejectStats
Description
File
Amount Total requested loan amount
Requested
Application Date of borrower application
Date
Loan Title Loan title
Risk_Score Borrower risk (FICO) score
Dept-To- Ratio of borrower total monthly debt payments
Income Ratio divided by monthly income.
Zip Code The first 3 numbers of the borrower zip code
provided from loan application.
State Two digit State Abbreviation provided from loan
application.
Employment Employment length in years, where 0 is less than
Length 1 and 10 is greater than 10.
RejectStats
Description
File
Policy Code policy_code=1 if publicly available.
policy_code=2 if not publicly available
Because the LendingClub data are provided in a flat file, the

only two attributes necessary to describe the data are the attribute
name (e.g., Amount Requested) and a description of that attribute.
The description ensures that the data in each attribute are used and
analyzed in the appropriate way—it’s always important to remember
that technology will do exactly what you tell it to, so you must be
smarter than the computer! If you run analysis on an attribute
thinking it means one thing, when it actually means another, you
could make some big mistakes and bad decisions even when you’re
working with great data. It’s critical to get to know the data through
database schemas and data dictionaries thoroughly before
attempting to do any data analysis.
When you are working with data stored in a relational database,
you will have more attributes to keep track of in the data dictionary.
Table 2-3 provides an example of a data dictionary for a generic
Supplier table:
page 45
TABLE 2-3
Supplier Data Dictionary
PROGRESS CHECK
4. What is the purpose of the primary key? A foreign key? A
nonkey (descriptive) attribute?
5. How do data dictionaries help you understand the data from
a database or flat file?
EXTRACT, TRANSFORM, AND LOAD

(ETL) THE DATA
LO 2-
3
Explain
and
apply
extractio
n,
transfor
mation,
and
loading
(ETL)
techniqu
es.
Once you have familiarized yourself with the data via data
dictionaries and schemas, you are prepared to request the data from
the database manager or extract the data yourself. The ETL process
begins with identifying which data you need and is complete when
the clean data are loaded in the appropriate format into the tool to be
used for analysis.
This process involves:
1. Determining the purpose and scope of the data request.
2. Obtaining the data.
3. Validating the data for completeness and integrity. page 46
4. Cleaning the data.
5. Loading the data for analysis.
Extract
Determine exactly what data you need in order to answer your
business questions. Requesting data is often an iterative practice,
but the more prepared you are when requesting data, the more time
you will save for yourself and the database team in the long run.
Requesting the data involves the first two steps of the ETL
extraction process. Each step has questions associated with it that
you should try to answer.
Step 1: Determine the Purpose and Scope of the Data

Request
What is the purpose of the data request? What do you need the
data to solve? What business problem will they address?
What risk exists in data integrity (e.g., reliability, usefulness)? What
is the mitigation plan?
What other information will impact the nature, timing, and extent of
the data analysis?
Once the purpose of the data request is determined and scoped,
as well as any risks and assumptions documented, the next step is
to determine whom to ask and specifically what is needed, what
format is needed (Excel, PDF, database), and by what deadline.
Step 2: Obtain the Data

How will data be requested and/or obtained? Do you have access
to the data yourself, or do you need to request a database
administrator or the information systems department to provide the
data for you?
If you need to request the data, is there a standard data request
form that you should use? From whom do you request the data?
Where are the data located in the financial or other related
systems?
What specific data are needed (tables and fields)?
What tools will be used to perform data analytic tests or procedures
and why?
Obtaining the Data via a Data Request

Determining not only what data are needed, but also which tool will
be used to test and process the data will aid the database
administrator in providing the data to you in the most accessible
format.
It is also necessary to specify the format in which you would like
to receive the data; it is often preferred to receive data in a flat file
(i.e., if the data you requested reside in multiple tables or different
databases, they should be combined into one file without any
hierarchy or relationships built in), with the first row containing
column headings (names of the fields requested), and each
subsequent row containing data that correspond with the column
headings. Subtotals, breaks, and subheadings complicate data
cleaning and should not be included.4 When you receive the data,
make sure that you understand the data in each column; the data
dictionary should prove extremely helpful for this. If a data dictionary
is unavailable, then you should plan to meet with database users to
get a clear understanding of the data in each column.
page 47
In a later chapter, you will be provided a deep dive into the Audit
Data Standards (ADS) developed by the American Institute of
Certified Public Accountants (AICPA).5 The aim of the ADS is to
alleviate some of the headaches associated with requesting data for
audits by providing a guide to standardize audit data requests and
the format in which the data are provided from the company being
audited to the auditor. These include the following:
1. Order-to-Cash subledger standards
2. Procure-to-Pay subledger standards
3. Inventory subledger standards
4. General Ledger standards
While the ADSs provide an opportunity for standardization, they
are voluntary. Regardless of whether your request for data will
conform to the standards, a data request form template can make
communication easier between data requester and provider.
Example Standard Data Request Form:
Requester Name:
Requester Contact Number:
Requester Email Address:
Please provide a description of the information needed
(indicate which tables and which fields you require):
What will the information be used for?
Frequency (circle one) One-Off Annually Termly

Other:_____
Format you wish the data to be Spreadsheet Text File
delivered in (circle one): Word Other:
Document _____
Request Date:
Required Date:
Intended Audience:
Customer (if not requester):
Once the data are received, you can move on to the transformation
phase of the ETL process. The next step is to ensure that the data
that have been extracted are complete and correct.
page 48
Obtaining the Data Yourself

At times, you will have direct access to a database or information
system that holds all or some of the data you need. In this case, you
may not need to go through a formal data request process, and you
can simply extract the data yourself.
After identifying the goal of the data analysis project in the first
step of the IMPACT cycle, you can follow a similar process to how
you would request the data if you are going to extract it yourself:
1. Identify the tables that contain the information you need. You can do
this by looking through the data dictionary or the relationship model.
2. Identify which attributes, specifically, hold the information you need
in each table.
3. Identify how those tables are related to each other.
Once you have identified the data you need, you can start
gathering the information. There are a variety of methods that you
could take to retrieve the data. Two will be explained briefly here—
SQL and Excel—and there is a deep dive into SQL in appendices D
and H, as well as a deep dive into Excel’s VLookup in appendix C.
SQL: “Structured Query Language” (SQL, often pronounced
sequel) is a computer language that can be used to create, update,
and delete records and tables in relational databases, but in Data
Analysis, the focus is on extracting data—that is, to select the
precise attributes and records that fit the criteria of our data analysis
goal. Using SQL, we can combine data from one or more tables and
organize the data in a way that is more intuitive and useful for data
analysis than the way the data is stored in the relational database. A
firm understanding of the data—the tables, how they are related, and
their respective primary and foreign keys—is integral to extracting
the data.
Typically, data should be stored in the database and analyzed in
another tool such as Excel, IDEA, or Tableau. However, you can
choose to extract only the portion of the data that you wish to
analyze via SQL instead of extracting full tables and transforming the
data in Excel, IDEA, or Tableau. This is especially preferable when
the raw data stored in the database is large enough to overwhelm
Excel. Excel 2016 can only hold 1,048,576 rows on one
spreadsheet. When you attempt to bring in full tables that exceed
that amount, even when you use Excel’s powerful Power BI tools, it
will slow down your analysis if the full table isn’t necessary.
There is more description about writing queries and a chance to
practice creating joins in Appendix H.
Excel: When your data are not stored in a relational database, if
you want to do exploratory analysis (more on this in a later chapter)
or if the tables you wish to extract aren’t too large for Excel, the
tables that contain the data you need can be extracted in whole into
Excel and worked with directly in a spreadsheet. The advantage of
this is that further analysis will almost certainly be done in Excel, so it
could be beneficial to have all the data readily available for further
questions to drill down into once the initial question is answered.
Understanding the primary key and foreign key relationships is also
integral to working with the data directly in Excel.
When your data are stored directly in Excel, you can also use
Excel functions and formulas to combine data from multiple Excel
tables into one table, similar to how you can join data with SQL in
Access or another relational database. One of Excel’s most useful
tools for looking up data from two separate tables and matching
them based on a matching primary key/foreign key relationship is the
VLookup function. There are a variety of ways that the VLookup
function can be used, but for extracting and transforming data it is
best used to add a column to a table.
page 49
More information about creating VLookup functions in Excel is

provided in Appendix C.
The question of whether to use SQL or Excel’s tools (such as
VLookup) is primarily answered by where the data are stored. Since
data are most frequently stored in a relational database (as
discussed earlier in this chapter, due to the efficiency and data
integrity benefits relational databases provide), SQL will often be the
best option for retrieving data, after which that data can be loaded
into Excel or another tool for further analysis. Another benefit of SQL
queries is that they can be saved and reproduced at will or at regular
intervals. Having a saved SQL query can make it much easier and
more efficient to re-create data requests. However, if the data is
already stored in a flat file in Excel, there is little reason to use SQL.
Sometimes when you are performing exploratory analysis, even if
the data are stored in a relational database, it can be beneficial to
load entire tables into Excel and bypass the SQL step - this should
be considered carefully before doing so, though, because relational
databases handle large amounts of data much better than Excel can.
Writing SQL queries can also make it easier to load only the data
you need to analyze into Excel so that you do not overwhelm Excel’s
resources
Transform
Step 3: Validating the Data for Completeness and
Integrity
Any time data is moved from one location to another, it is possible
that some of the data could have been lost during the extraction. It is
critical to ensure that the extracted data are complete (that the data
you wish to analyze were extracted fully) and that the integrity of the
data remains (that none of the data have been manipulated or
tampered with during the extraction). Being able to validate the data
successfully requires you to not only have the technical skills to
perform the task, but also to know your data well. If you know what
to reasonably expect from the data in the extraction (How many
records should have been extracted? What are some checksums
you can rely on to ensure the data is complete and hasn’t been
tampered with?), then you have a higher likelihood of identifying
errors or issues from the extraction. The following four steps should
be completed to validate the data after extraction:
1. Compare the number of records that were extracted to the number
of records in the source database. This will give you a quick
snapshot into whether any data were skipped or didn’t extract
properly due to an error or datatype mismatch. This is a critical first
step, but it will not provide information about the data themselves
other than ensuring that the record counts match.
2. Compare descriptive statistics for numeric fields: Calculating the
minimums, maximums, averages, and medians will help ensure that
the numeric data were extracted completely.
3. Validate Date/Time fields in the same way as numeric fields by
converting the datatype to numeric and running descriptive statistic
comparisons.
4. Compare string limits for text fields: Text fields are unlikely to cause
an issue if you extracted your data into Excel because Excel allows
a generous maximum character number (for example, Excel 2016
allows 32,767 characters per cell). However, if you extracted your
data into a tool that does limit the number of characters in a string,
you will want to compare these limits to the source database’s limits
per field to ensure that you haven’t cut off any characters.
If an error is found, depending on the size of the dataset, you
may be able to easily find the missing or erroneous data by scanning
the information with your eyes. However, if the page 50
dataset is large, or if the error is difficult to find, it may
be easiest to go back to the extraction and examine how the data
were extracted, fix any errors in the SQL code, and re-run the
extraction.
Step 4: Cleaning the Data

After validating the data, you should pay close attention to the state
of the data and clean it as necessary to improve the quality of the
data and subsequent analysis. The following four items are some of
the more common ways that data will need to be cleaned after
extraction and validation:
1. Remove headings or subtotals: Depending on the extraction
technique used and the file type of the extraction, it is possible that
your data could contain headings or subtotals that are not useful for
analysis. Of course, these issues could be overcome in the
extraction steps of the ETL process if you are careful to request the
data in the correct format or to only extract exactly the data you
need.
2. Clean leading zeroes and nonprintable characters: Sometimes data
will contain leading zeroes or “phantom” (nonprintable) characters.
This will happen particularly when numbers or dates were stored as
text in the source database but need to be analyzed as numbers.
Nonprintable characters can be white spaces, page breaks, line
breaks, tabs, etc., and can be summarized as characters that our
human eyes can’t see, but that the computer interprets as a part of
the string. These can cause trouble when joining data because,
while two strings may look identical to our eyes, the computer will
read the nonprintable characters and will not find a match.
3. Format negative numbers: If there are negative numbers in your
dataset, ensure that the formatting will work for your analysis. For
example, if your data contain negative numbers formatted in
parentheses and you would prefer this formatting to be as a
negative sign, this needs to be corrected and consistent.
4. Correct inconsistencies across data, in general: If the source
database did not enforce certain rules around data entry, it is
possible that there are inconsistencies across the data—for
example, if there is a state field, Arkansas could be formatted as
“AR,” “Ark,” “Ar.,” etc. These will need to be replaced with a
common value before you begin your analysis if you are interested
in grouping data geographically.
A Note about Data Quality

As you prepare your data for analysis, you should pay close
attention to the quality of the data. Incorrect or invalid data can skew
your results and lead to inaccurate conclusions. Low-quality data
contain numerous errors, obsolete or incorrect data, or invalid data.
Here are five main data quality issues to pay attention to when
you analyze a dataset for the first time:
1. Dates: The most common problems revolve around the date format
because there are so many different ways a date can be presented.
For example, look at the different ways you can show July 6, 2022:
6-Jul-2022; 6.7.2022; 44018 (in Excel); 07/06/2022 (in the United
States); 06/07/2022 (in Europe); and the list goes on. You need to
format the date to match the acceptable format for your tool. The
ISO 8601 standard indicates you should format dates in the year-
month-day format or 2022-07-06 and most professional query tools
accept this format. If you use Excel to transform dates to this
format, highlight your dates and go to Home > Number > Format
Cells and choose Custom. Then type in YYYY-MM-DD and click
OK.
2. Numbers: Numbers can be misinterpreted, particularly if they are
manually entered. For example, 1 or I; 0 or O; 3 or E; 7 or seven.
Watch for invalid number formats when you start sorting page 51
and analyzing your date then go back and correct them.
Additionally, accounting artifacts such as dollar signs, commas,
parenthesis are pervasive in spreadsheet data (e.g. $12,345.22 or
(1,422.53)). As you clean the data remove any extra accounting
characters so numbers appear in their raw form (e.g. 12345.22 or
-1422.53).
Format Cells Window in Excel
3. International characters and encoding: When you work with data

that spans multiple countries it is likely that you will come across
accent marks or special characters, such as accent marks (á or À),
umlats (Ü), invisible computer characters (TAB, RETURN,
linebreak, null), or special characters that are used in query and
scripting languages (*, #, “, ’). In many cases, these can be
corrected with a find and replace or contained in quote marks so
they are ignored by the query language. Additionally, while most
modern computer programs use UNICODE as the text encoding
language, older databases will generate data in the ASCII format. If
your tool fails to populate your dataset accurately, having
international characters and symbols is likely to be a cause.
4. Languages and measures: Similar to international characters, data
elements may contain a variety of words or measures that have the
same meaning. For example, cheese or fromage; ketchup or
catsup; pounds or lbs; $ or €; Arkansas or AR. In order to properly
analyze the comparable data, you’ll need to translate them into a
common format by choosing one word as the standard and
replacing the equivalent words. Also make sure the measure
doesn’t change the meaning. The total value in U.S. dollars is not
the same thing as the total value in euros. Make sure you’re
comparing apples to apples or euros to euros.
page 52
5. Human error: Whenever there is manual input into the
data, there is a high probability that data will be bad simply because
it was mistyped or entered into the wrong place. There’s no hard
and fast rule for dealing with input errors other than being vigilant
and making corrections (e.g., find and replace) when they occur.
Load
Step 5: Loading the Data for Data Analysis
If the extraction and transformation steps have been done well by
the time you reach this step, the loading part of the ETL process
should be the simplest step. It is so simple, in fact, that if your goal is
to do your analysis in Excel and you have already transformed and
cleaned your data in Excel, you are finished. There should be no
additional loading necessary.
However, it is possible that Excel is not the last step for analysis.
The data analysis technique you plan to implement, the subject
matter of the business questions you intend to answer, and the way
in which you wish to communicate results will all drive the choice of
which tool you use to perform your analysis.
Throughout the text, you will be introduced to a variety of different
tools to use for analyzing data beyond Access and Excel. These will
include Tableau, Weka, and IDEA. As these tools are introduced to
you, you will learn how to load data into them.
ETL or ELT?
If loading the data into Excel is indeed the last step, are you actually
“extracting, transforming, and loading,” or is it “extracting, loading,
and transforming?”
The term ETL has been in popular use since the 1970s, and even
though methods for extracting and transforming data have gotten
easier to use, more accessible, as well as more robust, the term has
stuck. Increasingly, however, the procedure is shifting toward ELT.
Particularly with tools such as Microsoft’s Power BI suite, all of the
loading and transforming can be done within Excel, with data directly
loaded into Excel from the database, and then transformed (also
within Excel). The most common method for mastering the data that
we use throughout this textbook is more in line with ELT than ETL;
however, even when the order changes from ETL to ELT, it is still
more common to refer to the procedure as ETL.
PROGRESS CHECK
6. Describe two different methods for obtaining data for
analysis.
7. What are five common data quality issues that must be
fixed before analysis can take place?
Summary
The first step in the IMPACT cycle is to identify the
questions that you intend to answer through your data
analysis project. Once a data analysis problem or
question has been identified, the next step in the
IMPACT cycle is mastering the data, which can be
broken down to mean obtaining the data needed and
preparing it for analysis.
page 53
In order to obtain the right data, it is
important to have a firm grasp of what data are available
to you and how that information is stored.
Data are often stored in a relational database, which

helps to ensure that an organization’s data are
complete and to avoid redundancy. Relational
databases are made up of tables with uniquely
identified records (this is done through primary keys)
and are related through the usage of foreign keys.
To obtain the data, you will either have access to extract

the data yourself or you will need to request the data
from a database administrator or the information
systems team. If the latter is the case, you will complete
a data request form, indicating exactly which data you
need and why.
Once you have the data, they will need to be validated
for completeness and integrity—that is, you will need to
ensure that all of the data you need were extracted and
that all data are correct. Sometimes when data are
extracted, some formatting or sometimes even entire
records will get lost, resulting in inaccuracies. Correcting
the errors and cleaning the data is an integral step in
mastering the data.
Finally, after the data have been cleaned, there may be
one last step of mastering the data, which is to load
them into the tool that will be used for analysis. Often,
the cleaning and correcting of data occur in Excel and
the analysis will also be done in Excel. In this case,
there is no need to load the data elsewhere. However, if
you intend to do more rigorous statistical analysis than
Excel provides, or if you intend to do more robust data
visualization than can be done in Excel, it may be
necessary to load the data into another tool following the
transformation process.
Key Words
composite primary key (43) A special case of a
primary key that exists in linking tables. The composite
primary key is made up of the two primary keys in the
table that it is linking.
data dictionary (44) Centralized repository of

descriptions for all of the data attributes of a dataset.
data request form (47) A method for obtaining data if

you do not have access to obtain the data directly
yourself.
descriptive attributes (43) Attributes that exist in

relational databases that are neither primary nor foreign
keys. These attributes provide business information, but
are not required to build a database. An example would
be “Company Name” or “Employee Address.”
ETL (45) The extract, transform, and load process that

is integral to mastering the data.
flat file (41) A means of storing data in one place, such
as in an Excel spreadsheet, as opposed to storing the
data in multiple tables, such as in a relational database.
foreign key (43) An attribute that exists in relational

databases in order to carry out the relationship between
two tables. This does not serve as the “unique identifier”
for each record in a table. These must be identified when
mastering the data from a relational database in order to
extract the data correctly from more than one table.
mastering the data (40) The second step in the

IMPACT cycle; it involves identifying and obtaining the
data needed for solving the data analysis problem, as well
as cleaning and preparing the data for analysis.
primary key (42) An attribute that is required to exist in

each table of a relational database and serves as the
“unique identifier” for each record in a table.
relational database (41) A means of storing data in

order to ensure that the data are complete, not redundant,
and to help enforce business rules. Relational databases
also aid in communication and integration of business
processes across an organization.
page 54
ANSWERS TO PROGRESS
CHECKS
1. The unique identifier of the Supplier table is [Supplier
ID], and the unique identifier of the Purchase Order
table is [PO No.]. The Purchase Order table contains
the foreign key.
2. The foreign key attributes in the Purchase Order table
that do not relate to any tables in the view are
EmployeeID and CashDisbursementID. These
attributes probably relate to the Employee table (so that
we can tell which employee was responsible for each
Purchase Order) and the Cash Disbursement table (so
that we can tell if the Purchase Orders have been paid
for yet, and if so, on which check). The Employee table
would be a complete listing of each Employee, as well
containing the details about each employee (for
example, phone number, address, etc.). The Cash
Disbursement table would be a listing of the payments
the company has made.
3.
4. The purpose of the primary key is to uniquely identify

each record in a table. The purpose of a foreign key is
to create a relationship between two tables. The
purpose of a descriptive attribute is to provide
meaningful information about each record in a table.
Descriptive attributes aren’t required for a database to
run, but they are necessary for people to gain business
information about the data stored in their databases.
5. Data dictionaries provide descriptions of the function
(e.g., Primary Key or Foreign Key when applicable),
datatype, and field names associated with each column
(attribute) of a database. Data dictionaries are
especially important when databases contain several
different tables and many different attributes in order to
help analysts identify the information they need to
perform their analysis.
6. Depending on the level of security afforded to a
business analyst, she can either obtain data directly
from the database herself or she can request the data.
When obtaining data herself, the analyst must have
access to the raw data in the database and a firm
knowledge of SQL and data extraction techniques.
When requesting the data, the analyst doesn’t need the
same level of extraction skills, but she still needs to be
familiar with the data enough in order to identify which
tables and attributes contain the information she
requires.
7. Four common issues that must be fixed are removing
headings or subtotals, cleaning leading zeroes or
nonprintable characters, formatting negative numbers,
and correcting inconsistencies across the data.
page 55

1. Mastering the data can also be described via the ETL
process. The ETL process stands for:
a. extract, total, and load data.
b. enter, transform, and load data.
c. extract, transform, and load data.
d. enter, total, and load data.
2. Which of the following describes part of the goal of the
ETL process:
a. identify which approach to data analytics should be
used.
b. load the data into a relational database for storage.
c. communicate the results and insights found through
the analysis.
d. identify and obtain the data needed for solving the
problem.
3. The advantages of storing data in a relational database
include which of the following?
a. Help in enforcing business rules
b. Increased information redundancy
c. Integrating business processes
d. All of the above
e. Only A and B
f. Only B and C
g. Only A and C
4. The purpose of transforming data is:
a. to validate the data for completeness and integrity.
b. to load the data into the appropriate tool for analysis.
c. to obtain the data from the appropriate source.
d. to identify which data are necessary to complete the
analysis.
5. Which attribute is required to exist in each table of a
relational database and serves as the “unique identifier”
for each record in a table?
a. Foreign key
b. Unique identifier
c. Primary key
d. Key attribute
6. The metadata that describes each attribute in a
database is which of the following?
a. Composite primary key
b. Data dictionary
c. Descriptive attributes
d. Flat file
7. As mentioned in the chapter, which of the following is
not a common way that data will need to be cleaned
after extraction and validation?
a. Remove headings and subtotals.
b. Format negative numbers.
c. Clean up trailing zeroes.
d. Correct inconsistencies across data.
page 56
8. Why is Supplier ID considered to be a
primary key for a Supplier table?
a. It contains a unique identifier for each supplier.
b. It is a 10-digit number.
c. It can either be for a vendor or miscellaneous
provider.
d. It is used to identify different supplier categories.
9. What are attributes that exist in a relational database
that are neither primary nor foreign keys?
a. Nondescript attributes
b. Descriptive attributes
c. Composite key
d. Relational table attributes
10. Which of these is not included in the five steps of the
ETL process?
a. Determine the purpose and scope of the data request.
b. Obtain the data.
c. Validate the data for completeness and integrity.
d. Scrub the data.
1. The advantages of a relational database include limiting
the amount of redundant data that are stored in a
database. Why is this an important advantage? What
can go wrong when redundant data are stored?
2. The advantages of a relational database include
integrating business processes. Why is it preferable to
integrate business processes in one information
system, rather than store different business process
data in separate, isolated databases?
3. Even though it is preferable to store data in a relational
database, storing data across separate tables can
make data analysis cumbersome. Describe three
reasons it is worth the trouble to store data in a
relational database.
4. Among the advantages of using a relational database is
enforcing business rules. Based on your understanding
of how the structure of a relational database helps
prevent data redundancy and other advantages, how
does the primary key/foreign key relationship structure
help enforce a business rule that indicates that a
company shouldn’t process any purchase orders from
suppliers who don’t exist in the database?
5. What is the purpose of a data dictionary? Identify four
different attributes that could be stored in a data
dictionary, and describe the purpose of each.
6. In the ETL process, the first step is extracting the data.
When you are obtaining the data yourself, what are the
steps to identifying the data that you need to extract?
7. In the ETL process, if the analyst does not have the
security permissions to access the data directly, then he
or she will need to fill out a data request form. While
this doesn’t necessarily require the analyst to know
extraction techniques, why does the analyst still need to
understand the raw data very well in order to complete
the data request?
8. In the ETL process, when an analyst is completing the
data request form, there are a number of fields that the
analyst is required to complete. Why do you think it is
important for the analyst to indicate the page 57
frequency of the report? How do you think
that would affect what the database administrator does
in the extraction?
9. Regarding the data request form, why do you think it is
important to the database administrator to know the
purpose of the request? What would be the importance
of the “To be used in” and “intended audience” fields?
10. In the ETL process, one important step to process
when transforming the data is to work with NULL, N/A,
and zero values in the dataset. If you have a field of
quantitative data (e.g., number of years each individual
in the table has held a full-time job), what would be the
effect of the following?
a. Transforming NULL and N/A values into blanks
b. Transforming NULL and N/A values into zeroes
c. Deleting records that have NULL and N/A values from
your dataset
(Hint: Think about the impact on different aggregate
functions, such as COUNT and AVERAGE.)
Problems
The following problems correspond to the College
Scorecard data. You should be able to answer each
question by just looking at the data dictionary included in
Appendix K, but if you would like to use the raw data, feel
free to do so (CollegeScorecard_RawData.txt).
1. Which attributes from the College Scorecard data would
you need to compare cost of attendance across types
of institutions (public, private nonprofit, or private for-
profit)?
you need to compare SAT scores across types of
institutions (public, private nonprofit, or private for-
profit)?
you need to compare levels of diversity across types of
profit)?
4. If you were conducting a data analysis in order to
compare the percentage of students who receive
federal loans at universities above and below the
median cost of attendance across all institutions, your
analysis would require several steps. One of the steps
is to know what question needs to be answered first in
order to complete the analysis. Come up with a set of
questions that need to be answered.
5. If you were analyzing the levels of diversity across
public and private institutions using the College
Scorecard data, how would you define diversity in terms
of the data provided? Would it be beneficial to combine
attributes?
you need to compare completion rate across types of
profit)?
you need to compare the percentage of students who
receive federal loans at universities above and below
the median cost of attendance across all institutions
(public, private nonprofit, or private for-profit)?
you need to determine if different regions of the country
have significantly different costs of attendance?
9. Use the College Scorecard data to determine if different
regions of the country have significantly different costs
of attendance (same as Problem 8) and fill out a data
request form in order to extract the appropriate data.
Use the template from the chapter as a guide.
page 58
Lab 2-1 Create a Request for Data

Extraction
One of the biggest challenges you face with data analysis
is getting the right data. You may have the best questions
in the world, but if there are no data available to support
your hypothesis, you will have difficulty providing value.
Additionally, there are instances in which the IT workers
may be reluctant to share data with you. They may send
incomplete data, the wrong data, or completely ignore
your request. Be persistent, and you may have to look for
creative ways to find insight with an incomplete picture.
Company summary
Sláinte is a fictional brewery that has recently gone
through big changes. Sláinte sells six different products.
The brewery has only recently expanded its business to
distributing from one state to nine states, and now its
business has begun stabilizing after the expansion. With
that stability comes a need for better analysis. You have
been hired by Sláinte to help management better
understand the company’s sales data and provide input
for its strategic decisions.
Data
Data request form
Technique
No technical experience is necessary for this lab.
Software needed
Word processor

Part 2: Generate a request for data.
Part 3: Assess the data you receive.

One of Sláinte’s first priorities is to identify its areas of
success as well as areas of potential improvement. Your
manager has asked you to focus specifically on sales
data at this point. This includes data related to sales
orders, products, and customers.
Q1. Given that you are new and trying to get a
grasp on Sláinte’s operations, list three
questions related to sales that would help
you begin your analysis. For example, how
many products were sold in each state?
Q2. Now hypothesize the answers to each of the
questions. Remember, your answers don’t
have to be correct at this point. They will help
you understand what type of data you are
looking for. For example: 500 in Missouri,
6,000 in Pennsylvania, 4,000 in New York,
etc.
Q3. Finally, for each question, identify the
specific tables and attributes that are
needed to answer your questions. Use Lab
Exhibit 2-1A for guidance on what tables and
attributes are available. For example, to
answer the question about state sales, you
would need the [Customer_St] attribute that is
located in the [Customer] master table as well
as the [Sales_Order_Quantity_Sold] attribute
in the [Sales] table. If you had access to store
or distribution center location data, you may
also look for a [State] field there, as well.
page 59
LAB EXHIBIT 2-1A

Part 2: Generate a Request for Data
Now that you’ve identified the data you need for your
analysis, complete a data request form.
1. Open the Data Request Form.
2. Enter your contact information.
3. In the description field, identify the tables that you’d
like to analyze, along with the time periods (e.g., past
month, past year, etc.).
4. Select a frequency. In this case, this is a “One-off
request.”
5. Enter a request date (today) and a required date (one
week from today).
6. Choose a format (spreadsheet).
7. Indicate what the information will be used for in the
appropriate box (internal analysis).
8. Take a screenshot (label it 2-1A) of your completed
form.
Part 3: Perform an Analysis of the

Data
After a few days, Rachel, an IT worker, responds to your
request. She gives you the following tables and attributes:
LAB EXHIBIT 2-1B

Sales_Subset Table
Description of
Attribute
Attribute
Sales_Order_ID (PK) Unique identifier for
each sales order
Sales_Order_Date The date of the sales
order, regardless of the
date the order is
entered
Sales_Employee_ID (FK) Unique identifier (from
Employee_Listing
table) for person who
created the record
Sales_Order_Quantity_Sold Sales order line
quantity
Product_Sale_Price Sales order line price
per unit
FGI_Product Table
Attribute Description of Attribute
Product_Code (PK) Unique identifier for each
product
Product_Description Product description (plain
English) to indicate the name
or other identifying
characteristics of the product
Product_Sale_Price Price per unit of the associated
product
You may notice that while there are a few attributes

that may be useful in your sales analysis, the list may be
incomplete and be missing several values. This is normal
with data requests.
Q4. Take a moment and identify any attributes
that you are missing from your original
request that would be necessary to answer
the question “How many products were sold
in each state?”
page 60
Q5. Evaluate your original questions and

responses. Can you still answer the original
questions that you identified in step 1?
Q6. Is there another question you could answer
from the data Rachel provided?
End of Lab
Lab 2-2 Use PivotTables to

Denormalize and Analyze the Data
Efficient relational databases contain normalized data.
That is, each table contains only data that are relevant to
the object, and tables’ relationships are defined with
primary key/foreign key pairs. For example, each record
in a customer table is assigned a unique ID (e.g.,
customer 152883), and the remaining attributes (e.g.,
customer address) describe that customer. In a sales
order table, the only customer data you find is a foreign
key pointing to the customer (e.g., customer 152883) we
are selling merchandise to. The foreign key value
connects the sales order record to the customer record
and allows any or all of the linked attributes to appear on
the sales order form or report.
With Data Analytics, efficient databases are not as
helpful. Rather, we would like to “denormalize” the data or
combine all of the related data into one large file that can
be easily evaluated for summary statistics or be used to
create meaningful PivotTables. Excel calls this the Internal
Data Model. In Access (or other relational databases,
such as SQLite or SQL Server), we create a query. This
lab will take you through this process. This lab will help
you learn how to create relationships between related
spreadsheets in Excel using Excel’s Internal Data Model.
The Internal Data Model is available in Excel for PC
versions from 2013 onward, but the screenshots are all
taken from Excel 2016. If you are using a different version
of Excel, you may experience slightly different screens.
This lab will also help prepare you for using the
Internal Data Model (alternative 2) in future labs to
transform data, as well as to aid in understanding of
primary and foreign key relationships.
Company summary
through big changes. Sláinte sells six different products.
distributing from one state to nine states, and now its
business has begun stabilizing after the expansion. With
that stability comes a need for better analysis. One of
Sláinte’s first priorities is to identify its areas of success,
as well as areas of potential improvement.
Data
Sláinte dataset
Technique
Some experience with relational databases,
spreadsheets, and PivotTables is useful for this lab.
Relying on Appendix C about PivotTables might be
helpful.
Some experience with SQL is useful for this lab. Relying
on Appendix D on SQL might be helpful.
Software needed
Excel
Access or SQLite
+ Shift + 4)
page 61

Part 2: Master your data and prepare it for analysis in
Excel.
You will be presented with four different alternatives,
the first of which is “to do nothing” to master the data.
The following three alternatives present three
different methods of ETL to create the same results.
Work through each ETL alternatives (2 through 4)
to learn different methods of ETL and assess
when you would prefer to use each.
Alternatives 1 through 3 assume that the data is
already stored in Excel.
Alternative 4 assumes that the data is stored in a
relational database. You can opt to use the
Access database file or the SQLite file for this
alternative.
Part 3: Perform an analysis using PivotTables.
Part 4: Address and refine your results.
Part 5: Communicate your findings.

Sláinte has brought you in to help determine potential
areas for sales growth in the next year. Additionally,
management has noticed that the company’s margins
aren’t as high as they had budgeted and would like you to
help identify some areas where they could improve their
pricing, marketing, or strategy.
Specifically, they would like to know how many of each
product were sold, the product’s actual name (not just the
product code), and the months in which different products
were sold.
Q1. Given Sláinte’s request, identify the data
attributes and tables needed to answer the
question. You can rely on Lab Exhibit 2-2A to
help answer this question.
Part 2: Master the Data: Prepare Data

for Analysis in Excel
The requested Sláinte data are available in the
Slainte_Subset.xlsx file and include the following tables
and fields, presented in a UML diagram:
LAB EXHIBIT 2-2A

Depending on your desired analysis, there are a few
alternative approaches for ETL that you could use to
prepare the data for analysis:
1. Do nothing.
2. Use the Excel Internal Data Model.
3. Merge the data into a single table using Excel Query
Editor.
We will work through each alternative throughout this
lab to expose you to different methods of ETL.
page 62
Alternative 1 (data stored in Excel): Do nothing

If you are simply trying to calculate statistics or make
comparisons using attributes within a single table, there is
no need to transform the tables. Simply load the table,
make sure the data are clean, and proceed to analysis.
For example, to find the total number of each item
sold, you would need only the [Sales_Subset] table and
its attributes [Product_Code] and
[Sales_Order_Quantity_Sold].
Q2. List two other questions that could be
answered using only a single table from the
Slainte dataset.
Alternative 2 (data stored in Excel): Use the

Excel Internal Data Model
For analyses that require two or more tables, you need to
define the relationships between the tables before you
can proceed with your analysis.
With the Excel’s Internal Data Model, you can create
these relationships directly within Excel.
For example, if you want to find the total number of
each item sold and show the product name instead of just
the code, you would need [Sales_Order_Quantity_Sold]
from the [Sales_Subset] table and [Product_Description]
from the [FGI_Product] table. These two tables are joined
together on the [Product_Code] primary/foreign key.
1. Open the Slainte_Subset.xlsx file.
2. Click the Data tab on the ribbon.
3. Click the Relationships button in the Data Tools
section.
4. In the Manage Relationships window, click New. . .
5. Create a relationship between [Sales_Subset] and
[FGI_Product] as shown in Exhibit 2-2B. Start with the
table that contains the foreign key, then choose the
related table that contains the primary key. In this case:
a. Table: [Sales_Subset]
b. Related Table: [FGI_Product]
c. Column (Foreign): [Product_Code]
d. Related Column (Primary): [Product_Code]
LAB EXHIBIT 2-2B

Define the Primary Key/Foreign Key Relationships in
Excel
6. Click OK to save the relationship. The window will
close, and you will return to the Manage Relationships
window. Click New. . .
7. Now create a relationship between [Sales_Subset] and
[Customer]. In this case:
a. Table: [Sales_Subset]
b. Related Table: [Customer]
c. Column (Foreign): [Customer_ID]
d. Related Column (Primary): [Customer_ID]
8. Click OK to save the relationship.
9. Take a screenshot (label it 2-2A) of the Manage
Relationships window with both relationships created.
page 63
10. Click Close in the Manage Relationships
window to return to the spreadsheets. While the
spreadsheets do not appear to have changed with the
new relationships, we have created a powerful engine
for analyzing our data. We will have access to any of
the records and related fields in any of the tables
without additional work, such as Find and Replace or
VLookup.
11. Save your workbook as Slainte_Relationships.xlsx.
This spreadsheet is now ready for analysis. But before
we analyze the data, work through Alternative 3 to learn a
different method of ETL.
Q3. How comfortable are you with identifying
primary key/foreign key relationships?
Alternative 3 (data stored in Excel): Merge the

data into a single table using Excel Query
Editor
While relationships are incredibly useful when dealing
with multiple tables, there are times when it is useful to
combine all of the data together in one table. Both queries
and PivotTables can be more straightforward when you
don’t have to continually define the relationships. The
downside to working with a single table is that you must
work with a larger file size and there are a lot of redundant
data.
1. Create a new blank spreadsheet in Excel.
2. Click the Data tab on the ribbon.
3. Click the New Query button in the Get & Transform
Data section.
This button is named Get Data in Excel 365.
4. Choose From File > From Workbook.
5. Locate the Slainte_Subset.xlsx file on your computer,
and click Import.
6. In the Navigator, check Select multiple items, then
check the three tables to import, shown in Exhibit 2-2C:
a. [Customer]
b. [FGI_Product]
c. [Sales_Subset]
LAB EXHIBIT 2-2C

Select Multiple Tables to Join
page 64
7. Click Edit. The Power Query Editor will
appear.
This button is named Transform Data in Excel 365.
8. In the queries menu (which can be found on your far left
within the Power Query Editor), you will find the three
tables you just imported. Click Sales_Subset to see a
preview of the Sales_Subset data.
We now want to merge all three of these tables
together, using Sales_Subset as our base. We’ll start with
merging the Sales_Subset and Customer tables.
9. With Sales_Subset as the active table, click the Home
tab, then choose Merge Queries from the Combine
section. From the dropdown menu, again select Merge
Queries. A new Merge window will appear.
10. In the Merge window, the [Sales_Subset] query will
appear at the top.
LAB EXHIBIT 2-2D

Select the Primary and Foreign Keys in the Merge
Window to Create a Large Table in Excel’s Query
Editor
11. To join the [Sales_Subset] and [Customer] queries,

do the following:
a. Select the [Customer] query from the middle drop-
down list.
b. Click the [Customer_ID] column in both the
[Sales_Subset] and [Customer] queries.
c. From the Join Kind drop-down list, choose Left
Outer (all from first, matching from second). This
means [Sales_Subset]. [Customer_ID] will be
identified as the foreign key.
d. Click OK to return to the Query Editor.
e. If your new column isn’t already namedCustomer,
double-click the newly added attribute title and
rename it [Customer].
f. Finally, click the opposing arrows icon next to the
[Customer] column title. Select Expand and click
OK. The customer attributes have now been added to
the [Sales_Subset] query.
12. To join the [Sales_Subset] and [FGI_Product] queries,
do the following:
a. Click Merge Queries again from the Home tab
(ensure that Sales_Subset is still the active table).
b. Select [FGI_Product] from the middle drop-down list.
c. Click the [Product_Code] column in both the
[Sales_Subset] and [FGI_Product] queries.
d. From the Join Kind drop-down list, choose Left
Outer (all from first, matching from second). This
means [Sales_Subset]. [Product_Code] will be
identified as the foreign key.
page 65
e. Click OK to return to the Query Editor.
f. Double-click the newly added [NewColumn] attribute
title and rename it [Product].
g. Finally, click the opposing arrows icon next to the
[Product] column title. Select Expand and click OK.
The product attributes have now been added to the
[Sales_Subset] query.
13. Maximize the Query Editor window, and take a
screenshot (label it 2-2B).
14. Click Close & Load twice (top left corner of the Power
Query Editor) to return to Excel.
15. In the Workbook Queries pane on the right (named
Queries & Connections in Excel 365), select
Sales_Subset to view the data. (If the data has not
automatically imported into a table, right-click
[Sales_Subset] and choose Load To. . .).
16. Rename the active sheet with the full set of data
[Sales_Order_Merge].
Note: You can also directly load your merged table into a
PivotTable if that is the analysis you’re going to perform.
17. Save your workbook as Slainte_Merge.xlsx.
Q4. Have you used the Query Editor in Excel
before? Double-click the [Sales_Subset] query
and click through the tabs on the ribbon. Which
options do you think will be useful in the future?
Alternative 4 (data stored in a relational

database): Use SQL queries to retrieve the data.
Extracting data with SQL queries can not only be very
efficient, but sometimes necessary. There are instances
where datasets will be too large for Excel, and Access or
another query tool (such as SQLite) will be the only way
to analyze the data effectively.
1. Open the Slainte_Subset.accdb file.
2. Open the SQL editor by navigating to the Create tab on
the ribbon.
3. Click the Query Design button in the Queries section to
open the SQL Designer. Here you should see three
tables.
4. Click Close on the Show Table window.
5. In the top left corner, click SQL icon and then select
SQL View to open the SQL Editor.
6. In the SQL Editor, type the following lines of code:
SELECT Sales_Subset.*, FGI_Product.*, Customer.*
FROM Customer
INNER JOIN
(FGI_Product INNER JOIN Sales_Subset
ON FGI_Product.Product_Code =
Sales_Subset.Product_Code)
ON Customer.Customer_ID =
Sales_Subset.Customer_ID;
7. Click Run from the Design tab on the Ribbon to view
your combined query output.
8. Take a screenshot of your query output (label it 2-2C).
9. Right-click the Query and click Save to save your query
as Slainte_Merge.
page 66
10. From here you can either click External
Data menu to export your data or you can close your
file without exporting. For practice, export to Excel.
Then, open Excel and choose Data > Get Data > From
Database > From Microsoft Access Database, then
navigate to your database and import the query.
Part 3: Perform an Analysis Using

PivotTables and Queries
Now that the data have been organized, you’re ready for
some basic analysis. Given the sales data, management
has asked you to prepare a report showing the total
number of each item sold each month between January
and April 2020. This means that we should create a
PivotTable with a column for each month, a row for each
product, and the sum of the quantity sold where the two
intersect.
1. Open any of the files you created in Part 2 in Excel,
and click the Insert tab on the ribbon.
2. Click PivotTable in the Tables section.
3. In the Create PivotTable window click Add this data
to the Data Model. Note: If you have only one table,
choose Select a table or range and choose your
sheet.
4. Click OK to create the PivotTable. A PivotTable Fields
pane appears on the right.
Note: If at any point while working with your
PivotTable, your PivotTable Fields list disappears, you
can make it reappear by ensuring that your active cell
is within the PivotTable itself. If the Field List still
doesn’t reappear, navigate to the Analyze tab in the
Ribbon, and select Field List.
5. Click the arrow toggle next to each table to show the
available fields. If you don’t see your three tables, click
the All option directly below the PivotTable Fields
pane title.
6. Take a screenshot of your PivotTable field list (label it 2-
2D).
7. Because you defined relationships or merged the tables
in Part 2, you can drag any of the attributes from your
list of fields to their respective Filters, Columns, Rows,
or Values. Do that now:
a. Columns: [Sales_Order_Date] (Month) from
[Sales_Subset]. Note: When you add a date, Excel
will automatically try to group the data by Year,
Quarter, etc. For now, remove the other options.
b. Rows: [Product_Description] from [FGI_Products].
Note: If you imported the data from the Access
database query, it will be named
Product.Product_Description
8.
a. Values: [Sales_Order_Quantity_Sold] from
[Sales_Subset].
b. Filters: None.
9. Finally, to show only the four months from January to
April, click the drop-down arrow next to Column Labels
in the PivotTable (not in the Field list) and uncheck Nov
and Dec.
10. Optional step: Clean up your PivotTable. Rename
labels and the title of the report to something more
useful.
11. Take a screenshot of your PivotTable (label it 2-2E).
12. Save a copy of your workbook as Slainte_Pivot.xlsx.
To perform a similar, but less flexible analysis in Access,
do the following:
1. Open your Slainte_Subset.accdb file from Part 2.
2. Click Create > Query Design. Close the window that
appears.
3. Click SQL View in the top-left corner.
page 67
4. Enter the following query:
SELECT Product_Description,
Sum(Sales_Order_Quantity_Sold) AS Total_Sales
FROM Slainte_Merge
WHERE Sales_Order_Date Between #1/1/2020# And
#4/30/2020#
GROUP BY Product_Description;
5. Click Run to show the results.
6. Take a screenshot (label it 2-2F).
7. Save your query as Total_Sales_By_Product and
close your database.
Part 4: Address and Refine Your

Results
Now that you’ve completed a basic analysis to answer
management’s question, take a moment to think about
how you could improve the report and anticipate
questions your manager might have.
Q5. If the owner of Sláinte wishes to identify which
product sold the most, how would you make
this report more useful?
Q6. If you wanted to provide more detail, what
other attributes would be useful to add as
additional rows or columns to your report, or
what other reports would you create?
Part 5: Communicate Your Findings

Let’s make this easy for others to understand using
visualization and explanations.
Q7. Write a brief paragraph about how you would
interpret the results of your analysis in plain
English. For example, which data points stand
out?
Q8. In Chapter 4, we’ll discuss some visualization
techniques. Describe a way you could present
these data as a chart or graph.
End of Lab
Lab 2-3 Resolve Common Data

Problems in Excel and Access
There are several issues with this dataset that we’ll need
to resolve before we can process the data. This will
require some cleaning, reformatting, and other
techniques.
Company summary
LendingClub is a peer-to-peer marketplace where
borrowers and investors are matched together. The goal
of LendingClub is to reduce the costs associated with
these banking transactions and make borrowing less
expensive and investment more engaging. LendingClub
provides data on loans that have been approved and
rejected since 2007, including the assigned interest rate
and type of loan. This provides several opportunities for
data analysis.
Data
Download the LendingClub datasets: ApproveStats
from the Connect website
page 68
Technique
Some experience with Excel is useful for this lab.
Software needed
Excel
+ Shift + 4)

Understand and clean the data to enable analysis for
various problems.

You’ve already identified some analysis questions for
LendingClub in Chapter 1. Here, you’ll focus on data
quality. Think about some of the common issues with data
you receive from other people. For example, is the date
field in the proper format? Do number fields contain text
or vice versa?
Q1. What do you expect will be major data quality
issues with LendingClub’s data?

The LendingClub collects different sets of data, including
LoanStats for approved loans and RejectStats for rejected
loans. There are significantly more data available for
LoanStats.
For this lab, we will analyze approved loans using the
Lab_2-3_ApproveStats.xlsx file.
There are 145 different attributes. To save some time,
we’ve identified 19 of the most interesting in Lab Exhibit
2-3A.
LAB EXHIBIT 2-3A
member_id Membership ID
loan_amnt Requested loan amount
purpose Loan purpose: e.g., wedding,
medical, debt_consolidation, car
zip_code The first three digits of the
applicant’s zip code
addr_state State
dti Debt-to-income ratio
delinq_2y Late payments within the past 2
years
Source: LoanStatsXXXX.csv
Q2. Given this list of attributes, what types of
questions do you think you could answer
regarding approved loans? (If you worked
through Lab 1-2, what concerns do you have
with the data’s ability to predict answers to the
questions you identified in Chapter 1)?
page 69
1. Take a moment and explore the data.

Q3. Is there anything in the data that you think will
make analysis difficult? For example, are there
any special symbols, nonstandard data, or
numbers that look out of place?
Q4. What would you do to clean the data in this file?
Let’s identify some issues with the data.
There are many attributes without any data, and that may
not be necessary.
The [int_rate] values are written in ##.##%, but analysis
will require #.####.
The [term] values include the word “months,” which
should be removed for numerical analysis.
The [emp_length] values include “n/a”, “<”, “+”, “year”,
and “years”—all of which should be removed for
numerical analysis.
Dates, including [issue_d], can be more useful if we
expand them to show the day, month, and year as
separate attributes. Dates cause issues in general
because different systems use different date formats
(e.g., 1/9/2009, Jan-2009, 9/1/2009 for European dates,
etc.), so typically some conversion is necessary.
First, remove the unwanted data:

2. Save your file as “Loans2007-2011.xlsx” to take
advantage of some of Excel’s features.
3. Delete the first row that says “Notes offered by
prospectus. . .”.
4. Delete the last two rows that include “Total amount
funded. . .”.
5. Delete columns that have no values, including [id],
[member_id], and [url].
6. Repeat for any other blank columns or unwanted
attributes.
Next, fix your numbers:

7. Select the [int_rate] column.
8. In the Home tab, go to the Number section and change
the number type from Percentage to General using the
drop-down menu.
9. Repeat for any other attributes with percentages.
10. Take a screenshot (label it 2-3A) of your partially
cleaned data file.
Then, remove any words from numerical values:

11. Select the [term] column.
12. Use Find & Replace (Ctrl+H or Home > Editing >
Find & Select > Replace) to find the words “months”
and “month” and replace them with a null/blank value.
Important: Be sure to include the space before the
words and go from the longest variation of the word to
the shortest. In this case, if you replaced “month” first,
you would end up with a lot of values that still had the
letter “s” from “months.”
13. Now select the [emp_length] column and find and
replace the following values:
Original Value New Value
na or n/a 0
< 1 year 0
1 year 1
2 years 2
3 years 3
4 years 4
5 years 5
6 years 6
7 years 7
8 years 8
9 years 9
10+ years 10
, (comma) (blank)
page 70
14. Take a screenshot (label it 2-3B) of your partially

cleaned data file, showing the [term] column.
Analytics Tool: Python

Note: Finding and replacing 13 values by hand may
be tedious, but it is efficient for a one-off analysis
and a small file. If you plan to re-perform this
analysis multiple times or find and replace dozens of
items or you have a file that is larger than Excel can
handle, you’re better off using a scripting language,
such as Python. You can download Python free from
python.org, and a quick search on Google will help
you find tutorials to start with the basics.
Here’s what the script would look like for the find
and replace function where you would list the
original value as item and the replacement value as
replacement:
import csv
ifile = open(‘file’, 'rb')
reader = csv.reader(ifile,delimiter='\t')
ofile = open('file', 'wb')
writer = csv.writer(ofile, delimiter='\t')
s = ifile.read()
for item, replacement in zip(findlist, replacelist):
s = s.replace(item, replacement)
s = s.replace(item, replacement)
ofile.write(s)
ifile.close()
ofile.close()
Finally, transform those dates:

15. Right-click the column to the right or the [issue_d]
column, and choose Insert to add a blank column.
16. Name the new column [issue_month].
17. Use the =MONTH([column address for issue_d])
formula to extract the month from the date in your new
column and copy your formula to the bottom of the
sheet. You should see a month number value in each
cell. If it still has a date format, change the number
format to General in the Home tab.
18. Now convert the formulas to data values. Select the
new [issue_month] column that contains your formula.
page 71
19. First, copy all of the values in the
[issue_month] column (you can use the keyboard
shortcut Copy [Ctrl+C]) and then replace the existing
data in the same column with the values by using the
following keyboard shortcut: Paste Special
[Ctrl+Alt+V]. Choose Values [V], then click OK.
20. Save your file.
21. Add another blank column and name it [issue_year].
22. Use the =YEAR([column address for issue_d]) formula
to extract the year from the date in your new column
and copy your formula to the bottom of the sheet. You
should see a year number value in each cell. If it still
has a date format, change the number format to
General in the Home tab.
23. Now, convert the formula to data values like you did
with the Issue_month column (Select the new
[issue_year] column that contains your formula, then
Copy [Ctrl+C] and Paste Special [Ctrl+Alt+V].
Choose Values [V]. Click OK).
24. Save your file.
25. Take a screenshot (label it 2-3C) of your cleaned data
file, showing the new date columns.
Q5. Why do you think it is useful to reformat and
extract parts of the dates before you conduct
your analysis? What do you think would happen
if you didn’t?
Q6. Did you run into any major issues when you
attempted to clean the data? How did you
resolve those?
End of Lab
Lab 2-4 Generate Summary
Statistics in Excel
When you’re working with a new or unknown set of data,
validating the data is very important. When you make a
data request, the IT manager who fills the request should
also provide some summary statistics that include the
total number of records and mathematical sums to ensure
nothing has been lost in the transmission. This lab will
help you calculate summary statistics in Excel.
Company summary
LendingClub is a peer-to-peer marketplace where
borrowers and investors are matched together. The goal
of LendingClub is to reduce the costs associated with
these banking transactions and make borrowing less
expensive and investment more engaging. LendingClub
provides data on loans that have been approved and
rejected since 2007, including the assigned interest rate
and type of loan. This provides several opportunities for
data analysis.
Data
Download the LendingClub dataset: Lab_2-
4_ApproveStats.xlsx from the Connect website
Technique
Software needed
Excel
+ Shift + 4)
Calculate summary statistics using Excel.
page 72
Quickly View Summary Statistics in Excel

For basic validation, we’ll use Excel. Remember, there is
a limitation on the number of records that Excel can
handle, so this is best for smaller- to medium-sized files.
Excel’s toolbar at the bottom of the window provides quick
access to a summary of any selected values.
1. Open Lab_2-4_ApproveStats.xlsx.
2. Select the [loan_amnt] column. At the bottom right of
the window, you will see the Average, Count, and Sum
calculations, shown in Lab Exhibit 2-4A. Compare those
to the validation given by LendingClub:
3. Funded loans: $3,503,840,175
Number of approved loans: 235,629
LAB EXHIBIT 2-4A

Summary Statistics Provided by the Excel Toolbar at the Bottom
Showing Average, Count, and Sum

Q1. Do your numbers match the numbers
provided by LendingClub? What explains the
discrepancy, if any?
4. Right-click on the summary toolbar and choose
Numerical Count from the list. You should now see
four values in the bar.
Q2. Does the Numerical Count provide a more
useful/accurate value for validating your data?
Why or why not do you think that is the case?
Q3. What other summary values might be useful for
validating your data?
5. Take a screenshot (label it 2-4A) showing your
expanded summary toolbar with four (or more) values.
View More Detailed Summary Statistics in Excel

If we want to view more summary statistics than just the
ones provided in the summary toolbar, or even if we
would prefer to retrieve an output of summary statistics
instead of having to select the data to view it, we can run
the “Descriptive Statistics” tool using the Data Analysis
ToolPak.
6. Ensure that the Data Analysis ToolPak is added in to
your Excel (for instructions on doing so are provided in
Appendix B at the end of the textbook).
7. Click the Data tab on the ribbon and select Data
Analysis to open the Data Analysis ToolPak options.
page 73
LAB EXHIBIT 2-4B
Data tab on the Ribbon
8. Select Descriptive Statistics and click OK.
LAB EXHIBIT 2-4C

Data Analysis ToolPak options
9. For the input range, select the column of data

associated with loan_amnt.
10. If you selected the First Row (the label that says “Loan
Amnt”), place a checkmark next to Labels in First Row.
11. Place a check mark next to Summary Statistics.
12. Click OK.
LAB EXHIBIT 2-4D

Descriptive Statistics Options
13. The output with the summary statistics will be provided

on a new spreadsheet.
14. Take a screenshot (label it 2-4B) showing your
Summary Statistics output.
End of Lab
Lab 2-5 College Scorecard

Extraction and Data Preparation
This lab will help you learn how to extract data from a text
file in preparation for analysis in Excel. This lab is in
preparation for future labs in Chapter 3 that will analyze
college scorecard data.
page 74
Data summary
The data used are a subset of the College Scorecard
dataset that is provided by the U.S. Department of
Education. These data provide federal financial aid and
earnings information, insights into the performance of
schools eligible to receive federal financial aid, and the
outcomes of students at those schools. You can learn
more about how the data are used and view the raw data
yourself at https://collegescorecard.ed.gov/data/.
However, for this lab, you should use the text file provided
to you.
Data
CollegeScorecard Datasets: CollegeScorecard_RawData
Technique
Software needed
Text Editor (Windows: Notepad; Mac: TextEdit)
Excel
+ Shift + 4)

Extract data into a text editor and transform it into
structured, ready-to-analyze data in Excel.

Because this lab is focused on mastering the data, the
question has been identified for you. We will begin with a
simple question with two variables, SAT average and
completion rate for first-time, full-time students at four-
year institutions.
1. Open the text file CollegeScoreCardRawData.txt.
2. Select all of the data in the text file and copy them.
3. Open a new Excel workbook. With the active cell as A1,
paste the text data.
4. Take a screenshot (label it 2-5A).
5. The data defaulted to pasting into only column A. To
view each attribute in its own column, you will need to
parse the data using the Text to Columns feature in
Excel. Navigate to the Data tab in Excel’s ribbon.
6. Make sure that the all of the data in column A are
selected (not just the first row of data), then click the
Text to Columns button to open up a wizard to guide
your data transformation.
7. The wizard provides two options: delimited and fixed
width. Delimited means separated, and in Excel (and
other applications), it references a character that
separates (or delimits) data.
Q1. By looking through the data in the text file, what
do you think the delimiter is?
8. Leaving delimited checked (as is the default), click
Next in the wizard, and select the appropriate delimiter.
Make sure to un-check the default option, Tab.
9. Click Finish in the wizard.
10. Take a screenshot (label it 2-5B).
page 75
11. To ensure that you captured all of the data
through the extraction from the txt file, we need to
validate it. Validate the following checksums:
You should have 7,704 records (rows).
Compare the attribute names (column headers) to the
attributes listed in the data dictionary (found in
Appendix K of this chapter). Are you missing any, or do
you have any extras?
The average SAT score should be 1,059.07 (this is
leaving NULL values as NULL).
Q2. In the checksums, you validated that the
average SAT score for all of the records is
1,059.07. When we work with the data more
rigorously, several tests will require us to
transform NULL values. If you were to transform
the NULL SAT values into 0, what would
happen to the average (would it stay the same,
decrease, or increase)? How would that change
to the average affect the way you would
interpret the data? Do you think it’s a good idea
to replace NULL values with 0s in this case?
12. Now that the data have been validated, you can clean
the data. How you clean the data is determined by the
question you intend to answer. In this case, we’re
preparing our data to run a regression test using the
two attributes SAT_AVG and C150_4. As you’ll learn in
Chapter 3, a regression test won’t run with non-numeric
values (i.e., we can’t leave the NULL values in, and we
can’t transform them to blanks). Earlier you discussed
the cons of replacing NULL values with 0s.
To avoid the issues with NULL, blanks, and 0s, we
will remove all of the records that contain NULL values
in either SAT_AVG or C150_4. Do so.
13. Perform a =COUNT() to verify the number of records
that remain after removing all records associated with
NULL values in SAT_AVG or C150_4. 1,271 records
should remain.
14. Take a screenshot that includes the data and the result
of the =COUNT() function. (label it 2-5C).
Your data is now ready for the test plan. This lab will
continue in Chapter 3.

Dillard’s Store Data: How to Create
an Entity-Relationship Diagram
Company summary
Dillard’s is a department store with approximately 330
at finance.yahoo.com (Ticker symbol = DDS) and the
Data
your instructor for login credentials.
page 76
Software needed
Microsoft SQL Server Management Studio (available on
the Remote Desktop at the University of Arkansas)

Create an ERD (entity-relationship diagram), which
provides some quick information on the data that’s
provided in the database. In one diagram, you can view
all tables to see the entire database, or you can pick just
the two you’re working with to focus on those attributes.

Dillard’s is trying to figure out when its customers spend
more on individual transactions. We ask questions
regarding how Dillard’s sells its products.
Q1. How would a view of the entire database or
certain tables out of that database allow us to
get a feel for the data?
Q2. What types of data would you guess that
Dillard’s, a retail store, gathers that might be
useful? How could Dillard’s suppliers use
these data to predict future purchases?
Part 2: Master the Data and Part 3:

Perform an Analysis of the Data
For this lab, we will create an ERD (entity-relationship
diagram) to view the structure of the database.
1. Log on to Remote Desktop at the University of
Arkansas.
2. Press the Windows key to open the Start menu.
Microsoft SQL Server Management Studio will be on
the right of the menu that appears. Open SQL Server to
access the WCOB_DILLARDS data.
3. Input the Server Name in the Connect to Server
window that your instructor provided you.
page 77
4. Leave the default for authentication to
Windows Authentication, and click Connect.
5. Expand the Databases folder in the Object Explorer
window.
6. Scroll down to the WCOB_DILLARDS database and

expand it.
7. Right-click Database Diagrams to reveal the below
window.
Source: Microsoft SQL Server Management Studio
8. Select New Database Diagram.

9. A window indicating that you cannot create tables will
appear, click OK. Because we do not need to create or
edit any tables, viewing them is sufficient.
Source: Microsoft SQL Server Management Studio.
page 78
10. Select the tables you would like to view. For
this lab, highlight all of the tables and select Add.
11. Take a screenshot (label it 2-6A). To ensure you can
see all of the tables in your screenshot, you can select
View < Zoom and select To Fit.
Part 4: Address and Refine Results

Q3. What is the primary key for the TRANSACT
table? What is the primary key for the SKU
table?
Q4. How do we connect the SKU database to the
TRANSACT table? How do we join tables
from two different related tables?
End of Lab

Dillard’s Store Data: How to Preview
Data from Tables in a Query
Company summary
Data
Software needed

Learn how to get a snippet of the data to better
understand the data fields, what they contain, and what
their data structure looks like to inform additional queries
and database analysis.

Data Analytics requires a lot of give and take. Often, you
will learn a bit, modify the search, modify the analysis,
and try again, especially while you are getting to know the
data. All the while, you are asking yourself questions,
trying to make your search efficient and effective.
Questions like these:
Q1. How would a diagram of the entire database
allow us to get a feel for the data? How would
looking at examples of the actual data stored
in some of the tables help you understand the
data?
page 79
Q2. What types of data would you guess that

Dillard’s, a retail store, gathers that might be
useful? How could Dillard’s suppliers use
these data to predict future purchases?

Arkansas.
2. Open Microsoft SQL Server Management Studio to
access the WCOB_DILLARDS data.
window that you were provided through the
Walton.uark.edu/enterprise website.
4. Leave the default for authentication to Windows

Authentication and click Connect.
5. Select New Query from the menu at the top of the SQL
Server application.
6. If the drop-down indicating which database you are

intending to query doesn’t say “WCOB_DILLARDS”
(e.g., it frequently defaults to “Master”), select the drop-
down window and scroll down to WCOB_DILLARDS,
then click Enter. You could also type
WCOB_DILLARDS instead of waiting to scroll to it.
page 80
7. Because this dataset is massive, it can take
a very long time for the system to return the complete
set of data for some of the bigger tables (such as
TRANSACT). If you would like to view just the top few
rows of a dataset to get the feel for what type of data is
in the table, you can do so with a query.
In the SELECT line, you can type TOP # before the
columns you would like to see. Any type of filtering,
aggregating, and ordering will still work through the rest
of the query, but selecting the top few will help the
query run faster by returning a subset of the result.
8. To view the top 10 rows in the TRANSACT table, type
the following query into the query window:
SELECT TOP 10 *
FROM TRANSACT
9. To see the result of the query, click Execute. F5 also
works to run queries as a PC shortcut.
10. Take a screenshot of your results (label it 2-7A).

11. When you look at these results, you may wonder what
some of the attributes represent. For example,
TRAN_TYPE only returns values with P. To view other
types of values, you could filter out any record that has
a TRAN_TYPE of P. Execute the following query:
SELECT TOP 10 *
FROM TRANSACT
WHERE TRAN_TYPE <> 'P'

Q3. What do you think ‘P’ and ‘R’ represent in the
TRAN_TYPE table? How might transactions
differ if they are represented by ‘P’ or ‘R’?
Q4. What benefit can you gain from selecting only
the top few rows of your data, particularly
from a large dataset?
End of Lab

Dillard’s Store Data: Connecting
Excel to a SQL Database
Company summary
Arkansas and the Walton College of Business, page 81
which may be why he shared transaction data
with us to make available for this lab and labs throughout
this text.
Data
Software needed
Excel 2016 (available on the Remote Desktop at the
University of Arkansas)

Learn how to access databases, run queries, and
perform analyses in Excel.

Because, as accountants, we are most familiar with
Microsoft Excel, we’d like to learn how to access the data
and run queries in Excel. But the question is still why use
Excel?
Q1. What can you do in Excel that is much more
difficult to do in other data management
programs?
Q2. Because most accountants are familiar with
Excel, name three data management
functions you can do easier in Excel than any
other program. How does that familiarity help
you with your analysis?

Arkansas.
Connecting Excel to a SQL Server Database

While executing queries in SQL Server is a great method
for viewing data, if you want to eventually load data into
Excel for additional analysis or visualization, it is easiest
to write the queries directly in Excel. This puts the entire
ETL (or ELT) process in one tool, Excel.
2. From the Remote Desktop, click the windows key to
access the Start Menu and open a new Excel
workbook.
3. From the Data tab, click New Query > From Database
> From SQL Server Database.
page 82
4. In the Microsoft SQL database pop-up window, input
the server name that you were provided through the
Walton.uark.edu/enterprise website. The database
name is WCOB_DILLARDS.
5. Click OK.
6. If presented an option to change credentials, keep the
default to use your current credentials, and then click
Connect.
7. If an Encryption Support window pops up, click OK.

8. The tables in the WCOB_DILLARDS database are
available for you to select in the Navigator window.
Click once on STORE to preview the data.
page 83
9. The data will preview on the right side of the Navigator

window. Click Load to load the data into a table in
Excel.
As long as the dataset that you have loaded is
under the Excel row limit of 1,048,576, the entire table
will be available for you to work with in Excel. You can
analyze the data using Excel’s formulas, functions, and
statistical tools, as well as create PivotTables and
charts.
10. Create a PivotTable for this set of data by selecting all
of the data from the Store table and then clicking
PivotTable on the Insert tab of the Excel ribbon.
page 84
11. We can quickly view a count of how many
stores are in each state. Drag and drop STATE into the
ROWS section of the PivotTable Fields window and
STORE into the VALUES section.
page 85
12. It is likely that the PivotTable assumed you
wanted to SUM the Store ID, which provides nonsense
data. We need to change that aggregate to a COUNT
instead.
Click the drop-down next to Sum of STORE in the
VALUES section of the PivotTable Fields window and
select Value Field Settings.
13. Select Count to change the way the data for number of
stores per state are summarized, and then click OK.
14. You have now created a summarized view of the data
that shows the number of stores in each state. To make
this even easier to interpret, you can sort the data so
that you see the states that have the highest number of
stores first. To do so, have your active cell anywhere in
the Count of Store column, right-click the cell, select
Sort, then select Sort Largest to Smallest.
page 86

15. Take a screenshot (label it 2-8A) of the PivotTable.
Q3. Reference your PivotTable and find which state
has the highest number of Dillard’s stores.
Which states have the fewest? How many
stores are there across the country?
Q4. Counting the number of stores per state is one
example of how the data that have been loaded
from SQL Server into Excel can become useful
information through a PivotTable. What are
other ways that you could organize the STORE
data in a PivotTable to come up with meaningful
information?
Writing Queries Directly in Excel

While executing queries in SQL Server is a great method
for viewing data, if you want to eventually load data from a
query result into Excel for analysis or visualization, it is
easiest to write the queries directly in Excel. This puts the
entire ETL process in one tool, Excel.
1. Open a new Excel workbook.
2. From the Data tab, click New Query > From Database
> From SQL Server Database.

3. In the Microsoft SQL database pop-up window, input
the server information that you received when
accessing the WCOB_DILLARDS data. The Database
name is WCOB_DILLARDS.
page 87
Important Note: If you just worked through the first part

of this lab (connecting to data), this step is where the
process begins to be different. Instead of clicking OK, you
will click SQL statement (optional).
4. For this query, we will pull in enough data to answer a

variety of questions about transaction line items in each
state. We’ll select all of the columns from the
TRANSACT table and the STATE column from the
STORE table. In order to do that, we’ll join the two
tables together in our query.
Q5. Joins are made based on their primary
key/foreign key relationship. Looking at the
ERD or the dataset, which two columns form
the relationship between the TRANSACT and
STORE tables?
5. Type this query into the SQL statement box:
SELECT TRANSACT.*, STATE
FROM TRANSACT
INNER JOIN STORE
ON TRANSACT.STORE = STORE.STORE
WHERE TRAN_DATE BETWEEN '20160901' AND
'20160905'
page 88
6. Click OK to continue.
7. If presented with a box questing your credentials, click
Connect using your current credentials in the next
window.
8. If presented with a window about encryption support,

click OK on the Encryption Support window.
9. Excel will provide you a preview of your data before
loading it. If the query loads successfully (i.e., if you see
the preview, instead of an error), click Load to load the
data into an Excel table.
10. It may take a few minutes to load. Even though the
query we ran was only for 15 days of transactions,
there are still more than 1 million transactions (or rows)
to return.
Part 4: Calculate Summary Statistics

Calculating summary statistics such as mean, median,
and mode for quantitative data can be helpful to get a
quick feeling for the components of a large dataset.
11. While you can calculate these statistics by hand, you
can also have Excel calculate them automatically
through the Data Analysis ToolPak. If you haven’t
added this component into Excel yet, follow this menu
path: File > Options > Add-ins. From this window,
select the Go. . . button, and then place a check mark
in the box next to Analysis ToolPak. Once you click
OK, you will be able to access the ToolPak from the
Data tab on the Excel ribbon.
12. We will calculate descriptive statistics for the attributes
ORIG_PRICE, SALE_PRICE, TRAN_AMT.
Q6. Looking at the first several rows of data,
compare the amounts in ORIG_PRICE,
SALE_PRICE, TRAN_AMT. What do you think
TRANS_AMT represents?
13. Click the Data Analysis button from the Data tab on
the Excel ribbon and select Descriptive Statistics.
page 89
14. For the Input Range, select the three columns

associated with the three attributes that we are
measuring. Leave the default to columns, and place a
check-mark in Labels in First Row.
15. Place a check mark next to Summary Statistics, then
press OK.
It may take awhile for the statistics to run because you’re

working with so many rows.
Q7. What are the means for each of the attributes
(ORIG_PRICE, SALE_PRICE, TRAN_AMT)?
Q8. The mean from TRAN_AMT is lower than the
means for both ORIG_PRICE and
SALE_PRICE, why do you think that is? (Hint:
It is not an error).

Q9. How does doing a query within Excel allow
quicker and more efficient access and
analysis of the data?
page 90
Q10. Is 5 days of data sufficient to
capture the statistical relationship among and
between different variables? What will Excel
do if you have more than 1 million rows?
Note: There are statistical programs such as SAS and
SPSS that allow for transformation and statistical analysis
of bigger datasets. You can also store your data in the
Data Model in Excel and create a PivotTable from the
data model. This will allow you to analyze massive
datasets in Excel, as long as your PivotTable doesn’t
exceed the row limit.
End of Lab

Dillard’s Store Data: Joining Tables
Company summary
Data
Technique
This lab is most easily performed if Labs 2-6 and 2-7
have already been completed.
Software needed

Learn how to do a table join, joining two tables all on
your own and running an analysis of the data.

1. Consult the entity-relationship diagram to view the
variables available in the TRANSACT table and the
CUSTOMER table. Consult Lab 2-6 for instructions to
access the ERD.
Q1. If we wanted to join the TRANSACT and the
CUSTOMER tables, what fields (or variables)
would we use to join them?

Arkansas.
3. Open Microsoft SQL Server Management Studio to
access the WCOB_Dillards data.
window that you were provided through the
Walton.uark.edu/enterprise website.
page 91
5. Leave the default for authentication to Windows

Authentication, and click Connect.
6. Select New Query from the menu at the top of the SQL
Server application.
7. If the drop-down indicating which database you are

intending to query doesn’t say “WCOB_DILLARDS”
(e.g., it frequently defaults to “Master”), select the drop-
down window and scroll down to WCOB_DILLARDS,
then click Enter. You could also type
WCOB_DILLARDS instead of waiting to scroll to it.
8. Given the description in the text and in Labs 2-6 and 2-

7, you have the tools you need to join two tables,
TRANSACT and CUSTOMER and run a query on
customer state that will count the number of sales made
to customers by state. Run the query for the entire
dataset; do not filter based on a limited set of days.
9. This query may take a few minutes to run. Once the
results have returned, you can check your results by
looking at how many transactions at Dillard’s have
originated Customers from Arkansas (AR): 3953688.
Q2. How many different states are listed?
Q3. Why are there so many more states listed than
50?
Q4. What do you assume the blank values
represent? If you were to analyze these data to
learn more about the number of customers from
different places have shopped at Dillard’s,
what would you do with these data: group them,
leave them out, leave them alone? Why?
End of Lab
1B. Resnick, “Researchers Just Released Profile Data on 70,000 OkCupid Users without
Permission,” 2016, http://www.vox.com/2016/5/12/11666116/70000-okcupid-users-data-
release (accessed October 31, 2016).
2J. P. Isson and J. S. Harriott, Win with Advanced Business Analytics: Creating Business
Value from Your Data (Hoboken, NJ: Wiley, 2013).
3G. C. Simsion and G. C. Witt, Data Modeling Essentials (Amsterdam: Morgan Kaufmann,
2005).
4T. Singleton, “What Every IT Auditor Should Know about Data Analytics,” n.d., from
http://www.isaca.org/Journal/archives/2013/Volume-6/Pages/What-Every-IT-Auditor-
Should-Know-About-Data-Analytics.aspx#2.
5For a description of the audit data standards, please see this website:
https://www.aicpa.org/interestareas/frc/assuranceadvisoryservices/pages/assuranceandad
visory.aspx.
page 92
Chapter 3
Performing the Test Plan and
Analyzing the Results

Data Analytics involves various models and techniques used to understand
the environment, make comparisons, and predict the future. In this chapter,
we evaluate several different approaches and models and identify when to
use them and how to interpret the results. We also provide specific
accounting-related examples of when each of these specific data
approaches and models is appropriate to address our particular question.
A Look Back
Chapter 2 provided a description of how data are prepared and scrubbed
to be ready to use to answer business questions. We explained how to
extract, transform, and load data and then how to validate and normalize
the data. In addition, we explained how data standards are used to
facilitate the exchange of data between both senders and receivers.
A Look Ahead
Chapter 4 will demonstrate various techniques that can be used to
effectively communicate the results of your analyses. Additionally, we
discuss how to refine your results and translate your findings into useful
information for decision makers.
page 93
jaruek/123RF
Liang Zhao Zhang, a San Francisco–based janitor, made more than $275,000 in 2015. The
average janitor in the area earns just $26,180 a year. Zhang, a Bay Area Rapid Transit
(BART) janitor, has a base pay of $57,945 and $162,050 in overtime pay. With benefits, the
total was $276,121. While some call his compensation “outrageous and irresponsible,” Zhang
signed up for every available overtime slot that became available. To be sure, Zhang worked
more than 4,000 hours last year and received overtime pay. Can BART predict who might
take advantage of overtime pay? Should it set a policy restricting overtime pay? Would it be
better for BART to hire more regular, full-time employees instead of offering so much
overtime? Can Data Analytics help with these questions?
Using a profiling data analytics approach detailed in this chapter, BART could generate
summary statistics of its workers and their overtime pay to see the extent that overtime is
required and taken advantage of.
Using regression and classification approaches to Data Analytics would help to classify
which employees are most likely to exceed normal bounds and why. BART, for example, has
a policy of offering overtime by seniority. So do the most senior employees sign up first and
leave little overtime to others? Will a senior employee get paid more for overtime than more
junior-level employees? If so, is that the best policy for the company and its employees?
Source: http://www.cnbc.com/2016/11/04/how-one-bay-area-janitor-made-276000-last-
year.html.
OBJECTIVES
LO 3-1 Understand four categories of Data Analytics.

LO 3-2 Describe some descriptive analytics approaches, including
summary statistics and data reduction.
LO 3-3 Explain the diagnostic approach to Data Analytics,
including profiling and clustering.
LO 3-4 Understand predictive analytics, including regression and
classification.
LO 3-5 Describe the use of prescriptive analytics, including
machine learning and artificial intelligence.
page 94
PERFORMING THE TEST PLAN

LO 3-
1
Understa
nd four
categorie
s of data
analytics.
The third step of the IMPACT cycle model, or the “P,” is “performing test
plan.” In this step, different Data Analytics approaches help us understand
what happened, why it happened, what we can expect to happen in the
future, and what we should do. These Data Analytics approaches or
techniques help to address our business questions and provide information
to support accounting and management decisions.
Data Analytics approaches rely on a series of tasks and models that are
used to understand data and gain insight into the underlying cause and
effect of business decisions. Many accounting courses introduce students
to basic models that describe the results of periodic transactions (e.g.
ratios, trends, and variance analysis). These simple calculations help
accountants fulfill their traditional role as historians summarizing the results
from the past to inform stakeholders of the status of the business.
While these simple techniques provide important information, their
value is limited to providing information in hindsight. The contributing value
of Data Analytics increases as the focus shifts from hindsight to foresight
and from summarizing information to optimizing business outcomes as we
go from descriptive analytics to prescriptive analytics (as illustrated in
Exhibit 3-1). For example, lean accounting relies more heavily on data
analysis to accurately predict changes in budgets and forecasts to
minimize disruption to the business. These models that more accurately
predict the future and prescribe a course of action come at a cost of
increasing complexity in terms of manipulating and calculating appropriate
data, and the implications of the results.
EXHIBIT 3-1
Four Main Categories of Data Analytics
There are four main categories of data analytics, shown in Exhibit 3-1:
Descriptive analytics are procedures that summarize existing data to
determine what has happened in the past. Some examples of descriptive
analytics include summary statistics (e.g. Count, Min, Max, Average,
Median), distributions, and proportions.
Diagnostic analytics are procedures that explore the current data to
determine why something has happened the way it has, typically
comparing the data to a benchmark. As an example, diagnostic analytics
allow users to drill-down in the data and see how it compares to a budget,
a competitor, or trend.
Predictive analytics are procedures used to generate a model that can
be used to determine what is likely to happen in the future. Examples of
predictive analytics include regression analysis, forecasting, classification,
and other predictive modeling.
page 95
Prescriptive analytics are procedures that model data to
enable recommendations for what should be done in the future. These
typically include developing more advanced machine learning and artificial
intelligence models to recommend a course of action based on a current
problem.
The choice of Data Analytics model depends largely on the type of
question that you’re trying to answer and your access to the data needed
to answer the question. Descriptive and diagnostic analytics are typically
paired when you would want to describe the past data and then compare it
to a benchmark to determine why the results are the way they are, similar
to the accounting concepts of planning and controlling. Likewise, predictive
and prescriptive analytics make good partners when you would want to
predict an outcome and then make a recommendation on how to follow up,
similar to an auditor flagging a transaction as high risk and then following a
decision flowchart to determine whether to request additional evidence or
include it in audit findings.
As you move from one Data Analytics approach to the next, you trade
hindsight and information, which are traditionally accounting domain areas
for foresight and optimization.
Ultimately, the model you use comes down to the questions you are
trying to answer. We highlighted the Data Analytics approaches in Chapter
1. Here we categorize them into the four main analytics categories,
summarized in Table 3-1:
TABLE 3-1
Summary of Data Analytics Approaches
Type of
Example in accounting
analytic
Descriptive Understand what happened.

analytics
Summary Calculate the average and medium income, age range,

statistics and highest and lowest purchases of customers during the
4th quarter.
Data Filter data to only include transactions within the current

reduction or reporting period.
filtering
Diagnostic Understand why it happened.

analytics
Profiling Identify outlier transactions to determine exposure and risk.

Compare individual segments to a benchmark.
Clustering Identify groups of store locations that are outperforming the

rest.
Similarity Understand the underlying behavior of high-performing

matching divisions.
Co- Identify related party and intracompany transactions based

occurrence on the individuals involved in a the transaction.
grouping
Predictive Estimate a future value or category.

analytics
Regression Calculate the fixed and variable costs in a mixed cost

equation or determine the number of days a shipment is
likely to take.
Type of
Example in accounting
analytic
Classification Determine whether or not certain demongraphics, such as

age, zip code, income level or gender are likely to engage
in fraudulent transactions.
Link Predict participation of individuals based on their

prediction underlying common attributes, such as an incentive.
Prescriptive Make recommendations for a course of action.

analytics
Decision Tax software takes input from a preparer and recommends

support whether to take a standard deduction or an itemized
systems deduction.
Artificial Audit analysis monitors changing transactions and

intelligence suggests follow up when new abnormal patterns appear.
1. Descriptive analytics:
Summary statistics describe a set of data in terms of their location
(mean, median), range (standard deviation, minimum, maximum), shape
(quartile), and size (count).
Data reduction or filtering is used to reduce the amount of observations
to focus on relevant items (i.e., highest cost, highest risk, largest impact,
etc.). It does this by taking a large set of data (perhaps the population)
and reducing it to a smaller set that has the vast majority of the critical
information of the larger set. For example, auditing may use data
reduction to narrow transactions based on relevance or size. While
auditing has employed various random and stratified sampling over the
years, Data Analytics suggests new ways to highlight which transactions
do not need the same level of vetting as other transactions.
page 96
2. Diagnostic analytics:
Profiling identifies the “typical” behavior of an individual, group, or
population by compiling summary statistics about the data (including
mean, standard deviations, etc.) and comparing individuals to the
population. By understanding the typical behavior, we’ll be able to identify
abnormal behavior more easily. Profiling might be used in accounting to
identify transactions that might warrant some additional investigation
(e.g., outlier travel expenses or potential fraud).
Clustering helps identify groups (or clusters) of individuals (such as
customers) that share common underlying characteristics—in other
words, identifying groups of similar data elements and the underlying
drivers of those groups. For example, clustering might be used to
segment a customer into a small number of groups for additional analysis
and risk assessment. Likewise, transactions might also be put into
clusters to understand underlying relationships.
Similarity matching is a grouping technique used to identify similar
individuals based on data known about them. The opening vignette in
Chapter 1 mentioned Alibaba and its attempt to identify seller and
customer fraud based on various characteristics known about them to
see if they were similar to known fraud cases.
Co-occurrence grouping discovers associations between individuals
based on common events, such as transactions they are involved in.
Amazon might use this to sell another item to you by knowing what items
are “frequently bought together” or “Customers who bought this item also
bought . . .” as shown in Chapter 1.
3. Predictive analytics:
Regression estimates or predicts the numerical value of a dependent
variable based on the slope and intersect of a line and the value of an
independent variable. An R2 value indicates how closely the line fits to
the data used to calculate the regression. An example of regression
analysis might be, given a balance of total accounts receivable held by a
firm, what is the appropriate level of allowance for doubtful accounts for
bad debts?
Classification predicts a class or category for a new observation based
on the manual identification of classes from previous observations.
Membership of a class may be binary in the case of decision trees or
indicate the distance from a decision boundary. Some examples of
classification include predicting which loans are likely to default, credit
applications that are expected to be approved, the classification of an
operating or financing lease, or identification of suspicious transactions.
In each of these cases, prior data must be manually identified as
belonging to each class to build the predictive model.
Link prediction predicts a relationship between two data items, such as
members of a social media platform. For example, if two individuals have
mutual friends on social media and both attended the same university, it
is likely that they know each other and the site may make a
recommendation for them to connect. Chapter 1 provides an example of
this used in Facebook. Link prediction in an accounting setting might
work to use social media to look for relationships between related parties
that are not otherwise disclosed to identify related party transactions.
4. Prescriptive analytics:
Decision support systems are rule-based systems that gather data and
recommend actions based on the input. Tax preparation software,
investment advice tools, and auditing tools recommend courses of
actions based on data that are input as part of an interview or
interrogation process.
Machine learning and artificial intelligence are learning models or
intelligent agents that adapt to new external data to recommend a course
of action. For example, an artificial intelligence model may observe
opinions given by an audit partner and adjust the model to reflect
changing levels of risk appetite and regulation.
While these are all important and applicable data approaches, in the
rest of the chapter we limit our discussion to the more common models,
including summary statistics, data reduction, profiling, clustering,
regression, classification, and artificial intelligence. You’ll find that these
data approaches are not mutually exclusive and that actual analysis may
involve parts of several approaches to arrive at the intended test of the
data and result.
page 97
PROGRESS CHECK
1. Using Table 3-1, identify the appropriate approach for the
following questions:
a. Will a customer purchase item X if given incentive A?
b. Should we offer a customer a line of credit?

c. How many items will the customer purchase?
2. What is the main difference between descriptive and diagnostic
methods?
DESCRIPTIVE ANALYTICS
LO 3-
2
Describe
some
descriptiv
e
analytics
approach
es,
including
summary
statistics
and data
reduction.
Descriptive analytics help summarize what has happened in the past. For
example, a financial accountant would sum all of the sales transactions
within a period to calculate the value for Sales Revenue that appears on
the income statement. An analyst would count the number of records in a
data extract to ensure the data are complete before running a more
complexanalysis. An auditor would filter data to limit the scope to
transactions that represent the highest risk. In all these cases, basic
analysis provides an understanding of what has happened in the past to
help decision makers achieve good results and correct poor results.
Here we look at two main approaches that are used by accountants
today: summary statistics and data reduction.
Summary Statistics
Summary statistics describe the location, spread, shape, and
dependence of a set of observations. These commonly include the count,
sum, minimum, maximum, mean or average, standard deviation, median,
quartiles, correlation covariance, and frequency that describe a specific
measurable value, shown in Table 3-2.
TABLE 3-2
Description of Summary Statistics
Statistic Excel formula Description
Sum SUM() The total value of all numerical values
Mean =AVERAGE() The center value; sum of all observations

divided by the number of observations
Median =MEDIAN() The middle value that divides the top half
of the data from the bottom half
Minimum =MIN() The smallest value
Maximum =MAX() The largest value
Count =COUNT() The number of observations
Frequency =FREQUENCY() The number of observations in each of a

series of numerical or categorical buckets
Standard =STDEV() The variability or spread of the data from

deviation the mean; a larger standard deviation
means a wider spread away from the
mean
Quartile =QUARTILE() The value that divides a quarter of the data

from the rest; indicates skewness of the
data
Correlation =CORREL() How closely two datasets are correlated or

coefficient predictive of one another
page 98
The use of summary statistics helps the user understand what the data
look like. For example, the sum function can be used to determine account
balances. The mean and median can be used to aggregate transactions by
employee, location, or division. The standard deviation and frequency help
to identify normal behavior and trends in the data.
Data Reduction
As you recall, the data reduction approach attempts to reduce the amount
of detailed information considered to focus on the most critical, interesting,
or abnormal items (i.e., highest cost, highest risk, largest impact, etc.). It
does this by filtering through a large set of data (perhaps the total
population) and reducing it to a smaller set that has the vast majority of the
critical information of the larger set. The data reduction approach is done
primarily using structured data—that is, data that are stored in a database
or spreadsheet and are readily searchable.
Data reduction involves the following steps (using an example of an
employee creating a fictitious vendor and submitting fake invoices):
1. Identify the attribute you would like to reduce or focus on. For example, an
employee may commit fraud by creating a fictitious vendor and submitting
fake invoices. Rather than evaluate every employee, an auditor may be
interested only in employee records that have addresses that match
vendor addresses.
2. Filter the results. This could be as simple as using filters in Excel, or using
the WHERE phrase in a SQL query. It may also involve a more
complicated calculation. For example, employees who create fictitious
vendors will often use addresses that are similar, but not exactly the
same, as their own address to foil basic SQL queries. Here the auditor
should use a tool that allows fuzzy matching, which uses probability to
identify likely similar addresses.
3. Interpret the results. Once you have eliminated irrelevant data, take a
moment to see if the results make sense. Calculate the summary
statistics. Have you eliminated any obvious entries? Looking at the list of
matching employees, the auditor might tweak the probability in the fuzzy
match to be more or less precise to narrow or broaden the number of
employees who appear.
4. Follow up on results. At this point, you will continue to build a model or
use the results as a targeted sample for follow-up. The auditor should
review company policy and follow up with each employee who appears in
the reduced list as it represents risk.
Example of Data Reduction in Internal and

External Auditing
While auditing has employed various random and stratified sampling over
the years, Data Analytics suggests new ways to highlight transactions that
do not need the same level of vetting or further analysis as other
transactions. One example might be to filter the travel and entertainment
(T&E) transactions to find specific values, including whole-dollar amounts
of T&E expenses. Whole-dollar amounts have a greater likelihood of being
made up or fraudulent (as illustrated in Exhibit 3-2).
EXHIBIT 3-2
Use filters to reduce data
Auditors may filter data to consider only those transactions being paid
to specific vendors, such as mobile payment processors. Because anyone
can create a payment account using processors such as Square
Payments, there is a higher potential for the existence of a fictitious or
employee-created vendor. The data reduction approach allows us to focus
more time and effort on those vendors and transactions that might require
additional analysis to make sure they are legitimate.
page 99
Another example of the data reduction approach is gap detection,

where we look for missing numbers in a sequence, such as payments
made by check. Finding out why certain check numbers were skipped and
not recorded requires additional analysis and consideration.
Data reduction may also be used to filter all the transactions between
known related party transactions. Focusing specifically on related party
transactions allows the auditor to focus on those transactions that might
potentially be sensitive and/or risky.
Finally, data reduction might be used to compare the addresses of
vendors and employees to ensure that employees are not siphoning funds
to themselves. Use of fuzzy match looks for correspondences between
portions, or segments, of the text of each potential match, shown in Exhibit
3-3. Once potential matches between vendors and employees are found,
additional analysis must be conducted to figure out if funds have been, or
potentially could be, siphoned.
EXHIBIT 3-3
A fuzzy matching shows a likely match of an employees and vendor
Examples of Data Reduction in Other Accounting

Areas
Data reduction approaches are also used in operational audit settings. For
example, filtering the data to find cases where there are duplicate invoice
payments might be an efficient way to find errors or fraud. Once duplicate
invoice payments are found, additional work can be done to identify the
reasons this has occurred. It may also be a way to reduce costs when
duplicate payments are found and procedures are set in place to mitigate
duplicate payments from occurring in the future.
Data reduction approaches may also be useful in a financial statement
analysis setting, perhaps performed by financial analysts, pension funds, or
individual investors. Among other uses, XBRL (eXtensible Business
Reporting Language) is used to facilitate the exchange of financial
reporting information between the company and the Securities and
Exchange Commission (SEC). The SEC then makes it available to all
interested parties, including suppliers, competitors, investors, and financial
analysts. XBRL requires that the data be tagged according to the XBRL
taxonomy. Using these tagged data in common size page 100
financial statements, financial analysts, loan officers,
auditors, accountants, information system specialists, and investors
develop models to access all the relevant financial or nonfinancial data to
help interpret the financial data to predict future earnings, forecast solvency
or liquidity, and analyze profitability. The use of XBRL and the modeling by
financial data takes all the details of the financial statements, footnotes,
and other financial and nonfinancial data and summarizes them in models
of future earnings, solvency, liquidity, and profitability. We’ll explore XBRL
further in Chapter 8.
PROGRESS CHECK
3. Describe how the data reduction approach could be used to
evaluate employee travel and entertainment expenses.
4. Explain how XBRL might be used by lenders to focus on specific
areas of interest.
DIAGNOSTIC ANALYTICS
LO 3-
3
Explain
the
diagnosti
c
approach
to Data
Analytics,
including
profiling
and
clustering
Diagnostic analytics provide insight into why things happened or how

individual data values relate to the general population. Once you
summarize data using descriptive techniques, you can drill-down and
discover the numbers that are driving an outcome. Benchmarks give
context to the data by giving analysts a reference point (or line) to compare
the data to. For example, the arithmetic mean of a data set gives you
context for a specific value. These benchmarks may be based on past
activity, a comparison with a major competitor or an entire industry.
Two primary methods of diagnostic analytics include profiling and
cluster analysis. In both of these cases the analysis provides insight into
where a specific value lies relative to the rest of the sample or population.
The farther the distance from the rest of the observations, the more
interesting the individual value becomes. These outliers could represent
risk or opportunities to learn more about the business process or
partnerships driving the behavior.
Profiling
As you recall, profiling involves gaining an understanding of a typical
behavior of an individual, group, or population (or sample). Profiling is done
primarily using structured data—data that are stored in a database or
spreadsheet and are readily searchable. Using these data, analysts can
use common summary statistics to describe the individual, group, or
population, including knowing its mean, standard deviation, sum, etc.
Profiling is generally performed on data that are readily available, so the
data have already been gathered and are ready for further analysis.
Profiling is used to discover patterns of behavior. In Exhibit 3-4, for
example, the higher the Z-score (farther away from the mean), the more
likely a customer will have a delayed shipment (blue circle). As shown in
the Exhibit, a Z-score of three represents three standard deviations away
from the mean. We use profiling to explore the attributes of that vendor that
we may want to avoid in the future.
EXHIBIT 3-4
Z-Scores Provide an Example of Profiling That Helps Identify Outliers (in this case,
customers with unusually high average days to ship)
Data profiling can be as simple as calculating summary statistics on

transactional data, such as the average number of days to ship a product,
the typical amount we pay for a product, or the number of hours an
employee is expected to work. On the other hand, profiling can be used to
develop complex models to predict potential fraud. For example, you might
create a profile for each employee in a company that may include a
combination of salary, hours worked, and travel and entertainment
purchasing behavior. Sudden deviations from an employee’s past behavior
may represent risk and warrant follow-up by the internal auditors.
Similar to evaluating behavior, data profiling is typically used to assess
data quality and internal controls. For example, data profiling may identify
customers with incomplete or erroneous master data or mistyped
transactions.
page 101
Data profiling typically involves the following steps:

1. Identify the objects or activity you want to profile. What data do you want
to evaluate? Sales transactions? Customer data? Credit limits? Imagine a
manager wants to track sales volume for each store in a retail chain. She
might evaluate total sales dollars, asset turnover, use of promotions and
discounts, and/or employee incentives.
2. Determine the types of profiling you want to perform. What is your goal?
Do you want to set a benchmark for minimum activity, such as monthly
sales? Have you set a budget that you wish to follow? Are you trying to
reduce fraud risk? In the retail store scenario, the manager would likely
want to compare each store to the others to identify which ones are
underperforming or overperforming.
3. Set boundaries or thresholds for the activity. This is a benchmark that may
be manually set, such as a budgeted value, or automatically set, such as
a statistical mean, quartile, or percentile. The retail chain manager may
define underperforming stores as those whose sales activity falls below
the 20th percentile of the group and overperforming stores as those
whose sales activity is above the 80th percentile. These thresholds are
automatically calculated based on the total activity of the stores, so the
benchmark is dynamic.
4. Interpret the results and monitor the activity and/or generate a list of
exceptions. Here is where dashboards come into play. Management can
use dashboards to quickly see multiple sets of profiled data and make
decisions that would affect behavior. As you evaluate the results, try to
understand what a deviation from the defined boundary represents. Is it a
risk? Is it fraud? Is it just something to keep an eye on? To evaluate her
stores, the retail chain manager may review a summary of the sales
indicators and quickly identify under- and overperforming stores. She is
likely to be more concerned with underperforming stores because they
represent major challenges for the chain. Overperforming stores may
provide insight into marketing efforts or customer base.
5. Follow up on exceptions. Once a deviation has been identified,
management should have a plan to take a course of action to validate,
correct, or identify the causes of the abnormalbehavior. When the retail
chain manager notices a store that is underperforming compared to its
peers, she may follow up with the individual store manager to understand
his concerns or offer a local promotion to stimulate sales.
page 102
As with most analyses, data profiles should be updated periodically to

reflect changes in firm activity and identify activity that may be more
relevant to decision making.
Example of Profiling in Management Accounting
Advanced Environmental Recycling Technologies (ticker symbol
AERT) makes wood-plastic composite for decking that doesn’t rot and
keeps its form, color, and shape indefinitely. It has developed a recipe and
knows the standards of how much wood, plastic, and coloring goes into
each foot of decking. AERT has developed standard costs and constantly
calculates the means and standard deviations of the use of wood, plastic,
coloring, and labor for each foot of decking. As the company profiles each
production batch, it knows that when significant variances from the
standard cost occur, those variances need to be investigated further.
Management accounting relies heavily on diagnostic analytics in the
planning and controlling process. By comparing the actual results of activity
to the budgeted expectation, management determines the processes and
procedures that resulted in favorable and unfavorable activity. For example,
in a manufacturing company like AERT, variance analysis compares the
actual cost, price, and volume of various activities with standard
equivalents, shown in Exhibit 3-5. The unfavorable variances appear in
orange as the actual cost exceeds the budgeted cost or are to the left of
the budget reference line. Favorable variances appear to the right of the
budget reference line in blue. Sales exceed the budgeted sales. As sales
volume increases, the costs (negative values) also increase, leading to an
unfavorable variance in orange.
EXHIBIT 3-5
Variance Analysis Is an Example of Data Profiling
page 103
Example of Profiling in an Internal Audit

Profiling might also be used by internal auditors to evaluate travel and
entertainment (T&E) expenses. In some organizations, total annual T&E
expenses are second only to payroll and so represent a major expense for
the organization. By profiling the T&E expenses, we can understand the
average amount and range of expenditures and then compare and contrast
with prior period’s mean and range to help identify changing trends and
potential risk areas for audit and potentially for tax purposes. This will help
indicate areas where there is lack of controls, changes in procedures, or
individuals more willing to spend excessively in potential types of T&E
expenses, etc., which might be associated with higher risk.
The use of profiling in internal audits might unearth when employees
misuse company funds, like in the case of Tom Coughlin, an executive at
Walmart, who misused “company funds to pay for CDs, beer, an all-terrain
vehicle, a customized dog kennel, even a computer as his son’s graduation
gift—all the while describing the purchases as routine business
expenses.”1
Example of Profiling in Auditing

Profiling is also useful in continuous auditing. If we consider the dollar
amount of each transaction, we can develop a Z-score by knowing the
mean and standard deviation. Using our statistics knowledge and
assuming a normal distribution, any transaction that has a Z-score of 3 or
above would represent abnormal transactions that might be associated
with higher risk. We can investigate further seeing if those transactions had
appropriate approvals and authorization.
An analysis of Benford’s law could also be used to assess a set of
transactions. Benford’s law is an observation about the frequency of
leading digits in many real-life sets of numerical data. The law states that in
many naturally occurring collections of numbers, the significant leading
digit is likely to be small. If the distribution of transactions for an account
like “sales revenue” is substantially different than Benford’s law would
predict, then we would investigate the sales revenue account further and
see if we can explain why there are differences from Benford’s law. Exhibit
3-6 shows an illustration of Benford’s law using the first digit of employee
transactions. An abnormal frequency of transaction beginning with the
number four may indicate that employees are attempting to circumvent
internal controls, such as an approval limit. While the number one also
exceeds the expected value, we would expect a larger volume of smaller
numbers. We will discuss additional applications of Benford’s law in
Chapter 6.
EXHIBIT 3-6
Benford’s Law Applied to Large Numerical Data Sets (including employee transactions)
page 104
Cluster Analysis
The clustering data approach works to identify groups of similar data
elements and the underlying relationships of those groups. More
specifically, clustering techniques are used to group data/observations into
a specific number of clusters or groups so that all the data within any
cluster are similar, while data across clusters are different. Cluster analysis
works by calculating the minimum distance between each observation and
the center of each cluster, shown in Exhibit 3-7.
EXHIBIT 3-7
Clustering Is Used to Find Three Natural Groupings of Vendors Based on Purchase Activity
When you are exploring the data for these patterns and don’t have a
specific question, you would use an unsupervised approach. For
example, consider the question: “Do our vendors form natural groups
based on similar attributes?” In this case, there isn’t a specific target
because you don’t yet know what similarities our vendors have. You may
use clustering to evaluate the vendor attributes and see which ones are
closely related. You could also use co-occurrence grouping to match
vendors by geographic region; data reduction to simplify vendors into
obvious categories, such as wholesale or retail or based on overall volume
of orders; or profiling to evaluate vendors with similar on-time delivery
behavior, shown in Exhibit 3-4. In any of these cases, the data drive the
analysis, and you evaluate the output to see if it matches our intuition.
These exploratory exercises may help to define better questions, but are
generally less useful for making decisions.
As an example, Walmart may want to understand the types of
customers who shop at its stores. Because Walmart has good reason to
believe there are different market segments of people, it may consider
changing the design of the store or the types of products to accommodate
the different types of customers, emphasizing the ones that are most
profitable to Walmart. To learn about the different types of customers,
managers may ask whether customers agree with the following statements
using a scale of 1–7 (on a Likert scale):
Enjoy: I enjoy shopping.
Budget: I try to avoid shopping because it is bad for the budget.
Eating: I like to combine my shopping with eating out.
Coupons: I use coupons when I shop.
Quality: I care more about the quality of the products than I do about the
price.
Apathy: I don’t care about shopping.
Comparing: You can save a lot of money by comparing prices between
various stores.
Additionally, they would ask about numerical customer behavior:
Income: The household income of the respondent (in dollars).
Shopping at Walmart: How many times a month do you visit Walmart?
Accountants may analyze the data and plot the responses to see if
there are correlations within the data on a scatterplot. The visual plot of the
relationship between responses to the various questions page 105
may help cluster the various customers into different
clusters and help Walmart cater to specific customer clusters better
through superior insights.
Example of the Clustering Approach in Auditing

The clustering data approach may also be used in an auditing setting.
Imagine a group insurance setting where fraudulent claims associated with
payment were previously found by internal auditors through happenstance
and/or through hotline tips. Based on current internal audit tests, payments
are the major concern of the business unit. Specifically, the types of related
risks identified are duplicate payments, fictitious names, improper/incorrect
information entered into the systems, and suspicious payment amounts.
Clustering is useful for anomaly detection in payments to insurance
beneficiaries, suppliers, etc. By identifying transactions with similar
characteristics, transactions are grouped together into clusters. Those
clusters that consist of few transactions or small populations are then
flagged for investigation by the auditors as they represent groups of
outliers. Examples of these flagged clusters include transactions with large
payment amounts and/or a long delay in processing the payment.
The dimensions used in clustering may be simple correlations between
variables, such as payment amount and time to pay, or more complex
combinations of variables, such as ratios or weighted equations. As they
explore the data, auditors develop attributes that they think will be relevant
through intuition or data exploration. Exhibit 3-8 illustrates clustering of
insurance payments based on the following attributes:
EXHIBIT 3-8
Cluster Analysis of Insurance Payments
1. Payment amount: The value of the transaction payment.
2. Days to Pay: The number of days from the original recorded transaction to
the payment date.
page 106
The data are normalized to reduce the distortion of the data and other
outliers are removed. They are then plotted with the number of days to pay
on the x axis and the payment amount on the y-axis. Of the eight clusters
identified, three clusters highlight potential anomalies that may require
further investigation as part of an internal or external audit.
Cluster 6 payments (purple) have a long duration between the processing
to payment dates.
Cluster 7 payments (pink) have high payment amounts.
Cluster 8 payments (brown) have high payment amounts and a long
duration between the processing date and the payment date.
With this insight auditors may assess the risk associated with these
payments and understand transaction behavior relative to acceptable
behavior defined in internal controls.
PROGRESS CHECK
5. Profiling is also used in law enforcement, such as offender or
criminal profiling. Offender profiling is a tool used by law
enforcement to identify likely suspects, analyze data patterns to
help predict future offenses by criminals, and identify potential
victims. Compare and contrast this type of profiling with the
profiling data approach used in accounting (mentioned earlier in
this section).
6. Identify a reason the sales amount of any single product may or
may not follow Benford’s law.
7. Name three clusters of customers who might shop at Walmart.
8. In Exhibit 3-8, Cluster 1 of the group insurance highlighted claims
have a long period from death to payment dates. Why would that
cluster be of interest to internal auditors?
PREDICTIVE ANALYTICS
LO 3-
4
Understa
nd
predictive
analytics,
including
regressio
n and
classificat
ion.
Before we discuss predictive analytics, we need to bring you up to speed
on some data-specific terms:
A target is an expected attribute or value that we want to evaluate. For
example, if we are trying to predict whether a transaction is fraudulent, the
target might be a specific “fraud score.” If we’re trying to predict an interest
rate, the target would be “interest rate.”
A class is a manually assigned category applied to a record based on an
event. For example, if the credit department has rejected a credit line for a
customer, the credit department assigns the class “Rejected” to the
customer’s master record. Likewise, if the internal auditors have confirmed
that fraud has occurred, they would assign the class “fraud” to that
transaction.
On the other hand, we may ask questions with specific outcomes, such
as: “Will a new vendor ship a large order on time?” When you are
performing analysis that uses historical data to predict a future outcome,
you will use a supervised approach. You might use regression to predict
a specific value to answer a question such as, “How many days do we
predict it will take a new vendor to ship an order?” Again, the prediction is
based on the activity we have observed from other vendors, shown in
Exhibit 3-9. We use historical data to create the new model. Using a
classification model, you can predict whether a new vendor belongs to
one class or another based on the behavior of the others, shown in Exhibit
3-10. Causal modeling,similarity matching, and link prediction are
additional supervised approaches where you attempt to identify causation
(which can be expensive), identify a series of characteristics that predict a
model, or attempt to identify other relationships, respectively.
page 107
Regression
Regressions allow the accountant to develop models to predict expected
outcomes. These expected outcomes might be to predict the number of
days to ship products relative to the volume of orders placed by the
customer, shown in Exhibit 3-9.
Regression is a supervised method used to predict specific values. In
this case, the number of days to ship is dependent on the number of items
in the order. Therefore, we can use regression to predict the number of
days it takes Vendor A to ship based on the volume in the order. (Vendor A
is represented by the gold star in Exhibits 3-9 and 3-10).
EXHIBIT 3-9
Regression
EXHIBIT 3-10
Classification
Regression analysis involves the following process:

1. Identify the variables that might predict an outcome. The inputs are called
independent variables, where the output is a dependent variable.
2. Determine the functional form of the relationship. Is it a linear relationship
where each input plots to another? Are you trying to divide the records
into different groups or classes?
3. Identify the parameters of the model. What are the relative weights of
each variable or the thresholds of each branch in a classification?
4. Evaluate the goodness of fit. Calculate the correlation coefficient or R2
value to determine whether the data are close to the line or not. In
general, the better the fit (e.g., R2 > 0.8), the more accurate the prediction
will be.
The following discussion primarily identifies the structure of the model—
that is, the relationship between the dependent variable and the plausible
independent variables—in this way:
Dependent variable = f(Independent variables)
page 108
The dependent variable might be the amount that should be considered

in an allowance for doubtful accounts; the independent variables that might
predict the level needed to reserve it may be current aged loans, loan type,
customer loan history, and collections success. Dependent variables could
be continuous (e.g., salary or loan history) or categorical (e.g., home
ownership or customer type). Continuous variables may provide more
interesting results than a one or two class categorical variable.
Examples of the Regression Approach in

Managerial Accounting
Accounting firms experience a great amount of employee turnover each
year (between 15 and 25 percent each year).2 Understanding and
predicting employee turnover is a particularly important determination for
accounting firms. Each year, they must predict how many new employees
might be needed to accommodate growth, to supply needed areas of
expertise, and to replace employees who have left. Accounting firms might
predict employee turnover by predicting the following regression model in
this way:
Employee turnover = f (Current professional salaries, Health of the

economy [GDP], Salaries offered by other accounting firms or by corporate
accounting, etc.)
Using such a model, accounting firms could then begin to collect the
necessary data to test their model and predict the level of employee
turnover.
Examples of the Regression Approach in

Auditing
One of the key tasks of auditors of a bank is to consider the amount of the
allowance for loan losses or for non-banks to consider the allowance for
doubtful accounts (i.e., those receivables that may never be collected).
These allowances are often subject to manipulation to help manage
earnings.3 The Financial Accounting Standards Board (FASB) recently
issued Accounting Standards Update 2016-13, which requires that banks
provide an estimate of expected credit losses (ECLs) by considering
historical collection rates, current information, and reasonable and
supportable forecasts, including estimates of prepayments.4 Using these
historical and industry data, auditors may work to test a model to establish
a loan loss reserve in this way:
Allowance for loan loses amount = f (Current aged loans, Loan type,
Customer loan history, Collections success)
Other Examples of the Regression in Accounting

For example, in Chapter 1, we worked to understand why LendingClub
rejected certain loan applications. As we considered all of the possible
explanations, we found that there were at least three possible indicators
that a loan might be rejected, including the debt-to-income ratios, length of
employment, and credit (risk) scores, suggesting a model that:
Loan rejection = f (Debt-to-income ratio, Length of employment, Credit

[risk] score)
page 109
Another example of the regression approach might be the approval of

individual credit card transactions. Assume you go on a trip; in the morning
you are in Pittsburgh and by the very next day, you are in Shanghai. Will
your credit card transaction in Shanghai automatically be rejected? Credit
card companies establish models to predict fraud and decide whether to
accept or reject a proposed credit card transaction. A potential model may
be the following:
Transaction approval = f (Location of current transaction, Location of last

transaction, Amount of current transaction, Prior history of travel of credit
card holder, etc.)
Classification
The goal of classification is to predict whether an individual we know very
little about will belong to one class or another. For example, will a customer
have his or her balance written off? The key here is that we are predicting
whether the write-off will occur or not (in other words, there are two
classes: “Write-Off” and “Good”).
Classification is a supervised method that can be used to predict the
class of a new observation. In this case, blue circles represent “on-time”
vendors. Green squares represent “delayed” vendors. The gold star
represents a new vendor with no history.
Classification is a little more involved as we are now dealing with
machine learning and complex probabilistic models. Here are the general
steps:
1. Identify the classes you wish to predict.
2. Manually classify an existing set of records.
3. Select a set of classification models.
4. Divide your data into training and testing sets.
5. Generate your model.
6. Interpret the results and select the “best” model.
Classification Terminology
First, a bit of terminology to prepare us for our discussion.
Training data are existing data that have been manually evaluated and
assigned a class. We know that some customer accounts have been
written off, so those accounts are assigned the class “Write-Off.” We will
train our model to learn what it is that those customers have in common so
we can predict whether a new customer will default or not.
Test data are existing data used to evaluate the model. The
classification algorithm will try to predict the class of the test data and then
compare its prediction to the previously assigned class. This comparison is
used to evaluate the accuracy of the model or the probability that the
model will assign the correct class.
Decision trees are used to divide data into smaller groups, and
decision boundaries mark the split between one class and another.
Exhibit 3-11 provides an illustration of both decision trees and decision
boundaries. Decision trees split the data at each branch into two or more
groups. In this example, the first branch divides the vendor data by
geographic distance and inserts a decision boundary through the middle of
the data. Branches 2 and 3 split each of the two new groups by vendor
volume. Note that the decision boundaries in the graph on the right are
different for each grouping.
EXHIBIT 3-11
Example of Decision Trees and Decision Boundaries
Pruning removes branches from a decision tree to avoid overfitting the

model. In other words, pruning reduces the number of times we split the
groups of data into smaller groups, as shown in Exhibit 3-11. Pre-pruning
occurs during the model generation. The model stops creating new
branches when the information usefulness of an additional branch is low.
Post-pruning evaluates the complete model and discards branches after
the fact. Exhibit 3-12 provides an illustration of how pruning might work in a
decision tree.
page 110
EXHIBIT 3-12
Illustration of Pruning a Decision Tree
Linear classifiers are useful for ranking items rather than simply
predicting class probability. These classifiers are used to identify a decision
boundary. Exhibit 3-13 shows an illustration of linear classifiers segregating
the two classes.
EXHIBIT 3-13
Illustration of Linear Classifiers
A linear discriminant uses an algebraic line to separates the two

classes. In the example noted here, the classification is a function of both
volume and distance:
We don't expect linear classifiers to perfectly segregate classes. For

example, the green square that appears below the line in Exhibit 3-13
would be incorrectly classified as a circle and considered an error.
Support vector machine is a discriminating classifier that is defined by
a separating hyperplane that works first to find the widest margin (or
biggest pipe) and then works to find the middle line. page 111
Exhibits 3-14 and 3-15 provide an illustration of support
vector machines and how they work to find the best decision boundary.
EXHIBIT 3-14
Support Vector Machines With support vector machines, first find the widest margin (biggest
pipe); then find the middle line.
EXHIBIT 3-15
Support Vector Machine Decision Boundaries SVMs have two decision boundaries at the
edges of the pipes.
Evaluating Classifiers
When classifiers wrongly classify an observation, they are penalized. The
larger the penalty (error), the less accurate the model is at predicting a
future value, or classification.
Overfitting
Rarely will datasets be so clean that you have a clear decision boundary.
You should always be wary of classifiers that are too accurate. Exhibit 3-16
provides an illustration of overfitting and underfitting. You want a good
amount of accuracy without being too perfect. Notice how the error rate
declines from 6 to 3 to 0. You want to be able to generalize your results,
and complete accuracy creates a complex model with little predictive value.
EXHIBIT 3-16
Illustration of Underfitting and Overfitting the Data with a Predictive Model
Exhibit 3-17 provides a good illustration of the trade-offs between the

complexity of the model and the accuracy of the classification. While you
may be able to come up with a very complex model with the training data,
chances are it will not improve the accuracy of correctly page 112
classifying the test data. There is, in some sense, a sweet
spot, where the model is most accurate without being so complex to thus
allow classification of both the training as well as the test data.
EXHIBIT 3-17
Illustration of the Trade-Off between the Complexity of the Model and the Accuracy of the
Classification
PROGRESS CHECK
9. If we are trying to predict the extent of employee turnover, do you
believe the health of the economy, as measured using GDP, will
be positively or negatively associated with employee turnover?
10. If we are trying to predict whether a loan will be rejected, would
you expect credit score to be positively or negatively associated
with loan rejection by a bank such as LendingClub?
PRESCRIPTIVE ANALYTICS
LO 3-
5
Describe
the use of
prescripti
ve
analytics,
including
decision
support
systems,
machine
learning
and
artificial
intelligen
ce.
Prescriptive analytics answer the question “What do we do next?” We have

collected the data; analyzed and profiled the data; and in some cases,
developed predictive models to estimate the proper class or target value.
Once those analyses have been performed, the decision process can be
aided by rules-based decision support systems, machine learning models,
or added to an existing artificial intelligence model to improve future
predictions.
These analytics are the most complex and expensive because they rely
on multiple variable and inputs, structured and unstructured data, and in
some cases the ability to understand and interpret natural language
command into data-driven queries.
Decision Support Systems
Decision support systems are information systems that support decision-
making activity within a business by combining data and expertise to solve
problems and perform calculations. They are designed to be interactive
and adapt to the information collected by the user. In the accounting
domain, they are typically built around a series of rules or If . . . then . . .
branching statements that guide the user through the process to the result.
One of the best examples of decision support systems is the calculation
of income tax using off the shelf tax software. Tools like TurboTax guide a
non-technical user through a series of interview questions and have them
enter a numerical income value or answer a yes/no question. The answers
to those questions determine what calculations to include, which schedules
to complete, and what the value of the tax return will be.
Decision support systems can help with application of accounting rules
as well. For example, when a company classifies a lease as a financing or
operating lease, it must consider whether the lease meets a number of
criteria. Using a decision support system, a controller could evaluate a new
lease and answer five questions to determine the proper classification,
shown in Exhibit 3-18.
EXHIBIT 3-18
Lease Classification Flowchart
page 113
Under a previous version of the FAB lease standard, there would have
been bright lines to indicate hard rules to determine the lease (for example,
“The lease term is greater than or equal to 75 percent of the estimated
economic life of the leased asset.”). Decision support systems are easier to
use when you have clear rules. Under the newer standard, more judgment
is needed to reach the most appropriate conclusion for the business. More
on this later.
Auditors use decision support systems as part of their audit procedures.
For example, they indicate a series of parameters such as tolerable and
expected error rates. A tool like IDEA will calculate the appropriate sample
size for evaluating source documents. Once the procedure has been
performed, i.e., source documents are evaluated, the auditor will then input
the number or extent of exceptional items and the decision support system
might classify the audit risk as low, medium, or high for that area.
Machine Learning and Artificial Intelligence

We have discussed some machine learning techniques, including
classification and cluster analysis in the previous sections. What these all
have in common is the use of algorithms and statistical models to generate
a previously unknown model that relies on patterns and inferences. Both
unsupervised exploratory analysis and supervised model generation
provide insight and predictive foresight into the business and decisions
made by the accountants and auditors. They can also model judgment and
decision making to recommend a class or action based on new, unknown
data.
Take lease classification, for instance. With the recent accounting
standard, the language has moved from bright lines (“75 percent of the
useful life”) to judgment (“major part”). While it may be tempting to rely on
the accountant to manually make this decision for each new lease,
machine learning will do it more quickly and more accurately than the
manual classification. A company with a sufficiently large portfolio of
previously classified leases may use those leases as a training set for a
machine learning model. Using the data attributes from these leases (e.g.,
useful life, total payments, fair value, originating firm) and the prior manual
classification (e.g., financing, operating) of the company’s leases, the
model can evaluate a new lease and assign the appropriate classification.
Post-classification verification and correction in the case of page 114
an inappropriate outcome is then fed into the model to
improve the performance of the model.
Artificial intelligence models work similarly in that they learn from the
inputs and corrections to improve decision making. For example, image
classification allows auditors to take aerial photography of inventory or
fixed assets and automatically identify the objects within the photo rather
than having an auditor manually check each object. Classification of
closed-circuit footage enables automatic counting of foot traffic in a retail
location for managers. Modeling of past judgment decisions by audit
partners makes it possible to determine whether an allowance or estimate
falls within a normal range for a client and is acceptable or should be
qualified. Artificial intelligence models that track sentiment in social media
and popular press posts to predict positive stock market returns for
analysts.
For most application of artificial intelligence models, the computational
power is such that most companies will outsource the underlying system to
companies like Microsoft, Amazon, or Google rather than develop it
themselves. These companies provide the datasets to train and build the
model, and the platforms provide the algorithms and code. When public
accounting firms outsource data clients may be hesitant to allow their
financial data to be used in these platforms without additional assurance
surrounding the privacy and security of their data.
PROGRESS CHECK
11. How might you expect managers to use decision support systems
when evaluating employee bonuses?
12. How do machine learning and artificial intelligence models
improve their recommendations over time?
Summary
In this chapter, we addressed the third step of the IMPACT
cycle model: the “P” for “performing test plan.” That is, how are
we going to test or analyze the data to address a problem we
are facing?
We identified descriptive analytics that help describe what
happened with the data, including summary statistics, and
data reduction and filtering.
We provided examples of diagnostic analytics that help users
identify relationships in the data that uncover why certain
events happen through profiling, clustering, similarity
matching, and co-occurrence grouping.
We explained examples of predictive analytics and introduced
some data mining concepts related to regression,
classification, and link prediction that can help predict future
events or values.
We discussed predictive analytics, including decision support
systems and artificial intelligence and provided some example
of how these systems can make recommendations for future
actions.
We introduced some specific models and terminology related
to these tools, including Benford's law, test and training data,
decision trees and boundaries, linear classifiers, and support
vector machines. We identified cases where creating models
that overfit existing data are not very accurate at predicting the
future.
We presented some classification terminology—including test
and training data, decision trees and boundaries, linear
classifiers, and support vector machines—and talked about
the perils of under- and overfitting the training data and their
consequences in predictions using the test data.
page 115
Key Words
Benford’s law (103) An observation about the frequency of
leading digits in many real-life sets of numerical data. The law
states that in many naturally occurring collections of numbers,
the significant leading digit is likely to be small.
causal modeling (106) A data approach similar to regression,

but used when the relationship between independent and
dependent variables where it is hypothesized that the
independent variables cause or are associated with the
dependent variable.
classification (106) A data approach used to assign each unit

in a population into a few categories potentially to help with
predictions.
clustering (103) A data approach used to divide individuals (like

customers) into groups (or clusters) in a useful or meaningful
way.
co-occurrence grouping (103) A data approach used to

discover associations between individuals based on transactions
involving them.
data reduction (98) A data approach used to reduce the
amount of information that needs to be considered to focus on
the most critical items (i.e., highest cost, highest risk, largest
impact, etc.).
decision boundaries (109) Technique used to mark the split

between one class and another.
decision support system (112) An information system that

supports decision-making activity within a business by
combining data and expertise to solve problems and perform
calculations.
decision tree (109) Tool used to divide data into smaller groups.
descriptive analytics (94) Procedures that summarize existing

data to determine what has happened in the past. Some
examples include summary statistics (e.g. Count, Min, Max,
Average, Median), distributions, and proportions.
diagnostic analytics (94) Procedures that explore the current

data to determine why something has happened the way it has,
typically comparing the data to a benchmark. As an example,
these allow users to drill-down in the data and see how it
compares to a budget, a competitor, or trend.
fuzzy match (124) A computer-assisted technique of finding

matches that are less than 100 percent perfect by finding
correspondencies between portions of the text of each potential
match.
link prediction (106) A data approach used to predict a

relationship between two data items.
predictive analytics (94) Procedures used to generate a model

that can be used to determine what is likely to happen in the
future. Examples include regression analysis, forecasting,
classification, and other predictive modeling.
prescriptive analytics (95) Procedures that model data to
enable recommendations for what should be done in the future.
These typically include developing more advanced machine
learning and artificial intelligence models to recommend a
course of action based on a current problem.
profiling (100) A data approach used to characterize the

“typical” behavior of an individual, group, or population by
generating summary statistics about the data (including mean,
standard deviations, etc.).
regression (106) A data approach used to estimate or predict,

for each unit, the numerical value of some variable using some
type of statistical model.
similarity matching (106) A data approach used to identify

similar individuals based on data known about them.
structured data (100) Data that are organized and reside in a

fixed field with a record or a file. Such data are generally
contained in a relational database or spreadsheet and are
readily searchable by search algorithms.
summary statistics (97) Describe the location, spread, shape,

and dependence of a set of observations. These commonly
include the count, sum, minimum, maximum, mean or average,
standard deviation, median, quartiles, correlation covariance,
and frequency that describe a specific measurable value.
supervised approach/method (103) Approach used to learn

more about the basic relationships between independent and
dependent variables that are hypothesized to exist.
support vector machine (110) A discriminating classifier that is

defined by a separating hyperplane that works first to find the
widest margin (or biggest pipe).
page 116
training data (109) Existing data that have been manually
evaluated and assigned a class, which assists in classifying the
test data.
test data (109) A set of data used to assess the degree and
strength of a predicted relationship established by the analysis
of training data.
unsupervised approach/method (103) Approach used for data

exploration looking for potential patterns of interest.
XBRL (99) (eXtensible Business Reporting Language) A global

standard for exchanging financial reporting information that uses
XML.
ANSWERS TO PROGRESS CHECKS

1.
a. Link prediction
b. Classification
c. Regression
2. While descriptive analytics focuses on what happened,
diagnostic analytics focuses on why it happened. Descriptive
and diagnostic analytics are typically paired because you
would want to describe the past data and then compare it to a
benchmark to determine why the results are the way they are,
similar to the accounting concepts of planning and controlling.
3. Data reduction may be used to filter out ordinary travel and
entertainment expenses so an auditor can focus on those that
are potentially erroneous or fraudulent.
4. The XBRL tagging allows an analyst or decision maker to
focus on one or a category of expenses of most interest to a
lender. For example, lenders might be most interested in
monitoring the amount of long-term debt, interest payments,
and dividends paid to assess if the borrower will be able to
repay the loan. Using the capabilities of XBRL, lenders could
focus on just those individual accounts for further analysis.
5. In some sense, profiling techniques to find criminals and
accounting anomalies are very similar. Profiling to find
criminals often looks to the physical characteristics (race, sex,
mental state, etc.) to predict whether the person has or is
likely to commit a crime (and is illegal to use in some
jurisdictions). Accounting looks to other, nonphysical
characteristics such as the amounts, totals, and types of
expenditures to identify potential anomalies.
6. A dollar store might sell everything for exactly $1.00. In that
case, the use of Benford’s law for any single product or even
for every product would not follow Benford’s law!
7. Three clusters of customers who might consider Walmart
could include thrifty shoppers (looking for the lowest price),
shoppers looking to shop for all of their household needs
(both grocery and non-grocery items) in one place, and those
customers who live close to the store (good location).
8. The longer time between the death and payment dates begs
one to ask why it has taken so long for payment to occur and
if the interest required to be paid is likely large. Because of
these issues, there might be a possibility that the claim is
fraudulent or at least deserves a more thorough review to
explain why there was such a long delay.
9. We certainly could let the data speak and address this
question directly. In general, when the health of the economy
is stronger, there are fewer layoffs and fewer people out
looking for a job, which means less turnover. Additional
analysis could determine whether the turnover is voluntary or
involuntary.
10. Chapter 1 illustrated that Lending Club collects the credit
score data, and the initial analysis there suggested the higher
the credit score, the less likely to be rejected. Given this
evidence, we would predict a negative relationship between
credit score and loans that are rejected.
page 117
11. Decision support systems follow rules to determine the
appropriate amount of a bonus. Following a set of rules, the
system may evaluate management goals, such as a sales
target or number of new accounts, to calculate and
recommend the appropriate bonus compensation.
12. Machine learning and artificial intelligence models learn by
incorporating new data and through manual correction of data.
For example, when a misclassified lease is corrected, the
accuracy of the recommended classification of future leases
improves.

1. is a set of data used to assess the degree and
strength of a predicted relationship.
a. Training data
b. Unstructured data
c. Structured data
d. Test data
2. Data that are organized and reside in a fixed field with a
record or a file. Such data are generally contained in a
relational database or spreadsheet and are readily searchable
by search algorithms. The term matching this definition is:
a. training data.
b. unstructured data.
c. structured data.
d. test data.
3. An observation about the frequency of leading digits in many
real-life sets of numerical data is called:
a. leading digits hypothesis.
b. Moore’s law.
c. Benford’s law.
d. clustering.
4. Which approach to data analytics attempts to predict a
relationship between two data items?
a. Similarity matching
b. Classification
c. Link prediction
d. Co-occurrence grouping
5. In general, the more complex the model, the greater the
chance of:
a. overfitting the data.
b. underfitting the data.
c. pruning the data.
d. a more accurate prediction of the data.
6. In general, the simpler the model, the greater the chance of:
a. overfitting the data.
b. underfitting the data.
c. pruning the data.
d. the need to reduce the amount of data considered.
page 118
7. is a discriminating classifier that is
defined by a separating hyperplane that works first to find the
widest margin (or biggest pipe) and then works to find the
middle line.
a. Linear classifier
b. Support vector machine
c. Decision tree
d. Multiple regression
8. mark the split between one class and another.
a. Decision trees
b. Identified questions
c. Decision boundaries
d. Linear classifiers
9. Models associated with regression and classification data
approaches have all except this important part:
a. identifying which variables (we’ll call these independent
variables) might help predict an outcome (we’ll call this the
dependent variable).
b. the functional form of the relationship (linear, nonlinear, etc.).
c. the numeric parameters of the model (detailing the relative
weights of each of the variables associated with the
prediction).
d. test data.
10. Which approach to data analytics attempts to assign each unit
in a population into a small set of classes where the unit
belongs?
a. Classification
b. Regression
c. Similarity matching
d. Co-occurrence grouping
1. What is the difference between a target and a class?
2. What is the difference between a supervised and an
unsupervised approach?
3. What is the difference between training datasets and test (or
testing) datasets?
4. Using TABLE 3-1 as a guide, what are three data approaches
associated with the descriptive approach?
5. Using TABLE 3-1 as a guide, what are three data approaches
associated with the diagnostic approach?
6. How might the data reduction approach be used in auditing?
7. How might classification be used in approving or denying a
potential fraudulent credit card transaction?
8. How is similarity matching different from clustering?
9. How does fuzzy match work? Give an accounting situation
where it might be most useful.
10. Compare and contrast the profiling data approach and the
development of standard cost for a unit of production at a
manufacturing company. Are they substantially the same, or
do they have differences?
11. Exhibits 3-9, 3-10, and 3-13 suggest that volume and distance
are the best predictors of “days to ship” for a wholesale
company. Any other variables that would also be useful in
predicting the number of “days to ship”?
page 119
Problems
1. How could the fuzzy match be used to find undisclosed
related party transactions?
2. An auditor is trying to figure out if the inventory at an
electronics store chain is obsolete. What characteristics (e.g.,
data elements about the store, customers, employees, and
inventory) might be used to help establish a model predicting
inventory obsolescence?
3. An auditor is trying to figure out if the goodwill its client
recognized when it purchased a factory has become impaired.
What characteristics might be used to help establish a model
predicting goodwill impairment?
4. How might clustering be used to describe customers who owe
money (accounts receivable)?
5. Why would the use of data reduction be useful to highlight
related party transactions (e.g., CEO has her own separate
company that the main company does business with)?
6. How could an investor use XBRL to do an analysis of the
industry’s inventory turnover?
7. Name three accounts that would be appropriate and
interesting to apply Benford’s law in auditing those accounts.
Why would an auditor choose those three accounts? When
would a departure from Benford’s law encourage the auditor
to investigate further?
Appendix: Setting Up A Classification

Analysis
To answer the question “Will a new vendor ship a large order on
time?” using classification, you should clearly identify your
variables, define the scope of your data, and assign classes.
This is related to “master the data” in the IMPACT model.
Identify Your Variables

Because this question is related to vendors and order
shipments, take a moment to think about attributes that might be
predictive. What attributes would you need to address the
following questions: Would the total number of order items
potentially cause a delay? Are certain types of items shipped
more timely than others? How about the overall shipping weight.
. .does that impact the timeliness of shipments? Does the
vendor’s physical distance from a company's warehouse
matter? How about the age of vendor relationship or number of
vendor employees? What else?
Define the Scope

Because you are looking at vendor shipments, you would need
—at the basic level—data related to the original purchase order
(order date, number of items), shipping data (shipping date,
weight), and vendor master data (location, age, size). This will
help you narrow down your data request and make it more likely
that you’ll get the data you request by an established deadline.
As you’re preparing your data, you’ll want to join these tables so
that each record represents an order. You’ll also want to
calculate any figures of merit, such as the number of days (Ship
date – Order date), volume (total number of items on the order
or physical size) or distance (Vendor address – Warehouse
address) (see Table 3-A1).
Table 3-A1
Vendor Shipments
page 120
Distance Formula
You can use a distance formula in Excel to calculate the
distance in miles or kilometers between the warehouse and the
vendor. First, you determine the latitude and longitude based on
the address, then use the following formula. Note: Use first
number 3959 for miles or 6371 for kilometers.
3959 * ACOS(SIN(RADIANS([Lat])) * SIN(RADIANS([Lat2])) + COS(RADIANS([Lat])) *
COS(RADIANS([Lat2])) * COS(RADIANS([Long2]) – RADIANS([Long])))
Assign Classes
Take a moment to define your classes. You are trying to predict
whether a given order shipment will either be “On-time” or
“Delayed” based on the number of days it takes from the order
date to the shipping date. What does “on-time” mean? Let’s
define “on-time” as an order that ships in 5 days or less and a
“delayed” order as one that ships later than 5 days. You’ll use
this rule to add the class as a new attribute to each of your
historical records (see Table 3-A2).
Table 3-A2
Shipment Class
On-time = (Days to ship ≤ 5)

Delayed = (Days to ship > 5)
page 121
Lab 3-1 Data Reduction Using Fuzzy

Matching
Auditors use data reduction to focus their efforts on testing
internal controls and limiting their scope. For example, they may
want to look only at transactions for a given year. In this lab, you
will learn to use filters in Excel and perform some fuzzy matches
on vendor and employee records, a common auditor analysis.
Company summary
These data are for a generic manufacturing company. You have
been asked to see if there are any potentially fictitious vendors
or employees who may have created fake companies in an
effort to commit fraud.
Data
Fuzzy.xlsx—contains employee and vendor data
Technique
Some Excel experience is handy here. You will use tables,
filters, and the Fuzzy Lookup add-in.
Software needed
Excel
Fuzzy Lookup add-in: https://www.microsoft.com/en-
us/download/details.aspx?id=15011 Note: this add-in requires
the Windows version of Excel. For Excel 2019 or later, see
Software Instructions on Connect.
Part 1: Identify a problem that will require data reduction

techniques.
Part 2: Master the data and prepare for analysis.
Part 3: Perform data reduction.
Part 1: Identify the Problem

Fictitious vendors represent risk to a company. One way
employees can embezzle funds from a company is to create a
fictitious vendor (a.k.a., shell vendor) and then submit an invoice
for services that were never performed. Where there are poor
internal controls, the employee receives the payment and
deposits the check.
Q1. What data do you think might exist to show that a
vendor is related to an employee? Which attributes
would you focus on?
Q2. How might you attempt to detect these connections
between vendors and employees?
Q3. If you were the employee committing fraud, what

would you try to do with the data to evade detection?
Part 2: Master the Data and Prepare for

Analysis
You have requested the employee and vendor master data
tables to aid in your analysis. The IT supervisor has sent you an
Excel sheet with the following tables and attributes:
Employees
EmployeeID
EmployeeFirstName
EmployeeLastName
page 122
EmployeeGender
EmployeeHireDate
EmployeeStreetAddress
EmployeeCity
EmployeeState
EmployeeZip
EmployeePhone
Vendors
VendorID
VendorName
VendorType
VendorSince
VendorContact
VendorBillingAddress
VendorBillingCity
VendorBillingState
VendorBillingZip
VendorBillingPhone
Your first step is to understand the data and prepare it in

Excel to perform some matching.
1. Open Fuzzy.xlsx in Excel.
2. Quickly browse through the worksheets to ensure that they
are complete.
3. Go to the Employees tab and click any data element.
4. Select the entire data table (Ctrl + A).
5. Go to the Home tab, Styles section, and click Format as
Table. Any style will do.
6. In the Format as Table box that appears, make sure the My
table has headers box is checked, and click OK.
7. In the Table Tools > Design tab, under Properties, change
the table name from Table1 to Employees.
8. Now go to the Vendors tab and click any data element.
Repeat steps 4–7 and name the new table Vendors.
9. Take a screenshot of either table (label it 3-1A).
10. Save your file as Fuzzy-Tables.xlsx.
Part 3: Perform Data Reduction

Now you’re ready to find those fictitious vendors. There are
many different approaches for working with the data to narrow
your focus. These can be used with other data sources as well.
Tool: Filtering
Excel Filters allow you to quickly find data with common
attributes and help to limit the scope of your analysis. Assume
that the auditors have analyzed all vendors prior to 2019 and
have resolved any outstanding issues. By analyzing only the
vendors from 2019, you avoid unnecessary analysis and reduce
the time it will take for the computer to run the analysis.
11. Open Fuzzy-Tables.xlsx and click the Vendors worksheet.
12. Click the drop-down arrow next to VendorSince to show
filtering options, shown below.
page 123
13. To select only 2019 records, uncheck Select All
and then check the box next to 2019 and click OK.
Source: Microsoft Excel 2016.
14. Select the table and headers (Ctrl + A twice) and copy the
values (Ctrl + C).
15. Create a new worksheet tab called Vendors2019 and paste
the filtered values there (Ctrl + V).
16. Select your new table and format it as a table called
“Vendors2019.”
18. Save your file as Fuzzy-Tables-2019.xlsx.
Tool: Fuzzy Match

SQL queries and PivotTables require exact matches between
two data points to identify related data. Foreign keys must match
primary keys exactly in databases or else a relationship doesn’t
exist. Names and addresses, as well as other manually entered
text values are more prone to errors and manipulation. Think
about your questions from Part 1 of this lab. Wouldn’t one way
to avoid detection be to change something subtle in the address
—for example “Street” to “St.” or “Center Ave” to “Center”? A
human could understand that these are the same thing, but a
computer cannot without some help.
Fuzzy Lookup is a plugin for Excel that enables these mostly
similar matches and finds things that might otherwise evade
detection by a computer system.
19. Download and enable Fuzzy Lookup for Excel if you haven’t
already.
20. Open Fuzzy-Tables-2019.xlsx if you haven’t already.
page 124
21. In the ribbon, click Fuzzy Lookup > Fuzzy
Lookup. A panel will appear on the right showing the tables
you defined in Part 2 of this lab.
22. For the Left Table, choose Vendors2019, and for the Right
Table, choose Employees.
23. In the Left Columns list, click VendorBillingAddress, and
from the Right Columns list, click EmployeeStreetAddress.
24. Click the Join icon button in between the two lists. A new
relationship will appear in the Match Columns list.
25. In the Output Columns list, uncheck everything except:
a. Vendors2019.VendorName
b. Vendors2019.VendorContact
c. Vendors2019.VendorBillingAddress
d. Vendors2019.VendorBillingZip
e. Vendors2019.EmployeeFirstName
f. Vendors2019.EmployeeLastName
g. Vendors2019.EmployeeStreetAddress
h. Vendors2019.EmployeeZip
i. Vendors2019.FuzzyLookup.Similarity
26. Number of Matches should be 1, suggesting the need for an
exact match.
27. The Similarity Threshold slider represents the percentage
similarity. You can slide it left if you want less similar matches
or slide it right for more similar. 1.0 is an exact match.
28. Create a new worksheet tab called FuzzyMatch.

29. Click cell A1, then click Go in the Fuzzy Lookup panel. A list
will appear with all of the vendors and any potential fuzzy
match.
page 125
30. Format the output as a table named
FuzzyMatch, then filter out any records with 0.0000 Similarity.
Q4. How many vendors have similar addresses to
employees?
Q5. What do you notice about the street vendor and
employee street addresses?
Q6. Are there any false positives (fuzzy matches that

aren’t really matches)?
31. Take a screenshot (label it 3-1C).
End of Lab
Lab 3-2 Regression in Excel
Company summary
The data used are a subset of the College Scorecard dataset
that is provided by the U.S. Department of Education. These
data provide federal financial aid and earnings information,
insights into the performance of schools eligible to receive
federal financial aid, and the outcomes of students at those
schools. You can learn more about how the data are used and
view the raw data yourself at
https://collegescorecard.ed.gov/data/. However, for this lab, you
should use the text file provided to you.
Data
CollegeScorecard Datasets: CollegeScorecard_CleanedData
from Lab 2-5
Technique
Software needed
Excel
Screen capture tool (Windows: Snipping Tool; Mac: Cmd + Shift
+ 4)
Part 1: Identify the questions.

Part 2: Load the data.
Part 3: Perform a regression analysis in Excel.
This lab relies upon the steps completed in Lab 2-5 in which the
data were prepared.
We will begin with a simple regression with two variables,
SAT average and completion rate for first-time, full-time students
at four-year institutions.
Q1. Would you expect SAT average and completion rate

to be correlated? If so, would you expect the
correlation to be positive or negative?
Q2. When determining relationships between variables,

one of the criteria for a potential causal relationship
is that the cause must happen before the effect.
Regarding SAT average and completion rate, which
would you determine to be the potential cause?
Which would be the effect?
Q3. Identifying the cause and effect as you did in Q2 can
help you determine the explanatory and response
variables. Which variable, SAT average or
completion rate, is the explanatory variable?
page 126

These steps were performed in Lab 2-5. You can either use the
already prepared data in the file Lab 3-2, College Scorecard
Data (Cleaned post-Lab 2-5).xlsx, or you can use the file that
you saved after completing Lab 2-5. (See Appendix K for the
data dictionary for this dataset.)
Part 3: Perform an Analysis of the Data

1. To perform a regression test in Excel, you need to first
download the Data Analysis ToolPak. To do so, Follow this
menu path: File > Options > Add-ins. From this window,
select the Go . . . button, and then place a checkmark in the
box next to Analysis ToolPak. Once you click OK, you will be
able to access the ToolPak from the Data tab on the Excel
ribbon.
2. Click the Data Analysis button from the Data tab on the Excel
ribbon and select Regression.
3. A regression window will pop up for you to input the Y range
and the X range.
LAB EXHIBIT 3-2A

4. Select the entire column of data that is associated with the

response variable for the Y range, then select the entire
column of data that is associated with the explanatory variable
for the X range. For a description of the variables, refer to the
data dictionary in Appendix K.
5. If you selected the labels in your ranges, place a checkmark in
the box next to Labels.
6. Click OK. This will run the regression test and place the
output on a new spreadsheet in your Excel workbook.
7. Take a screenshot of your regression output (label it 3-2A).
End of Lab
page 127
Lab 3-3 Classification
Company summary
LendingClub is a peer-to-peer marketplace where borrowers
and investors are matched together. The goal of LendingClub
is to reduce the costs associated with these banking
transactions and make borrowing less expensive and
investment more engaging. LendingClub provides data on
loans that have been approved and rejected since 2007,
including the assigned interest rate and type of loan. This
provides several opportunities for data analysis.
Data
Download the LendingClub datasets: LoanStats3a.csv,
RejectStatsA.csv, LCClass.csv from the Connect website
Software needed
Excel
Weka—available at www.cs.waikato.ac.nz/ml/weka
Screen capture tool (Windows: Snipping Tool; Mac: Cmd + Shift
+ 4)

Analyze the data using various classification models.
Thinking about LendingClub’s function as a marketplace for
investors and borrowers, what might stakeholders want to
know? Has LendingClub’s model changed over the years? If
we understood what affected an interest rate decision, could we
game the system to our advantage? Take a moment and come
up with some general questions that could be answered through
data analysis.
Q1. Thinking about loan applicants in general, how would
you expect them to fall into different groups?
Q2. When evaluating previous loan data, what would you
expect your target variable to be?
Q3. What factors do you think would affect whether a
loan will be accepted or rejected?
Q4. Identify the data you would need to answer your

questions and validate your hypothesis.

For this lab, you should download the LendingClub data and
prepare it for some more advanced analysis in Chapter 3.
Note: You have the choice to manually clean and combine the
data here from LoanStats3a.csv and RejectStatsA.csv or use
the cleaned data in LCClass.csv file. If you run into any issues
cleaning the data, skip to the completed file and continue to Part
3.
The LendingClub data contains two different file types

including LoanStats3a for approved loans and RejectStatsA for
rejected loans. There are significantly more data available for
LoanStats3a. There are 107 different attributes. To save some
time, we’ve identified 19 of the most interesting in Lab Tables 3-
3A and 3-3B.
LAB TABLE 3-3A

LoanStats3a.csv
member_id Membership id
loan_amount Requested loan amount
purpose Loan purpose: e.g., wedding, medical,
debt_consolidation, car
zip_code The first three digits of the applicant’s zip
code
addr_state State
dti Ratio of monthly debt payments to monthly
income
delinq_2y Late payments within the past two years
LAB TABLE 3-3B

RejectStatsA.csv
Amount Requested loan amount
Requested
Application Date Date of loan application
Loan Title Brief description of loan purpose
Risk_Score LendingClub’s calculated value
Debt-To-Income Ratio of monthly debt payments to monthly
Ratio income
Zip Code The first three digits of the applicant’s zip
code
State State
Employment Employment length
Length
Policy Code Internal number
Q5. Given this list of attributes, what concerns do you

have with the data’s ability to predict answers to the
questions you identified before?
Between the two groups of data files, we notice that there
are some attributes in common, though not very many (see Lab
Table 3-3C).
page 128
LAB TABLE 3-3C

Common Data Model
Common RejectStatsA.csv LoanStats3a.csv

Amount Amount Requested loan_amnt
Month =MONTH(“Application =MONTH(“issue_d”)
Date”)
Purpose Loan Title purpose
Common RejectStatsA.csv LoanStats3a.csv
DebtToIncome Debt-to-Income dti
State State addr_state
YearsOfEmployment Employment emp_length
Class REJECT ACCEPT
Q6. What does the lack of attributes in the RejectStatsA

files tell us about the data that LendingClub retains
on rejected loans?
Q7. How will that affect a classification analysis?
We will need to convert the data into a useful format before
we can perform any analysis. We need to generate two sets of
data, one for classification and one for regression and
clustering.
page 129
Cleaning the Data for Classification

Goal: Combine approved and rejected data for a given year,
assign a class to each record.
Issues
Approved and rejected loans contain different data attributes.
Date data values are recorded in different formats (1/9/2011 vs.
Jan-2011).
Years of employment contain text values and should be
numbers.
In Excel
1. Create a new spreadsheet.
2. Type the common attributes from Table 3-3C into the first row.
3. Open the LoanStats3a.csv and RejectStatsA.csv files in
Excel.
4. Delete all columns that don’t match those listed in Table 3-3C
and remove the summary data at the bottom of the file and
the link reference in the first row of the file.
5. Use the =MONTH formula to extract the month from the date.
6. Copy the Month column and Paste Special > Values into the
Month column.
7. Use find and replace to remove any % signs from the
DebtToIncome values.
8. Add a new Class column and enter REJECT to the rejected
loans and APPROVE to the approved loans.
9. Copy and paste the values from each .csv file into your new
spreadsheet. Note: The order of the columns may be different
in the source files, so make sure you could either reorganize
the columns in the source files or copy and paste each
column at a time into the correct column in the new sheet.
10. Find and replace the employment values using Lab Table 3-
3D.
LAB TABLE 3-3D
Original Value New Value

na or n/a 0
< 1 year 0
1 year 1
2 years 2
3 years 3
4 years 4
5 years 5
6 years 6
7 years 7
8 years 8
9 years 9
10+ years 10
, (comma) (blank)
11. Save your file as LCClass.csv. Be sure to choose .csv as the
file type.

We will try multiple classification models and compare their
results using Weka. Download Weka at
www.cs.waikato.ac.nz/ml/weka.
13. Open Weka and click the Explorer button.
14. Click Open file . . . Change the Files of Type to CSV data files
(*.csv) and then locate your LCClass.csv file. Then click
Open.
15. Click Visualize All found just above and right of the graph.
page 130
17. Click the Classify tab.
18. Run each of the following classification models. For each tool,
click on Choose . . . and then Start to run the analysis. Note:
watch the status bar at the bottom to know when a model is
finished running. The results will appear on the screen
showing the classification matrix and resulting model. As you
complete these models, make a note of the accuracy
percentage for each. This shows what percent of the loans
was correctly classified.
a. Weka > Classifiers > Trees > Random Forest.
b. Weka > Classifiers > Meta > AdaBoostM1.
c. Weka > Classifiers > Functions > Logistic.
d. Weka > Classifiers > Bayes > BayesNet.
Q8. Which model has the highest accuracy? How do you
know?

Review the accuracy rates of each of the models. As a rule of
thumb, a good model has an accuracy rate of 70% or higher,
though 100% may indicate overfitting. Accuracy of 50%–70% is
okay, but not great. Less than that and you’re better off flipping a
coin to choose a class.
Q9. How useful is your classification model in predicting
which applicants will be approved or rejected? How
do you know?
Review the details of models, including the accuracy

percentage and the decision boundaries of each of the variables
in Weka.
Q10. How would you interpret the results of your analysis
in plain English?
End of Lab
Lab 3-4 Comprehensive Case: Dillard’s

Store Data: Data Abstract (SQL) and
Regression (Part I)
Company summary
Dillard’s is a department store with approximately 330 stores in
29 states. Its headquarters is in Little Rock, Arkansas. You can
learn more about Dillard’s by looking at finance.yahoo.com
(Ticker symbol = DDS) and the Wikipedia site for DDS. You’ll
quickly note that William T. Dillard II is an accounting grad of the
University of Arkansas and the Walton College of Business,
which may be why he shared transaction data with us to make
Data
accessed through the University of Arkansas Remote Desktop.
Directions for accessing the Remote Desktop can be found at
www.mhhe.com/richardsondaa2e. See your instructor for login
credentials.
Software needed
Microsoft SQL Server Management Studio (available on the
Remote Desktop at the University of Arkansas)
Excel 2016 (available on the Remote Desktop at the University
of Arkansas)

Conduct analysis on three important questions that help us
understand when customers spend more on individual
transactions.
page 131

Dillard’s is trying to figure out when its customers spend more
on individual transactions. We ask questions regarding how
Dillard’s sells its products.
Q1. Customers in which states had the highest
transaction balances over the entire sample period?
Q2. Do customers in the state with the highest
transaction balances have a significantly higher
transaction balance from September 1, 2016, to
September 15, 2016, than all other states?
Q3. Are online transaction amounts statistically greater
than or lesser than non-online transactions during
the time period September 1, 2016, to September
15, 2016?

For this lab, you should access the TRANSACT and the STORE
tables from the Dillard’s WCOB_DILLARDS dataset from the
University of Arkansas. You may have learned how to do so
from past labs, or feel free to ask your instructor for access. You
can also learn how to do so in Appendix F.
1. Run the following SQL query in Microsoft SQL Server
Management Studio to address the first question to identify
which state had the highest customer transaction balance.
(Recall that transaction is defined for each individual item
purchased.)
Refer to Appendix J of the textbook to view an ER
Diagram for the Dillard’s data.
SELECT STATE, AVG(TRAN_AMT) AS Average
FROM TRANSACT
INNER JOIN STORE
WHERE TRAN_DATE BETWEEN ’20160901’ AND
’20160915’
GROUP BY STATE
ORDER BY AVG(TRAN_AMT) DESC
The output should return the following:
AR 28.455765
SC 20.544199
LA 20.530411
TX 20.482441
...
Take a screenshot of your results in SQL Server(label it 3-4A).

2. Noting that Arkansas (State =‘AR’) has the highest transaction
balance, let’s address our second question: “Do customers
in the state with the highest transaction balances
have a significantly higher transaction balance
from September 1, 2016, to September 15, 2016,
than all other states?”
3. To address Q2, you will need to connect to SQL Server data
through Excel using Data > Get & Transform and the Analysis
ToolPak. If you need a refresher on how to use that tool, refer
to the textbook’s Appendix B.
Once you connect to the SQL Server dataset in Excel and
expand the options to input a query, input the following SQL
query to extract the data needed for our analysis.
page 132
SELECT TRANSACT.*, STORE.STATE
FROM TRANSACT
INNER JOIN STORE
WHERE TRAN_DATE BETWEEN ’20160901’ AND
’20160909’
ORDER BY TRAN_DATE
4. Once the query runs, there is no need to transform the data in
Power Query for this lab. Immediately load the data into
Excel.
5. Once the data are in Excel, you’ll need to transform the State
data to perform regressionanalysis on the state of Arkansas to
address Q2. To do so, make a new column just right of the
existing dataset and label it Arkansas-dummy in column Q.
Type the formula =IF([@STATE]=“AR”,1,0) in each row. It will
assign a value of 1 to transactionsat stores in Arkansas and a
value of 0 for transactions at stores outside of the Arkansas.
Copy this formula all the way down to cover each row.
6. Perform a regression analysis by performing the following

steps.
a. Click on Data Analysis button in the Data tab in the ribbon.
If you do not have the Data Analysis ToolPak added in, see
the Appendix B, to learn how to add it to Excel.
b. Click Regression (as shown below), and then click OK.
c. Reference the cells that contain the Tran_AMT in the Input

Y Range and Arkansas-dummy in the Input X Range and
then click OK.
page 133
d. Your output should look like the screenshot below. The t

Stat greater than 2.0 suggests that the transaction amount
(Tran_Amt) is statistically greater in Arkansas than in all
other states.
7. Take a screenshot of your results (label it 3-4B).

We are now ready to address our third question: Are online
transaction amounts statistically greater than or
lesser than non-online transactions during the time
period September 1, 2016, to September 15, 2016?
Because we found that transactions in Arkansas are
statistically higher than all other states, we will include that
finding in our analysis as well, making this a multivariate
regression.
8. To address this question, we need to transform the Store
variable into an online-dummy variable. Dillard’s handles all
of their online sales through a fulfillment center in Maumelle,
Arkansas. This location has the store number 698. We will
use an IF statement to transform all transactions in store
number 698 into a 1 to indicate they are online sales, and all
other transactions will be indicated as a 0. The IF statement is
the following (we have hidden a few columns in the middle of
the table to make the screenshot easier to read):
page 134
Once this is complete, copy the calculation for all cells in

the column.
We’re now ready for regression analysis. Reference the
cells that contain the Tran_AMT in the Input Y Range and
reference Arkansas-dummy and Online-dummy in the
Input X Range and click OK.
The results of the regression analysis suggesting that both

Transactions in Arkansas and Transactions done online are
associated with greater transaction amounts are below.
page 135
9. Take a screenshot of your results (label it 3-4C).

of Q1 in plain English? Why do you think the state of
Arkansas had the highest transaction volume?
Q5. The analysis of Q2 addressed whether Arkansas had

a statistically higher transaction volume than other
states. How did the regression tests show or not
show a statistical difference? Does this have any
implications for the marketing for Dillard’s? Do you
think it is because Arkansas is the home base for
Dillard ’s?
Q6. The regression analysis suggests that online sales
are associated with greater transactions amounts.
Why do you think that is so?
End of Lab
Lab 3-5 Comprehensive Case: Dillard’s

Store Data: Data Abstract (SQL) and
Regression (Part II)
Company summary
Dillard’s is a department store with approximately 330 stores in
29 states. Its headquarters is in Little Rock, Arkansas. You can
learn more about Dillard’s by looking at finance.yahoo.com
(Ticker symbol = DDS) and the Wikipedia site for DDS. You’ll
quickly note that William T. Dillard II is an accounting grad of the
University of Arkansas and the Walton College of Business,
which may be why he shared transaction data with us to make
page 136
Data
accessed through the University of Arkansas Remote Desktop.
Directions for accessing the Remote Desktop can be found at
www.mhhe.com/richardsondaa2e. See your instructor for login
credentials.
Software needed
Excel 2016 (available on the Remote Desktop at the University
of Arkansas)

Conduct analysis on one important question that helps us
understand when customers spend more on individual
transactions.

Dillard’s is trying to figure out when its customers spend more
on individual transactions. We ask questions regarding how
Dillard’s sells its products.
Q1. Did customers who charged their purchases to a

Dillard’s credit card spend less on each transaction
during the time period September 1, 2016, to
September 9, 2016?

1. See Lab 3-4 to see how to access the dataset.

2. To get the necessary data to address Q1, you will need to run
the same query as Lab 3-4 and get the data into Excel ready
for analysis.
3. For those who use the Dillard’s credit card, it is noted as
“DLRD” in the TENDER_TYPE field. We need to compare
those who use the Dillard’s credit card to all other
transactions. To prepare for this analysis, we need to make a
DLRD-dummy variable (labeled as “DLRD-dummy”) that
carries the value of 1 if a Dillard’s credit card was used and a
value of 0 otherwise. Please make the transformation in this
way and copy down for every row as in Lab Exhibit 3-5A.
LAB EXHIBIT 3-5A

page 137
4. Once this is complete, we are ready for statistical
analysis. Given the results of Lab 3-4 that transaction amount
is positively associated with the state of Arkansas and online
sales, we will include them in our multivariate analysis. Click
the Data Analysis button and select Regression as pictured
below.
LAB EXHIBIT 3-5B

5. Click Regression and include Tran_Amt as the Y variable
and Arkansas-dummy, Online-dummy, and DLRD-dummy
as the X variables and run the regression.
6. Take a screenshot (label it 3-5A) of your results.

of Q1 in plain English? Was the relationship
statistically significant (Was the t-statistic greater
than 2)? Why do you think the use of the Dillard’s
card increased the amount spent on each
transaction?
Q3. Why did we also include Arkansas state sales and
online sales as other explanatory variables (X- or
independent variables) in this regression analysis?
Are these results still significant after the inclusion of
the use of the Dillard’s credit card?
Q4. Are there any other data from the TRANSACT table
that might help us predict the transaction amount?
Q5. If we had any other data to predict transaction

amount, what would you use? Brainstorm freely to
come up with what could explain these different
levels of transaction amounts!
End of Lab
1 http://www.washingtonpost.com/wp-dyn/content/article/2005/07/14/AR2005071402055.html
(accessed August 2, 2017).
2 http://www.cpafma.org/articles/inside-public-accounting-releases-2015-national-benchmarking-
report/ (accessed November 9, 2016).
3 A. S. Ahmed, C. Takeda, and S. Thomas, “Bank Loan Loss Provisions: A Reexamination of
Capital Management, Earnings Management and Signaling Effects,” Journal of Accounting and
Economics 28, no. 1 (1999), pp. 1–25.
4 http://www.pwc.com/us/en/cfodirect/publications/in-brief/fasb-new-impairment-guidance-financial-
instruments.html (accessed November 9, 2016).
page 138
Chapter 4
Communicating Results and
Visualizations

This chapter wraps up the introduction to the IMPACT model by
explaining how to communicate your results through data
visualization and through written reports. Creating a chart takes
more skill and practice than simply adding in a bar chart through the
Excel chart wizard, and this chapter will help you identify the purpose
for your data visualization so that you can choose the best chart for
your dataset. We will also help you learn how to refine your chart so
that it communicates as efficiently and effectively as possible. The
chapter concludes by describing how to provide a written report
tailored to specific audiences who will be interested in the results of
your data analysis.
A Look Back
In Chapter 3, we considered various models and techniques used for
data analytics and discussed when to use them and how to interpret
the results. We also provided specific accounting-related examples
of when each of these specific data approaches and models is
appropriate to address our particular question.
A Look Ahead
The next chapter looks at one application of DA in accounting—
auditing. Chapter 5 considers how both internal and external auditors
are using technology in general—and audit analytics specifically—to
evaluate firm data and generate support for management assertions.
We emphasize audit working papers, audit planning, continuous
monitoring, and continuous data assurance.
page 139
One of the first uses of a heat map as a form of data visualization is also one of
history’s most impactful. In the mid-1800s, there was a worldwide cholera pandemic.
Scientists were desperate to determine the cause to put a stop to the pandemic, and
one of those scientists, John Snow, studied a particular London neighborhood that
was suffering from a large number of cholera cases in 1854. Snow created a map of
the outbreak that included small bar charts on the streets indicating the number of
people affected by the disease across different locations in the neighborhood. He
suspected that the outbreak was linked to water, so he also drew small crosses on
the map to indicate water sources. Through this visualization, Snow was able to
identify that the people who were dying nearly all had one thing in common—they
were drinking out of the same water source. This led to the discovery of cholera
being conveyed through contaminated water. Exhibit 4-1A shows Snow’s 1854
cholera map.
EXHIBIT 4-1A
Source: John Snow. On the Mode of Communication of Cholera. 2nd ed. London:
John Churchill, 1855.
Software and methods for creating heat maps to visualize epidemics has
improved since 1854, but the purpose still exists. Using a heat map to visualize
clusters of people impacted by epidemics helps researchers, health professionals,
and policy makers identify patterns and ultimately inform decisions about how to
resolve epidemics. For example, in Exhibit 4-1B this map can help readers quickly
come to insight about where the overdose epidemic is most prevalent.
EXHIBIT 4-1B
Source: CDC
Without Snow’s hypothesis, methods for testing it, and ultimately communicating
the results through data visualization, the 1854 cholera outbreak would have
continued with scientists still being uncertain of the cause of cholera.
page 140
OBJECTIVES
LO 4-1 Determine the purpose of your data visualization.

LO 4-2 Choose the best chart for your dataset.
LO 4-3 Refine your chart to communicate efficiently and
effectively.
LO 4-4 Communicate your results in a written report.
page 141
Data are important, and data analytics are effective, but they are
only as important and effective as we can communicate and make
the data understandable. One of the authors often asks her students
what they would do if they were interns and their boss asked them to
supply information regarding in which states all of the customers
their organization served were located. Would they simply point their
boss to the Customers table in the sales database? Would they go a
step further and isolate the attributes to the Company Name and the
State? Perhaps they could go a step further and run a quick query or
PivotTable to perform a count on the number of customers in each
different state that the company serves. If they were to give their
boss what she actually wanted, however, they should provide a short
written summary of the answer to the research question, as well as
an organized chart to visualize the results. Data visualization isn’t
just for people who are “visual” learners. When the results of data
analysis are visualized appropriately, the results are made easier
and quicker to interpret for everybody. Whether the data you are
analyzing are “small” data or “big” data, they still merit synthesis and
visualization to help your stakeholders interpret the results with ease
and efficiency.
Think back to some of the first data visualizations and
categorizations you were exposed to (the food guide pyramid/food
plate, the animal kingdom, the periodic table) and, more modernly,
how frequently infographics are applied to break down a series of
complicated information on social media. These charts and
infographics make it easier for people to understand difficult
concepts by breaking them down into categories and visual
components.
DETERMINE THE PURPOSE OF YOUR
DATA VISUALIZATION
LO 4-
1
Determin
e the
purpose
of your
data
visualiza
tion.
As with selecting and refining your analytical model, communicating

results is more art than science. Once you are familiar with the tools
that are available, your goal should always be to share critical
information with stakeholders in a clear, concise manner. This could
involve a chart or graph, a callout box, or a few key statistics.
Visualizations have become very popular over the past three
decades. Managers use dashboards to quickly evaluate key
performance indicators (KPIs) and quickly adjust operational tasks;
analysts use graphs to plot stock price and financial performance
over time to select portfolios that meet expected performance goals.
In any project that will result in a visual representation of data, the
first charge is ensuring that the data are reliable and that the content
necessitates a visual. In our case, however, ensuring that the data
are reliable and useful has already been done through the first three
steps of the IMPACT model.
At this stage in the IMPACT model, determining the method for
communicating your results requires the answers to two questions:
1. Are you explaining the results of previously done analysis, or are
you exploring the data through the visualization? (Is your purpose
declarative or exploratory?)
2. What type of data is being visualized (conceptual, qualitative data or
data-driven, quantitative data)?
Scott Berinato, senior editor at Harvard Business Review,
summarizes the possible answers to these questions1 in a chart
shown in Exhibit 4-2. The majority of the work that we will do with the
results of data analysis projects will reside in quadrant 2 of Exhibit 4-
2, the declarative, data-driven quadrant. We will also do a bit of work
in Exhibit 4-2’s quadrant 4, the data-driven, exploratory quadrant.
There isn’t as much qualitative work to be done, although we will
work with categorical qualitative data occasionally. When we do work
with qualitative data, it will most frequently be page 142
visualized using the tools in quadrant 1, the
declarative, conceptual quadrant.
EXHIBIT 4-2
The Four Chart Types
Source: S. Berinato, Good Charts: The HBR Guide to Making Smarter, More Persuasive
Data Visualizations (Boston: Harvard Business Review Press, 2016).
Once you know the answers to the two key questions and have
determined which quadrant you’re working in, you can determine the
best tool for the job. Is a written report with a simple chart sufficient?
If so, Word or Excel will suffice. Will an interactive dashboard and
repeatable report be required? If so, Tableau may be a better tool.
Later in the chapter, we will discuss these two tools in more depth,
along with when each should be used.
Quadrants 1 and 3 versus Quadrants 2 and

4: Qualitative versus Quantitative
Qualitative data are categorical data. All you can do with these data
is count them and group them, and in some cases, you can rank
them. Qualitative data can be further defined in two ways, nominal
data and ordinal data. Nominal data are the simplest form of data.
Examples of nominal data are hair color, gender, and ethnic groups.
If you have a set of data on people with different hair color, you can
count the number of individuals who fit into the same hair color
category, but you cannot rank it (brown hair isn’t better than red
hair), nor can you take an average or do any other further
calculations beyond counting (you can’t take an average of
“blonde”). Increasing in complexity, but still categorized as qualitative
data, are ordinal data. Ordinal data can also be counted and
categorized like nominal data but can go a step further—the
categories can also be ranked. Examples of ordinal data include
gold, silver, and bronze medals, 1–5 rating scales on teacher
evaluations, and letter grades. If you have a set of data of students
and the letter grades they have earned in a given course, you can
count the number of instances of A, B, C, and so on, and you can
categorize them, just like with nominal data. You can also sort the
data meaningfully—an A is better than a B, which is better than a C,
and so on. But that’s as far as you can take your calculations—as
long as the grades remain as letters (and aren’t transformed into the
corresponding numerical grade for each individual), you cannot
calculate an average, standard deviation, or any other more complex
calculation.
Beyond counting and possibly sorting (if you have ordinal data),
the primary statistic used with qualitative data is proportion. The
proportion is calculated by counting the number of items in a
particular category, then dividing that number by the total number of
observations. For example, if you had a dataset of 150 people and
had each individual’s corresponding hair color with 25 people in your
dataset having red hair, you could calculate the proportion of red-
haired people in your dataset by dividing 25 (the number of people
with red hair) by 150 (the total number of page 143
observations in your dataset). The proportion of red-
haired people, then, would be 16.7 percent.
Qualitative data (both nominal and ordinal) can also be referred
to as “conceptual” data because such data are text-driven and
represent concepts instead of numbers.
Quantitative data are more complex than qualitative data
because not only can they be counted and grouped just like
qualitative data, but the differences between each data point are
meaningful—when you subtract 4 from 5, the difference is a
numerical measure that can be compared to subtracting 3 from 5.
Quantitative data are made up of observations that are numerical
and can be counted and ranked, just like ordinal qualitative data, but
that can also be averaged. A standard deviation can be calculated,
and datasets can be easily compared when standardized (if
applicable).
Similar to qualitative data, quantitative data can be categorized
into two different types: interval and ratio. However, there is some
dispute among the analytics community on whether the difference
between the two datasets is meaningful, and for the sake of the
analytics and calculations you will be performing throughout this
textbook, the difference is not pertinent. The simplest way to express
the difference between interval and ratio data is that ratio data have
a meaningful 0 and interval data do not. In other words, for ratio
data, when a dataset approaches 0, 0 means “the absence of.”
Consider money as ratio data—we can have 5 dollars, 72 dollars, or
8,967 dollars, but as soon as we reach 0, we have “the absence of”
money.
Interval data do not have a meaningful 0; in other words, in
interval data, 0 does not mean “the absence of” but is simply another
number. An example of interval data is the Fahrenheit scale of
temperature measurement, where 90 degrees is hotter than 70
degrees, which is hotter than 0 degrees, but 0 degrees does not
represent “the absence of” temperature—it’s just another number on
the scale.
Due to the “meaningful 0” difference between interval and ratio
data, ratio data are considered the most sophisticated form of data.
This is because the meaningful zero allows us to calculate fractions,
proportions, and percentages—ratios reflecting the relationship
between values. However, we can perform all other arithmetic
functions on both interval and ratio data. In Chapter 3, you learned
more about statistical tests such as hypothesis testing, regression,
and correlation. We can run all of these tests and calculate the
mean, median, and standard deviation on interval and ratio data.
Quantitative data can be further categorized as either discrete or
continuous data. Discrete data are data that are represented by
whole numbers. An example of discrete data is points in a basketball
game—you can earn 2 points, 3 points, or 157 points, but you
cannot earn 3.5 points. On the other hand, continuous data are
data that can take on any value within a range. An example of
continuous data is height: you can be 4.7 feet, 5 feet, or 6.27345
feet. The difference between discrete and continuous data can be
blurry sometimes because you can express a discrete variable as
continuous—for example, the number of children a person can have
is discrete (a woman can’t have 2.7 children, but she could have 2 or
3), but if you are researching the average number of children that
women aged 25–40 have in the United States, the average would be
a continuous variable. Whether your data are discrete or continuous
can also help you determine the type of chart you create because
continuous data lend themselves more to a line chart than do
discrete data.
A Special Case of Quantitative Data: The

Normal Distribution
Chapter 3 mentions the concept of the normal distribution in the
context of profiling in continuous auditing. The normal distribution is
a phenomenon that many naturally occurring datasets in our world
follow, such as SAT scores and heights and weights of newborn
babies. For a distribution of data to be considered normal, the data
should have equal median, mean, and mode, with half page 144
of the observations falling below the mean and the other
half falling above the mean. If you are comparing two datasets that
follow the normal distribution, even if the two datasets have very
different means, you can still compare them by standardizing the
distributions with Z-scores. By using a formula, you can transform
every normal distribution into a special case of the normal
distribution called the standard normal distribution, which has 0
for its mean (and thus, for its mode and median, as well) and 1 for its
standard deviation. The benefit of standardizing your data during a
comparison of two datasets is to no longer have to compare wildly
different numbers and attempt to eyeball how one observation differs
from the other—if you standardize both datasets, you can place both
distributions on the same chart and more swiftly generate insights.
Quadrants 1 and 2 versus Quadrants 3 and

4: Declarative versus Exploratory
In the context of the labs and tools we’re providing through this
textbook, the majority of your data visualizations created in step C of
the IMPACT model will be created with a declarative purpose.
Declarative visualizations are the product of wanting to “declare” or
present your findings to an audience. The data analysis projects
begin with a question, proceed through analysis, and end with
communicating those findings. This means that while the
visualization may prompt conversation and debate, the information
provided in the charts should be solid. Even if your analysis in the
previous steps of the IMPACT model had been exploratory, by the
time you have arrived to communicate your results, you are
declaring what you have found.
On the other hand, you will sometimes use data visualizations to
satisfy an exploratory visualization purpose. When this is done,
the lines between steps P (perform test plan), A (address and refine
results), and C (communicate results) are not as clearly divided.
Exploratory data visualization will align with performing the test plan
within visualization software—for example, Tableau—and gaining
insights while you are interacting with the data. Often the presenting
of exploratory data will be done in an interactive setting, and the
answers to the questions from step I (identify the questions) won’t
have already been answered before working with the data in the
visualization software.
Exhibit 4-3 is similar to the first four chart types presented to you
in Exhibit 4-2, but Exhibit 4-3 has more detail to help you determine
what to do once you’ve answered the first two questions. Remember
that the quadrant represents two main questions:
EXHIBIT 4-3
The Four Chart Types Quadrant with Detail
Source: S. Berinato, Good Charts: The HBR Guide to Making Smarter, More Persuasive
Data Visualizations (Boston: Harvard Business Review Press, 2016).
1. Are you explaining the results of the previously done analysis, or
are you exploring the data through the visualization? (Is your
purpose declarative or exploratory?) page 145
2. What type of information is being visualized
(conceptual, qualitative information or data-driven, quantitative
information)?
Once you have determined the answers to the first two questions,
you are ready to begin determining which type of visualization will be
the most appropriate for your purpose and dataset.
PROGRESS CHECK
1. What are two ways that complicated concepts were
explained to you via categorization and data visualization
as you were growing up?
2. Using the Internet or other resources (other textbooks, a
newspaper, or a magazine), identify an example of a data
visualization for each possible quadrant.
3. Identify which type of data scale the following variables are
measured on (qualitative nominal, qualitative ordinal, or
quantitative):
a. Instructor evaluations in which students select excellent,
good, average, or poor.
b. Weekly closing price of gold throughout a year.
c. Names of companies listed on the Dow Jones Industrial
Average.
d. Fahrenheit scale for measuring temperature.
CHOOSING THE RIGHT CHART

LO 4-
2
Choose
the best
chart for
your
dataset.
Once you have determined the type of data you’re working with and
the purpose of your data visualization, the next questions have to do
with the design of the visualization—color, font, graphics—and most
importantly, type of chart/graph. The visual should speak for itself as
much as necessary, without needing too much explanation for what’s
being represented. Aim for simplicity over bells and whistles that
“look cool,” but end up being distracting.
Charts Appropriate for Qualitative Data

Because qualitative and quantitative data have such different levels
of complexity and sophistication, there are some charts that are not
appropriate for qualitative data that do work for quantitative data.
When it comes to visually representing qualitative data, the
charts most frequently considered are:
Bar charts.
Pie charts.
Stacked bar chart.
The pie chart is probably the most famous (some would say
infamous) data visualization for qualitative data. It shows the parts of
the whole; in other words, it represents the proportion of each
category as it corresponds to the whole dataset.
Similarly, a bar chart also shows the proportions of each category
as compared to each of the others.
In most cases, a bar chart is more easily interpreted than a pie
chart because our eyes are more skilled at comparing the height of
columns (or the lengths of horizontal bars, depending on the
orientation of your chart) than they are at comparing sizes of pie,
especially if the proportions are relatively similar.
Consider the two different charts from the Sláinte dataset in
Exhibit 4-4. Each compares the proportion of each beer type sold by
the brewery.
page 146
EXHIBIT 4-4
Pie Charts and Column Chart Show Different Ways to Visualize Proportions
The magnitude of the difference between the Imperial Stout and

the IPA is almost impossible to see in the pie chart. This difference is
easier to digest in the bar chart.
Of course, we could improve the pie chart by adding in the
percentages associated with each proportion, but it is much quicker
for us to see the difference in proportions by glancing at the order
and length of the bars in a bar chart (Exhibit 4-5).
EXHIBIT 4-5
Pie Chart Showing Proportion
The same set of data could also be represented in a stacked bar
chart or a 100 percent stacked bar chart (Exhibit 4-6). This chart is
not a default option in Excel, but it does work in another data
visualization tool that we introduce later in this chapter, Tableau. The
first figure in Exhibit 4-6 is a stacked bar chart, which shows the
proportion of each type of beer sold expressed in the number of
beers sold for each product, while the latter shows the proportion
expressed in terms of percentage of the whole in a 100 percent
stacked bar chart.
EXHIBIT 4-6
Example of Stacked Bar Chart
While bar charts and pie charts are among the most common
charts used for qualitative data, there are several other charts that
function well for showing proportions:
Tree maps and heat maps. These are similar types of visualizations,
and they both use size and color to show proportional size of
values. While tree maps show proportions using physical space,
heat maps use color to highlight the scale of the values. However,
both are heavily visual, so they are imperfect for situations where
precision of the numbers or proportions represented is necessary.
Symbol maps. Symbol maps are geographic maps, so they should
be used when expressing qualitative data proportions across
geographic areas such as states or countries.
Word clouds. If you are working with text data instead of categorical
data, you can represent them in a word cloud. Word clouds are
formed by counting the frequency of each word mentioned in a
dataset; the higher the frequency (proportion) of a given word, the
larger and bolder the font will be for that word in the word cloud.
Consider analyzing the results of an open-ended response question
on a survey; a word cloud would be a great way to quickly spot the
most commonly used words to tell if there is a positive or negative
feeling toward what’s being surveyed. There are also settings that
you can put into place when creating the word cloud to leave out the
most commonly used English words—such as the, an, and a—in
order to not skew the data. Exhibit 4-7 is an example of a word
cloud for the text of Chapter 2 from this textbook.
page 147
EXHIBIT 4-7
Word Cloud Example from Chapter 2 Text
Charts Appropriate for Quantitative Data

The data visualization and chart possibilities for charting quantitative
data not only include those available for qualitative data (you can
group and count it), but they have even more sophistication. You can
use pie charts (with the same varying level of success) and bar
charts with quantitative data, but you can also use a lot more.
page 148
There are many different methods for visualizing quantitative

data. With the exception of the word cloud, all of the methods
mentioned in the previous section for qualitative data can work for
depicting quantitative data, but the following charts can depict more
complex data:
Line charts. Show similar information to what a bar chart shows, but
line charts are good for showing data changes or trend lines over
time. Line charts are useful for continuous data, while bar charts are
often used for discrete data. For that reason, line charts are not
recommended for qualitative data, which by nature of being
categorical, can never be continuous.
Box and whisker plots. Useful for when quartiles, median, and
outliers are required for analysis and insights.
Scatter plots. Useful for identifying the correlation between two
variables or for identifying a trend line or line of best fit.
Filled geographic maps. As opposed to symbol maps, a filled
geographic map is used to illustrate data ranges for quantitative
data across different geographic areas such as states or countries.
A summary of the chart types just described appears in Exhibit 4-
8. Each chart option works equally well for exploratory and
declarative data visualizations. The chart types are categorized
based on when they will be best used (e.g., when comparing
qualitative variables, a bar chart is an optimal choice), but this figure
shouldn’t be used to stifle creativity—bar charts can also be used to
show comparisons among quantitative variables, just as many of the
charts in the listed categories can work well with other datatypes and
purposes than their primary categorization below.
EXHIBIT 4-8
Summary of Chart Types
As with selecting and refining your analytical model,

communicating results is more art than science. Once you are
familiar with the tools that are available, your goal should always be
to share critical information with stakeholders in a clear, concise
manner. While visualizations can be incredibly impactful, they can
become a distraction if you’re not careful. For example, bar charts
can be manipulated to show a bias and, while novel, 3D graphs are
incredibly deceptive because they may distort the scale even if the
numbers are fine.
page 149
Tools to Help When Picking a Visual

There are many tools available for data visualization and exploratory
data analysis. Gartner annually assesses a collection of these tools
and creates the “magic quadrant” for business intelligence, depicted
in Exhibit 4-9. The magic quadrant can provide insight into which
tools you should consider using.
EXHIBIT 4-9
Gartner Magic Quadrant for Business Intelligence and Analytics Platforms
Source: Sallam, R. L., C. Howson, C. J. Idoine, T. W. Oestreich, J. L. Richardson, and J.

Tapadinhas, “Magic Quadrant for Business Intelligence and Analytics Platforms,” Gartner
RAS Core Research Notes, Gartner, Stamford, CT (2019).
Based on Gartner’s quadrant, it is easy to see that Tableau and

Microsoft are two of the best and most popular options available.
While other tools are used in industry, such as Qlik and TIBCO, we
focus on Tableau and Microsoft because of their position as leaders
in the data analytics space. The Microsoft tool that Gartner analyzed
and compared with the other products is not just Excel, it includes
the entire Microsoft BI suite, of which Excel is only a part. We will
focus on Excel as the main driver of the Microsoft toolkit in this text.
In previous years, Tableau was ranked slightly higher than
Microsoft on its ability to execute; in 2019, however, Gartner shifted
Microsoft above Tableau in not only completeness of vision, but also
ability to execute. Gartner assesses the tools in this quadrant not
only on their ability to visualize data, but on each tool’s ability to work
with data throughout the entire analytics process—data extraction,
transformation, data analysis and testing, and data visualization.
Microsoft’s tools slightly outperform Tableau in their execution of the
entire analytics process. Tableau is a newer product and has placed
the majority of its focus on data visualization, while Microsoft Excel
has a much more robust platform for data analysis. Excel’s biggest
advantage over Tableau (and over any other data visualization
software in the market) is its ubiquity. Excel has been on the market
longer than any of its competitors, and it is rare to find a business or
university that doesn’t have a version of Excel on every computer. If
your data analysis project is more declarative than exploratory,
it is more likely that you will perform your data visualization to
communicate results in Excel, simply because it is page 150
likely that you performed steps 2 through 4 in Excel,
and it is convenient to create your charts in the same tool that you
performed your analysis.
Tableau earns high praise for being intuitive and easy to use,
which makes it ideal for exploratory data analysis. You may even
find that you would prefer to immediately load your data from Excel
or your relational database (or wherever else your data are stored)
into Tableau during the second step of the IMPACT model and work
on your analysis inside the tool, instead of waiting for step 5 to just
communicate your results through Tableau. If your question isn’t fully
defined or specific, exploring your dataset in Tableau and changing
your visualization type to discover different insights is as much a part
of performing data analysis as crafting your communication. One of
the biggest disadvantages to Tableau is its cost, but fortunately,
Tableau is a tremendous supporter of education, and as a student,
you can download a free academic license to use Tableau on your
PC or Mac. The link to download your free license of Tableau is:
https://www.tableau.com/academic/students. Once you have
downloaded your license, we recommend opening the Superstore
sample workbook provided. You will find it at the bottom of the start
screen under “Sample workbooks” (Exhibit 4-10).
EXHIBIT 4-10
Source: ©Tableau Software, Inc. All rights reserved.
Once you open the workbook, you will see a variety of tabs at the
bottom of the workbook that you can page through and see different
ways that the same dataset can be analyzed and visualized. When
you perform exploratory analysis in Tableau, or even if you have
already performed your analysis and you have uploaded the dataset
into Tableau to communicate insights, we page 151
recommend trying several different types of charts to
see which one makes your insights stand out the most effectively. In
the top right corner of the Tableau workbook, you will see the Show
Me window, which provides different options for visualizing your
dataset (Exhibit 4-11).
EXHIBIT 4-11
In the Show Me tab, only the visualizations that will work for your
particular dataset will appear in full color.
For more information on using Tableau, see Appendix G.
Learning to Create a Good Chart by (Bad)

Example
Other than getting practice by looking at good visualizations and
modifying the way you visualize your dataset in Tableau to see how
different insights are showcased, one of the best ways to learn how
to create a good visualization is to look at some problematic
visualizations.
page 152
In the chart depicted in Exhibit 4-12, the Daily Mail, a UK-based

newspaper, tried to emphasize an upgrade in the estimated growth
of British economy. The estimate from the Office of National
Statistics indicated that Q4 growth would be 0.7 percent instead of
0.6 percent (a relatively small increase of about 15 percent). Yet the
visualization makes it appear as if this is a 200 percent increase
because of the scale the newspaper chose. Another issue is that
some time has passed between the estimates, and we don’t see that
disclosed here (Exhibit 4-12).
EXHIBIT 4-12
Bar Chart Distorting Data Comparison by Using Inappropriate Scale
Source: http://www.dailymail.co.uk/news/article-4248690/Economy-grew-0-7-final-three-
months-2016.html.
If we reworked the data points to show the correct scale (starting
at 0 instead of 0.55) and the change over time (plotting the data
along the horizontal axis), we’d see something like Exhibit 4-13. If we
wanted to emphasize growth, we might choose a chart like Exhibit 4-
14. Notice that both new graphs show an increase that is less
dramatic and confusing.
EXHIBIT 4-13
Bar Chart Using Appropriate Scale for Less Biased Comparison
EXHIBIT 4-14
Alternative Stacked Bar Chart Showing Growth
Our next example of a problematic method of data visualization is
in Exhibit 4-15. The data represented come from a study assessing
cyber-security attacks, and this chart in particular attempted to
describe the number of cyber-security attacks employees fell victim
to, as well as what their role was in their organization.
EXHIBIT 4-15
Difficult to Interpret Pie Chart
Source: http://viz.wtf/post/155727224217/the-authors-explain-furthermore-we-present-the.
Assess the chart provided in Exhibit 4-15. Is a pie chart really the
best way to present these data?
There are simply too many slices of pie, and the key referencing
the job role of each user is unclear. There are a few ways we could
improve this chart.
page 153
If you want to emphasize users, consider a rank-ordered bar

chart like Exhibit 4-16. To emphasize the category, a comparison like
that in Exhibit 4-17 may be helpful. Or to show proportion, maybe a
stacked bar (Exhibit 4-18).
EXHIBIT 4-16
More Clear Rank-Ordered Bar Chart
page 154
EXHIBIT 4-17
Bar Chart Emphasizing Attacks by Job Function
EXHIBIT 4-18
Stacked Bar Chart Emphasizing Proportion of Attacks by Job Function
PROGRESS CHECK
4. The following two charts represent the exact same data—
the quantity of beer sold on each day in the Sláinte Sales
Subset dataset. Which chart is more appropriate for
working with dates, the column chart or the line chart?
Which do you prefer? Why?
a.
Source: Microsoft Excel 2016
page 155
b.
5. The same dataset was consolidated into quarters. This
chart was made with the chart wizard feature in Excel,
which made the creation of it easy, but something went
wrong. Can you identify what went wrong with this chart?

6. The following four charts represent the exact same data
quantity of each beer sold. Which do you prefer, the line
chart or the column chart? Whichever you chose, line or
column, which of the pair do you think is the easiest to
digest?
a.
page 156
b.

c.

d.
FURTHER REFINING YOUR CHART

TO COMMUNICATE BETTER
LO 4-
3
Refine
your
chart to
communi
cate
efficientl
y and
effectivel
y.
After identifying the purpose of your visualization and which type of

visual will be most effective in communicating your results, you will
need to further refine your chart to pick the right data scale, color,
and format.
page 157
Data Scale and Increments
As tools such as Excel and Tableau become more intuitive and more
powerful, considering your data scale and increments is less of a
concern because both tools will generally come up with scales and
increments that make sense for your dataset. With that being said,
there are still four main questions to consider when creating your
data scale and increments:
1. How much data do you need to share in the visual to avoid being
misleading, yet also avoid being distracting? (For example, do you
need to display the past four years, or will the past two quarters
suffice?) When you consider leaving out some data, is it to show
only the insights that are meaningful, or is it an attempt to skew the
data or to hide poor performance? Be careful to not hide data that
are meaningful just because they don’t align with your expectations.
2. If your data contain outliers, should they be displayed, or will they
distort your scale to the extent that you can leave them out? If the
purpose of your chart is to call attention to the outliers, then they
need to remain (and you need to ensure that they are not errors, but
this should have been done in step 2 of the IMPACT model when
you mastered the data). If the purpose of your chart is to display the
middle pack of the data, the outliers may not be relevant to the
insights, and they could be left out.
3. Other than determining how much data you need to share, what
scale should you place those data on? Typically, charts should
begin with a baseline of 0, but if 0 is meaningless to your dataset,
you could find a different baseline that makes sense. Be careful to
not overexaggerate the height or the baseline so that your trendline
or bar chart is over- or underemphasized; your trendline should take
up two-thirds of the chart. Once you decide on a data scale, the
increments for your data scale should be “natural” such as 1s, 2s,
5s, 100s, etc. (e.g., not 3s or 0.02s).
4. Do you need to provide context or reference points to make the
scale meaningful? For example, if you were provided with a stock
price of $100, would you immediately be able to tell if that is a high
number or a low number? Not necessarily; without context of the
company’s stock price over time, the company’s industry and its
competitors’ stock prices, or some other piece of context, certain
numbers are not altogether useful.
Color
Similar to how Excel and Tableau have become stronger tools at
picking appropriate data scales and increments, both Excel and
Tableau will have default color themes when you begin creating your
data visualizations. You may choose to customize the theme.
However, if you do, here are a few points to consider:
When should you use multiple colors? Using multiple colors to
differentiate types of data is effective. Using a different color to
highlight a focal point is also effective. However, don’t use multiple
colors to represent the same type of data. Be careful to not use
color to make the chart look pretty—the point of the visualization is
to showcase insights from your data, not to make art.
We are trained to understand the differences among red, yellow,
and green, with red meaning something negative that we would
want to “stop” and green being something positive that we would
want to “continue,” just like with traffic lights. For that reason, use
red and green only for those purposes. Using red to show
something positive or green to show something negative is
counterintuitive and will make your chart harder to understand. You
may also want to consider a color-blind audience. If you are
concerned that someone reading your visuals may be color blind,
avoid a red/green scale and consider using page 158
orange/blue. Tableau has begun defaulting to
orange/blue color scales instead of red/green for this reason.
Once your chart has been created, convert it to grayscale to ensure
that the contrast still exists—this is both to ensure your color-blind
audience can interpret your visuals and also to ensure that the
contrast, in general, is stark enough with the color palette you have
chosen.
PROGRESS CHECK
7. Often, external consultants will use a firm’s color scheme
for a data visualization or will use a firm’s logo for points on
a scatter plot. While this might be a great approach to
support a corporate culture, it is often not the most effective
way to create a chart. Why would these methods harm a
chart’s effectiveness?
COMMUNICATION: MORE THAN

VISUALS—USING WORDS TO
PROVIDE INSIGHTS
LO 4-
4
Commun
icate
your
results in
a written
report.
As a student, the majority of the writing you do is for your professors.

You likely write emails to your professors, which should carry a
respectful tone, or essays for your Comp 1 or literature professors,
where you may have been encouraged to use descriptive language
and an elevated tone; you might even have had the opportunity to
write a business brief or report for your business professors. All the
while, though, you were still aware that you were writing for a
professor. When you enter the professional world, your writing will
need to take on a different tone. If you are accustomed to writing
with an academic tone, transitioning to writing for your colleagues in
a business setting requires some practice. As Justin Zobel says in
Writing for Computer Science, “good style for science is ultimately,
nothing more than writing that is easy to understand. [It should be]
clear, unambiguous, correct, interesting, and direct.”2 As an author
team, we have tremendous respect for literature and the different
styles of writing to be found, but for communicating your results of a
data analysis project, you need to write directly to your audience,
with only the necessary points included, and as little descriptive style
as possible. The point is, get to the point.
Content and Organization

Each step of the IMPACT model should be communicated in your
write-up, as noted here:
I: Explain what was being researched. Even if your audience is

the people who requested the research, you should still restate
the purpose of the project. Include any relevant history as well. If
your project is part of a larger program or if it’s a continued effort
to explain an issue or help a decision come to fruition, then
include the background.
M: Depending on your audience, you may not cover too much of
what your process was in the “master the data” step of the
IMPACT model, but an overview of the data source and which
pieces of data are included in the analysis should be present. If
your audience is technical and interested, you may go into detail
on your ETL process, but it is more likely that you will leave out
that piece.
P and A: Similar to how you write about mastering the data, you
may not need to include a thorough description of your test plan
or your process for refining your results depending on what your
audience is interested in and what they need to page 159
know, but including an overview of the type of
analysis performed and any limitations that you encountered will
be important to include.
C: If you are including a data visualization with your write-up, you
need to explain how to use the visual. If there are certain aspects
that you expect to stand out from the analysis and the
accompanying visual, you should describe what those
components are—the visual should speak for itself, but the write-
up can provide confirmation that the important pieces are
gleaned.
T: Discuss what’s next in your analysis. Will the visual or the
report result in a weekly or quarterly report? What trends or
outliers should be paid attention to over time?
Audience and Tone

Carefully considering your audience is critical to ensuring your
communication is effective. If you have three messages to write—
one letting your mom know that you are coming home this weekend
and you’ll need to do laundry, one to your professor letting her know
that you will miss class on Friday, and one to your best friend asking
if he wants to join you for Chipotle—efficiency would suggest that
you type it all into one email and click send. That would definitely be
the quickest way to get out the message. But is it a good idea?
Certainly not. Your mom does not need to know that you’re not going
to class on Friday, and you probably don’t want your professor to
show up at Chipotle to have lunch with you and your friend. Instead
of sending the same message to all three people, you tailor the
delivery—that is, you consider the audience. You include all of the
information that they need to know and nothing else.
You should do the same thing when crafting your communication
regarding your data analysis. If you have several different people to
communicate results to, you may consider crafting several different
versions: one that contains all of the extraction, transformation, and
loading (ETL) details for the programmers and database
administrators, one that is light on ETL but heavy on interpretation of
the visual and results for your managers, and so on. Consider the
knowledge and skill of your audience—don’t talk down to them, but
don’t overwhelm a nontechnical crowd with technical jargon. Explain
the basics when you should, and don’t when you shouldn’t.
An additional piece of communication to consider is the vehicle
for communication. We have a myriad of options available to us for
communicating: email, phone calls, Skype, instant messaging,
printed reports, even face-to-face conversations that can be either
informal or formal presentations in meetings. When crafting your
communication, consider the best way to provide the information to
your intended audience.
Is the concept difficult to understand? A written report will
probably not suffice; plan to supplement your written material with a
sit-down conversation or a phone call to explain the details and
answer questions.
Is the topic an answer to a question and fairly simple to
understand? An emailed response summarizing the visualization and
results will likely suffice.
How does the person you are sending the report to
communicate? Consider the professional culture of your
organization. It may be commonplace to communicate casually using
abbreviations and jargon in your workplace, but if that’s not the way
your workplace operates or even if it’s not the way that the recipient
of your message communicates, take the time to refine your
message and mirror the norms of the organization and the
recipients.
Is the report going to be updated and sent out at regular intervals
(daily, weekly, monthly)? If so, keep a consistent template so that it is
easy for the recipients to identify the information they seek on a
regular basis.
There are, of course, many more concepts to consider that will be
unique to each message that you craft. Take the time to always
consider your audience, their communication style, page 160
and what they need from the communication—and
provide it, via the right message, the right tone, and the right vehicle.
Revising
Just as you addressed and refined your results in the fourth step of
the IMPACT model, you should refine your writing. Until you get
plenty of practice (and even once you consider yourself an expert),
you should ask other people to read through your writing to make
sure that you are communicating clearly. Justin Zobel suggests that
revising your writing requires you to “be egoless—ready to dislike
anything you have previously written. . . . If someone dislikes
something you have written, remember that it is the readers you
need to please, not yourself.”3 Always placing your audience as the
focus of your writing will help you maintain an appropriate tone,
provide the right content, and avoid too much detail.
PROGRESS CHECK
Progress Checks 5 and 6 display different charts depicting the quantity of beer sold
on each day in the Sláinte Sales Subset dataset. If you had created those visuals,
starting with the data request form and the ETL process all the way through data
analysis, how would you tailor the written report for the following two roles?
8. For the CEO of the brewery who is interested in how well
the different products are performing.
9. For the programmers who will be in charge of creating a
report that contains the same information that needs to be
sent to the CEO on a monthly basis.
Summary
This chapter focused on the fifth step of the IMPACT
model, or the “C,” on how to communicate the results of
your data analysis projects. Communication can be
done through a variety of data visualizations and written
reports, depending on your audience and the data you
are exhibiting.
In order to select the right chart, you must first
determine the purpose of your data visualization. This
can be done by answering two key questions:
Are you explaining the results of a previously done

analysis, or are you exploring the data through the
visualization? (Is your purpose declarative or
exploratory?)
What type of data are being visualized (conceptual
qualitative data or data-driven quantitative data)?
The differences between each type of data (declarative

and exploratory, qualitative and quantitative) are
explained, as well as how each datatype affects both the
tool you’re likely to use (generally either Excel or
Tableau) and the chart you should create.
After selecting the right chart based on your purpose
and datatype, your chart will need to be further refined.
Selecting the appropriate data scale, scale increments,
and color for your visualization is explained through the
answers to the following questions:
How much data do you need to share in the visual to

avoid being misleading, yet also avoid being
distracting?
page 161
If your data contain outliers, should they
be displayed, or will they distort your scale to the
extent that you can leave them out?
Other than how much data you need to share, what
scale should you place those data on?
Do you need to provide context or reference points to
make the scale meaningful?
When should you use multiple colors?
Finally, this chapter discusses how to provide a written

report to describe your data analysis project. Each step
of the IMPACT model should be communicated in your
write-up, and the report should be tailored to the specific
audience to whom it is being delivered.
Key Words
continuous data (143) One way to categorize
quantitative data, as opposed to discrete data.
Continuous data can take on any value within a range. An
example of continuous data is height.
declarative visualizations (144) Made when the aim

of your project is to “declare” or present your findings to
an audience. Charts that are declarative are typically
made after the data analysis has been completed and are
meant to exhibit what was found in the analysis steps.
discrete data (143) One way to categorize quantitative

data, as opposed to continuous data. Discrete data are
represented by whole numbers. An example of discrete
data is points in a basketball game.
exploratory visualizations (144) Made when the lines

between steps P (perform test plan), A (address and
refine results), and C (communicate results) are not as
clearly divided as they are in a declarative visualization
project. Often when you are exploring the data with
visualizations, you are performing the test plan directly in
visualization software such as Tableau instead of creating
the chart after the analysis has been done.
interval data (143) The third most sophisticated type of

data on the scale of nominal, ordinal, interval, and ratio; a
type of quantitative data. Interval data can be counted and
grouped like qualitative data, and the differences between
each data point are meaningful. However, interval data do
not have a meaningful 0. In interval data, 0 does not
mean “the absence of” but is simply another number. An
example of interval data is the Fahrenheit scale of
temperature measurement.
nominal data (142) The least sophisticated type of

data on the scale of nominal, ordinal, interval, and ratio; a
type of qualitative data. The only thing you can do with
nominal data is count, group, and take a proportion.
Examples of nominal data are hair color, gender, and
ethnic groups.
normal distribution (143) A type of distribution in

which the median, mean, and mode are all equal, so half
of all the observations fall below the mean and the other
half fall above the mean. This phenomenon is naturally
occurring in many datasets in our world, such as SAT
scores and heights and weights of newborn babies. When
datasets follow a normal distribution, they can be
standardized and compared for easier analysis.
ordinal data (142) The second most sophisticated type

of data on the scale of nominal, ordinal, interval, and ratio;
a type of qualitative data. Ordinal can be counted and
categorized like nominal data and the categories can also
be ranked. Examples of ordinal data include gold, silver,
and bronze medals.
proportion (142) The primary statistic used with

qualitative data. Proportion is calculated by counting the
number of items in a particular category, then dividing that
number by the total number of observations.
qualitative data (142) Categorical data. All you can do

with these data are count and group, and in some cases,
you can rank the data. Qualitative data can be further
defined in two ways: nominal data and ordinal data. There
are not as many options for charting qualitative data
because they are not as sophisticated as quantitative
data.
quantitative data (143) More complex than qualitative

data. Quantitative data can be further defined in two
ways: interval and ratio. In all quantitative data, the
intervals between data points are meaningful, allowing the
data to be not just counted, grouped, and ranked, but also
to have more complex operations performed on them
such as mean, median, and standard deviation.
ratio data (143) The most sophisticated type of data on

the scale of nominal, ordinal, interval, and ratio; a type of
quantitative data. They can be counted and grouped just
like qualitative data, and the differences page 162
between each data point are meaningful like
with interval data. Additionally, ratio data have a
meaningful 0. In other words, once a dataset approaches
0, 0 means “the absence of.” An example of ratio data is
currency.
standard normal distribution (144) A special case of

the normal distribution used for standardizing data. The
standard normal distribution has 0 for its mean (and thus,
for its mode and median, as well), and 1 for its standard
deviation.
standardization (144) The method used for comparing

two datasets that follow the normal distribution. By using a
formula, every normal distribution can be transformed into
the standard normal distribution. If you standardize both
datasets, you can place both distributions on the same
chart and more swiftly generate your insights.
ANSWERS TO PROGRESS
CHECKS
1. Certainly, answers will vary given our own individual
experiences. But we can note that complex topics can
be explained and understood by linking them to
categorizations or pictures, such as the food pyramid.
2. Answers will vary.
3.
a. Qualitative ordinal
b. Quantitative (ratio data)
c. Qualitative nominal
d. Quantitative (interval data)
4. While this question does ask for your preference, it is
likely that you prefer image b because time series data
are continuous and can be well represented with a line
chart instead of bars.
5. Notice that the quarters are out of order (1, 2, then 4);
this looks like quarter 3 has been skipped, but quarter 4
is actually the last quarter of 2019 instead of the last
quarter of 2020, while quarters 1 and 2 are in 2020.
Excel defaulted to simply ordering the quarters
numerically instead of recognizing the order of the
years in the underlying data. You want to be careful to
avoid this sort of issue by paying careful attention to the
charts, ordering, and scales that are automatically
created through Excel (and other tools) wizards.
6. Answers will vary. Possible answers include the
following: Quantity of beer sold is a discrete value, so it
is likely better modeled with a bar chart than a line
chart. Between the two line charts, the second one is
easier to interpret because it is in order of highest sales
to lowest. Between the two bar charts, it depends on
what is important to convey to your audience—are the
numbers critical? If so, the second chart is better. Is it
most important to simply show which beers are
performing better than others? If so, the first chart is
better. There is no reason to provide more data than
necessary because they will just clutter up the visual.
7. Color in a chart should be used purposefully; it is
possible that a firm’s color scheme may be
counterproductive to interpreting the chart. The icons as
points in a scatter plot might be distracting, which could
make it take longer for a reader to gain insights from
the chart.
following: Explain to the CEO how to read the visual,
call out the important insights in the chart, tell the range
of data that is included (is it one quarter, one year, all
time?).
following: Explain the ETL process, exactly what data
are extracted to create the visual, which tool the data
were loaded into, and how the data were analyzed.
Explain the mechanics of the visual. The particular
insights of this visual are not pertinent to the
programmer because the insights will potentially
change over time. The mechanics of creating the report
are most important.
page 163

1. Gold, silver, and bronze medals would be examples of:
a. nominal data.
b. ordinal data.
c. structured data.
d. test data.
2. In the late 1960s, Ed Altman developed a model to
predict if a company was at severe risk of going
bankrupt. He called his statistic Altman’s Z-score, now a
widely used score in finance. Based on the name of the
statistic, which statistical distribution would you guess
this came from?
a. Normal distribution
b. Poisson distribution
c. Standardized normal distribution
d. Uniform distribution
3. Justin Zobel suggests that revising your writing requires
you to “be egoless—ready to dislike anything you have
previously written,” suggesting that it is _____ you need
to please:
a. yourself
b. the reader
c. the customer
d. your boss
4. Which of the following is not a typical example of
nominal data?
a. Gender
b. SAT scores
c. Hair color
d. Ethnic group
5. The Fahrenheit scale of temperature measurement
would best be described as an example of:
a. interval data.
b. discrete data.
c. nominal data.
d. continuous data.
6. _____ data would be considered the least sophisticated
type of data.
a. Ratio
b. Interval
c. Ordinal
d. Nominal
7. _____ data would be considered the most sophisticated
type of data.
a. Ratio
b. Interval
c. Ordinal
d. Nominal
8. Line charts are not recommended for what type of
data?
a. Normalized data
b. Qualitative data
page 164
c. Continuous data
d. Trend lines
9. Exhibit 4-8 gives chart suggestions for what data you’d
like to portray. Those options include all of the following
except:
a. relationship between variables
b. geographic data
c. outlier detection
d. normal distribution curves
10. What is the most appropriate chart when showing a
relationship between two variables (according to Exhibit
4-8)?
a. Scatter chart
b. Bar chart
c. Pie graph
d. Histogram
1. Explain Exhibit 4-2 and why these four dimensions are
helpful in describing information to be communicated.
Exhibit 4-2 lists conceptual and data-driven as being on
two ends of the continuum. Does that make sense, or
can you think of a better way to organize and
differentiate the different chart types?
2. According to Exhibit 4-8, which is the best chart for
showing a distribution of a single variable, like height?
How about hair color? Major in college?
3. Box and whisker plots (or box plots) are particularly
adept at showing extreme observations and outliers. In
what situations would it be important to communicate
these data to a reader? Any particular accounts on the
balance sheet or income statement?
4. Based on the data from datavizcatalogue.com, a line
graph is best at showing comparisons, relationships,
compositions, or distributions? Name the best two.
5. Based on the data from datavizcatalogue.com, what are
some major flaws of using word clouds to communicate
the frequency of words in a document?
6. Based on the data from datavizcatalogue.com, how
does a box and whisker plot show if the data are
symmetrical?
7. What would be the best chart to use to illustrate
earnings per share for one company over the past five
years?
8. The text mentions, “If your data analysis project is more
declarative than exploratory, it is more likely that you
will perform your data visualization to communicate
results in Excel.” In your opinion, why is this true?
9. According to the text and your own experience, why is
Tableau ideal for exploratory data analysis?
Problems
1. Why was the heat map associated with the opening
vignette regarding the 1854 cholera epidemic effective?
Now that we have more sophisticated tools and
methods for visualizing data, what else could have
been used to communicate this, and would it have been
more or less effective in your opinion?
2. Evaluate the use of color in the graphic associated with
the opening vignette regarding drug overdose deaths
across America. Would you consider its use effective or
ineffective? Why? How is this more or less effective
than communicating the same data in a bar chart?
page 165
3. According to Exhibit 4-8, which is the best
chart for comparisons of earnings per share over many
periods? How about for only a few periods?
4. According to Exhibit 4-8, which is the best chart for
static composition of a data item of the Accounts
Receivable balance at the end of the year? Which is
best for showing a change in composition of Accounts
Receivable over two or more periods?
5. The Big 4 accounting firms (Deloitte, EY, KPMG, and
PwC) dominate the audit and tax market in the United
States. What chart would you use to show which
accounting firm dominates in each state in terms of
audit revenues? Any there other interesting ways you
could use to find opportunities within the audit market?
6. Datavizcatalogue.com lists seven types of maps in its
listing of charts. Which one would you use to assess
geographic customer concentration by number? How
could you show if some customers buy more than other
customers on such a map? Would you use the same
chart or a different one?
7. In your opinion, is the primary reason that analysts use
inappropriate scales for their charts due to an error
related to naiveté (or ineffective training), or are the
inappropriate scales used so the analyst can sway the
audience one way or the other?
page 166
Lab 4-1 Use PivotCharts to

Visualize Declarative Data
This lab builds upon the steps completed in Lab 2-2
concerning the Sláinte brewery and the PivotTable report
you prepared showing the total number of each item sold
each month between January and April 2020.
When working with a data analysis project that is
declarative in nature, the analysis will likely be done in
Excel, and the data visualization will be done after the
analysis has been completed as a means to communicate
results.
Company summary
through big change. Sláinte sells six different products.
distributing from one state to distributing to nine states,
and now the business has begun stabilizing after the
expansion. With that stability comes a need for better
analysis. One of Sláinte’s first priorities is to identify its
areas of success, as well as areas of potential
improvement.
Data
Sláinte dataset
Technique
Some experience with spreadsheets and PivotTables is
useful for this lab.
Software needed
Excel
+ Shift + 4)
Parts 1–4 of the IMPACT Model

These steps were performed in Lab 2-2. You can either
use the already prepared data in the file
Lab_4_1_SlaintePivot.xlsx, or you can use the file that
you saved after completing Lab 2-2, which should have
been saved to your computer as Slainte_Pivot.xlsx.
Part 5: Communicate Your Findings

We demonstrate two alternate ways of communicating
findings. Please work through both alternatives.
Alternative 1: Create a PivotChart

1. Ensuring that the active cell in your workbook is
somewhere in the PivotTable, navigate to the Analyze
tab in the ribbon.
2. If you are working with a PC, you should see a button
for PivotChart. If you are working with a Mac, you can
insert a regular chart from the Insert tab on the ribbon.
Some of the functionality of this lab will be limited on a
Mac, so you may opt to use a virtual PC lab
environment to complete this lab.
3. Once you click into PivotChart, a window for Insert
Chart appears, along with a list of options for how you
can visualize your PivotTable. It defaults to column
chart. Another good option is the bar chart, which
displays your data in horizontal bars instead of
columns. Create either a bar chart or a column chart by
selecting the chart you prefer and clicking OK.
4. Take a screenshot that shows the PivotTable and the
PivotChart (label it 4-1A).
page 167
5. The advantage of working with PivotCharts
over regular charts is that you can slice and filter your
data in the PivotTable and the PivotChart at the same
time. If you are presenting your findings to a live
audience and you anticipate questions about specific
months or specific products, using Excel’s slicer tool is
a great way to filter your data in a way that is interactive
and transparent. On the Analyze tab on the ribbon, you
can select Insert Slicer.
6. In the window that pops up, select Product
Description. This will create an interactive filter so that
you can drill down into different product descriptions as
they perform over the months.
7. Create a second slicer for Sales_Order_Month and
click OK.
8. Take a screenshot that includes your PivotChart,
PivotTable, and both slicers (label it 4-1B).
Q1. Spend a few minutes filtering the data with the
slicers. Name three important insights that
were easy to identify through this
visualization.
Q2. What does the data visualization and the
interactivity of the slicer provide your
audience that the original PivotTable does
not?
Alternative 2: Visualize the PivotTable with
Conditional Formatting and Sparklines
Conditional formatting and sparklines are quick ways to
visualize and compare data and trends when a full-
fledged chart isn’t necessary.
9. To quickly visualize how each product’s total quantity
sold over time compares across all six of Sláinte’s
projects, we can apply conditional formatting to the
Grand Total column. Select the data in the Grand
Total column of your PivotTable, and navigate to the
Home tab on the ribbon.
LAB EXHIBIT 4-1A

10. From the Home tab, select the Conditional

Formatting button, and a menu with the different types
of formatting available will appear.
11. Select Data Bars and pick the first option for blue
gradient fill bars.
12. This conditional formatting is helpful because it allows
us to compare grand totals of each product. However, if
we would like to see how each product’s month-over-
month sales compare to one another, we can display
mini line charts next to each row with a sparkline. To do
so, select all of the “meat” of your PivotTable—that is,
don’t select any of the product labels (such as Imperial
IPA), month labels, or grand totals.
13. Navigate to the Insert tab on the ribbon, and select
Line in the Sparklines category.
14. A window will appear specifying the data range you just
selected and awaiting input for the Location Range.
We’d like to see the trend lines to the immediate right of
our PivotTable, so you can select the cells in the first
empty column after your Grand Totals.
page 168
LAB EXHIBIT 4-1B

LAB EXHIBIT 4-1C

15. Click OK, and your sparklines will be created.

16. You will notice that there are gaps in the lines, though. If
you’d rather see a continuous line to represent zero
values for the blank cells, you can change this option.
Ensure that one of the cells with the sparkline in it is
active, and navigate to the Sparkline Tools tab on the
ribbon.
17. Click the bottom half of the Edit Data button to make a
menu appear.
18. From the menu, select Hidden & Empty Cells. . .
19. Select the option to show empty cells as zero, and
click OK.
20. Take a screenshot to show the conditional formatting
and the sparklines (label it 4-1C).
Q3. When do you think a sparkline and/or
conditional formatting would be preferable
over creating a PivotChart?
Q4. What other visualizations would be useful to
interpret these data? If you were to create a
report to be run monthly, what are two
visualizations that should be included?
Q5. Provide a written report discussing the data
analysis project and the insights that should
be gained from this visualization.
End of Lab
Lab 4-2 Use Tableau to Perform

Exploratory Analysis and Create
Dashboards
When working with a data analysis project that is
exploratory in nature, the analysis can be done in
Tableau. You will likely enter the data analysis project with
an overarching question in mind, but as you answer that
question, your exploratory analysis will lead to ongoing
questions. The data visualization will help explore the
data, as well as ultimately be used as a means to
communicate results.
page 169
Company summary
through big change. Sláinte sells six different products.
distributing from one state to distributing to nine states,
and now the business has begun stabilizing after the
expansion. With that stability comes a need for better
analysis. One of Sláinte’s first priorities is to identify its
areas of success, as well as areas of potential
improvement.
Data
Sláinte dataset
Software needed
Tableau. Visit with your instructor for instructions or follow
this link to download Tableau,
https://www.tableau.com/academic/students, and click
Get Tableau for Free to register for a free student
license. Your student license will last one year.
+ Shift + 4)
Part 1: Identify appropriate questions.

Part 2: Complete the ETL process to load the data in
Tableau for analysis.
Part 3: Analyze the data you receive with data
visualization.
Part 4: Communicate the data you receive with a
digital dashboard.

If you completed Lab 2-1 or 2-2, you became familiar with
the Sláinte dataset and identified questions regarding
Sláinte.
In particular, we worked with this scenario: Sláinte has
brought you in to help determine potential areas for sales
growth in the next year. Additionally, the company has
noticed that its margins aren’t as high as it had budgeted
and would like you to help identify some areas where it
could improve its pricing, marketing, or strategy.
Specifically, Sláinte would like to know how many of each
product was sold.
We’ll start with the same question—identifying the
amount of each product sold, overall.
The Sláinte data include the following tables and
fields, presented in a UML diagram:
LAB EXHIBIT 4-2A

Q1. Using the UML diagram, identify which
table(s) and attributes you will need to answer
your initial question regarding amount of
products sold.

To complete the ETL process, we will need to extract the
data from Access and transform and load it into Tableau.
page 170
1. Open Tableau.
2. Select Microsoft Access from the Connect To a File
options.
3. Browse to the Slainte_Subset.accdb file and click

Open. This will extract the data.
4. The Data Source tab will open, with three tables for
you to select from. We can begin by just exploring the
Sales data. Double-click on the Sales_Subset table to
load it into Tableau.
page 171
5. The data should load into Tableau without
any problems. However, it is always a good idea to
check the datatypes that each attribute loaded in as.
Notice the Abc above Sales Order ID, and the
calendar icon above Sales Order Date. These indicate
that Sales Order ID data imported as text, while the
Sales Order Date imported as calendar data. The
number signs above Sales Order Quantity and
Product Sale Price indicate that those attributes were
imported as numerical data. This is all set up exactly as
we’d like, so there’s no need to transform the data.
Q2. If the Sales Order Date datatype had imported

as number, how might that cause a problem
with our analysis if we wanted to dig into the
data by month, for example?
Q3. Why did your Sales Order ID attribute import
as text when it looks like each field has
numerical data in it? Would there be any
benefit in Sales Order ID being stored as a
number? Why will it not present a problem in
our analysis to maintain these data as text?
Part 3: Perform Exploratory Analysis

Click into Sheet 1 in the bottom left of the Tableau tool to
begin working with the data. The Tableau screen can be
compared to the way Excel’s PivotTable Fields list is laid
out. The attributes from the Sales_Subset table are
categorized into dimensions and measures.
Dimensions are descriptive attributes—these are the
fields that we typically slice or group our data by in a
PivotTable.
Measures are numerical—these are the fields that you
would typically drag into the VALUES area in the
PivotTable to calculate a count, sum, or average of your
data.
6. To view the number of products sold, begin by double-
clicking on the measure Sales Order Quantity Sold.
Notice that Tableau doesn’t default to showing you one
number, but instead displays one bar of a bar chart. This
is a clear indication of how Tableau treats data differently
than Excel. Excel defaults to numerical data, while
Tableau defaults to visualization.
7. To group the total amount of products sold by the
products themselves, double-click on the dimension
Product Code.
The visualization you just created summarizes the
answer to our initial question (how many of each product
has sold), but this visual can be improved.
Q4. Identify two ways to improve this visual to
make it more easily understandable.
9. Sort the bars: Across the top of the Tableau screen are
a variety of icons. Toward the middle of that menu are
two icons for sorting data. Click the icon to sort your
data descending.
page 172
10. Add labels to the bars: To the left of your
data viz, there is the Marks card. It has a variety of
ways that you can enhance the way you’re viewing the
data. Click Label, then place a check mark in the box
next to Show mark labels.
11. Instead of showing Product Code, show the Product

Description; this will require you to join in another
table. Click back into the Data Source tab in the bottom
left.
12. Double-click on the FGI_Product table to load the
product data into Tableau. You will see the
FGI_Product data populate, as well as a Venn diagram
joining the two datasets. Click on the Venn diagram to
ensure the data are joined properly. You want to ensure
that the primary key of FGI_Product is matched with
the corresponding foreign key in the Sales_Subset
data (the same way the two tables are joined in the
UML diagram).
13. Return to Sheet 1 to work with the new data.

14. Double-click on the dimension Product Description to
add this detail to your data visualization.
15. Now that you have added the description to the
visualization, you can remove the Product Code
dimension. Remove Product Code from the data
visualization by dragging and dropping the Product
Code pill out of the Columns shelf. After removing the
Product Code pill, you will need to sort your data again
by using the same sort descending icon that you
clicked in step 9.
page 173
17. Sometimes when you’re performing exploratory data
analysis, you’ll want to save the visualization you just
made, while also giving yourself the opportunity to drill
down into the data. We’ll name this sheet after the
analysis you just did, then duplicate the data to work
with it further. Right-click Sheet 1 and select Rename
Sheet. Type Total Products Sold as the sheet’s name.
18. Right-click the sheet tab that you just renamed and
select Duplicate Sheet.
page 174
19. Let’s dig into how these products have
performed year over year. Drag and drop the dimension
Sales Order Date to the Columns shelf, and place it to
the left of Product Description.
image
Notice that the pill doesn’t just say the name of the
attribute, but it says YEAR and it has a button to expand
the pill.
If you expand the Sales Order Date function once, it
will split the data among quarters. If you expand again, it
will further drill down into months.
21. Rename this sheet Total Products Sold by Year.
22. Navigate to the Data Source tab and add in the
Customer table.
image
23. Ensure that the join is built on the appropriate primary

key/foreign key relationship between Sales_Subset
and Customer according to the UML diagram.
24. For each of the previous tables that we loaded into
Tableau, Sales_Subset and FGI_Product, we didn’t
need to transform the data. This time, we will want to
work with geographic data. Looking at the datatypes for
each attribute that loaded in, you can see that
Customer City and Customer Zip have globe icons for
their datatypes, indicating that Tableau was able to
intuit that these attributes are geographic. But the
Customer St attribute has an Abc next to it, indicating
that Tableau loaded it as only text (not geographic
data). Click on the Abc icon above Customer St to
change its datatype.
image
page 175
25. Select Geographic Role, and then select
State/Province.
image
26. Create a new sheet (do not duplicate any of the

previous sheets) by clicking the first icon to the right of
the Total Products Sold by Year tab.
image
27. This time, we will create a report that shows total

products sold by state. Double-click the dimension
Customer St. Tableau automatically populates a map
with a dot in each state that’s listed in the Customer
table.
28. Double-click on the measure Sales Order Quantity
Sold. The dots have changed to vary in size, which is
proportional to the amount of sales in each state.
29. We can make the results easier to interpret by changing
the visualization type. If the Show Me window isn’t
showing in the upper right corner, click Show Me, then
select the Filled Map.
image
page 176
30. Rename this sheet Total Products Sold
by State.
31. Take a screenshot (label it 4-2D).
Part 4: Communicate Results

Now that you have created three simple, but meaningful
data visualizations, you can create a dashboard to
communicate the results. Tableau makes it easy to place
all of these visualizations on one interactive pane.
32. Select the icon for New Dashboard, which is to the
right of the New Worksheet icon.
image
33. In the Dashboard view, instead of seeing the various

dimensions and measures to drag and drop, you see
the three sheets that you have created. You can drag
and drop them into the area that says Drop Sheets
Here, and you can arrange them any way you wish.
Replicate this arrangement:
image
34. You can also use each sheet as a filter. Click the Total
Products Sold section of your dashboard. There are
three small icons in the top right of the sheet when the
sheet is active. Clicking the middle one (which looks
like a funnel) will allow you to use the bars as filters for
the entire dashboard. Click to do so.
page 177
image
35. Follow the same process to make the states work as

filters for the dashboard by clicking Use as Filter in the
Total Products Sold by State sheet.
Now, you can click any of the bars in the Total
Products Sold chart or any of the states in the Total
Products Sold by State, and the data in each of the
three sheets will shift to focus on just those products
and/or states.
36. Filter by either a state or a product, and take a
screenshot (label it 4-2E).
Q5. After creating these sheets and the dashboard,
what additional data would you recommend that
Sláinte analyze? What is another data
visualization that would be helpful for Sláinte’s
decision making?
End of Lab

Dillard’s Store Data: Create
Geographic Data Visualizations in
Tableau and in Power BI
Company summary
Data
Software needed
Microsoft SQL Server Management Studio and Microsoft
Excel (available on the Remote Desktop at the University
of Arkansas)
Tableau (available on the Remote Desktop at the

Learn how to prepare data visualization in Tableau.
page 178

Question 2 of Lab 3-4 was as follows: Do customers in
the state with the highest transaction balances have a
significantly higher transaction balance during the period
September 1, 2016, to September 15, 2016, than all other
states?
In this lab, we will work to visualize these transaction
data in a way that helps users grasp the information
needed to make decisions.
Q1. How would this information, average
transaction balance by state, help a manager
make decisions?
Q2. How would you think managers would like to
visualize transaction balance by state? What
would be the most (and less) effective ways to
visualize these transactions?

Load the data into Tableau.
Arkansas. Open a new Tableau workbook and connect
to Microsoft SQL Server.
image
page 179
2. Input the Server and Database
information that you received from your instructor for
the Dillard’s data, and then click Sign In.
3. Wait for the connection to process, and then you have

two options: If you are certain that you will only want to
visualize one specific set of query results, you can input
a query from the Connections page. Alternatively, you
can connect to entire tables if you want the option to
drill down into the data and answer more than one
question.
Inputting a Custom Query into the Tableau
Connections Page
4. Double-click New Custom SQL.
image
page 180
5. Input your SQL query into the Edit
Custom SQL window, and then click OK.
SELECT state, avg(tran_amt) AS Average
FROM transact
INNER JOIN store
ON transact.store = store.store
GROUP BY state
image
6. It may take a couple minutes for the results to populate.

Once they do, we’ll preview the data.
The data should load without a problem, but because
Tableau is automatically interpreting the data, it is a good
idea to look through the data to ensure that we don’t need
to transform them in any way. In Tableau, you should
always check which datatype has been assigned to each
attribute. The datatype is denoted by a little icon that is an
Abc for a string of text, a number sign for numerical
data, a calendar for dates, or a globe for geographic
data.
The two attributes of state are denoted with an Abc
and a number sign:
image
7. Of particular concern is the way the state data were

imported. The Abc above the state column indicates
that they were imported into Tableau as plain text
instead of as a geographic attribute.
image
page 181
8. For us to view these data more
meaningfully, we’ll want to change the type of data that
state is designated as. Click the Abc, then Geographic
Role, and select State/Province.

9. Once Tableau has processed the change, click Sheet 1
on the bottom of the Tableau window to begin working
with the data.
page 182
10. Double-click on state in Dimensions.
You will see that Tableau immediately populates a map

with a blue dot in each state that has a Dillard’s store.
11. To make these data even more meaningful, we’ll add
average to this view. Double-click Average in
Measures.
12. Tableau might have defaulted to a symbol map. The

difference in averages is easier to interpret with a filled
map. Click Show Me in the top right corner of Tableau if
your Show Me window isn’t already available, then click
Filled Map.
End of this process
Joining Tables into the Tableau Connections Page

This option will produce the same visualization that the
steps above just created, but it will also provide more
flexibility for digging into the data because more data will
be loaded into Tableau.
Open up a new instance of Tableau and repeat steps 1

through 3 to connect to the Dillard’s data.
13. Instead of creating a new customer query, drag the
Transact table to the Drag tables here portion of the
Tableau window.
14. Drag the Store table to the Drag tables here portion of
the Tableau window.
15. Tableau will likely default to joining the tables on the
appropriate attributes, but double-check that it did by
clicking the visual representation of the join (it looks like
a Venn diagram).
page 183
16. The join should indicate that it is an inner
join based on the transact.store and store.store
attributes. If it says something different, modify the join.
17. Check that the attributes pulled in as the appropriate

datatypes. For example, City and Zip Code pulled in as
geographic datatypes, but state did not. Click the Abc
above the State attribute to change the datatype.
18. Click Geographic Role, then State/Province.
19. Click Sheet 1 in the bottom left corner of the Tableau

screen to begin working with the data.
page 184
20. Double-click State from Dimensions.
image
Tableau immediately populates a map with a blue dot in
each state that has a Dillard’s store.
21. To make these data even more meaningful, we’ll add
average transaction amount to this view. Start by
double-clicking on Tran Amt from the Measures.
22. It may take a couple minutes for Tableau to populate

the data, but the size of the blue dots will adjust to show
how the amounts vary across states. The default value
for this measure is SUM, though, so we need to edit it
to be average.
23. Hover over SUM(Tran Amt) in the Marks window to
make available an arrow for a drop-down window.
page 185
24. Click the drop-down, then click Measure
(Sum) to change the measure to Average.
image
25. Tableau might have defaulted to a symbol map. The

difference in averages is easier to interpret with a filled
map. Click Show Me in the top right corner of Tableau if
your Show Me window isn’t already available, then click
Filled Map.
26. Take a screenshot of your results (label it 4-3A).
End of this process
page 186

Data
Visualizing data often makes it easier to see the answers
to your questions, which then leads to more questions. In
this case, Arkansas clearly has a higher average
transaction amount than the other states. This may lead
you to want to drill down into the data to see if the
performance is the same across all of the stores in
Arkansas, or if there is a stand-out store.
27. If you click Arkansas, Tableau will give you the option
to filter out all of the other states so that you can drill
down into this data point. Click Keep Only.
28. From the dimensions, double-click City.
29. Tableau doesn’t recognize city lines, so it will change
from a filled map to a symbol map. This may be easier
to read as a bar chart, though, so click the Horizontal
Bars icon in the Show Me window.
Q3. Which city has the highest average

transaction amount? (It can be easier to
answer this question if you sort the data.
Clicking the “sort” button will re-order the bars
so that the city with the highest average
transaction amount will be the first bar listed.)
Q4. How would you think managers would like to

see transaction balance by state?
Q5. What are further questions that would be
meaningful to drill down into with this same
dataset, given what you have seen so far?
page 187
To dig deeper into the data, we can drill down into

which types of items are being sold the most in Maumelle.
To do so, we need to join in two more tables. Joining in
the SKU table will provide description of the items being
sold, and joining in the DEPARTMENT table will provide
categorical information for each individual item.
30. Click Data Source in the bottom left corner of the
Tableau application.
31. Join in the SKU and DEPARTMENT tables.

32. Return to your Tableau sheet with the horizontal bar

chart, and click Keep Only for Maumelle.
The DEPARTMENT and SKU data are hierarchical,
with an item belonging to a department, which groups into
a deptdec (decade) through a deptcent (century).
33. Begin by viewing the Maumelle store data by the
highest level of the hierarchy, the department century.
The description attribute will be the most useful to
interpret, so double-click on the Deptcent Desc
attribute from the DEPARTMENT dimensions.
34. To drill down further into the data, add the department
decade data to the chart. Double-click on Deptdec
Desc to add another level of detail.
35. You can also add drill-down capabilities by creating the
hierarchy in Tableau. Drag and drop Deptdec Desc on
top of Deptcent Cent in the Dimensions window:
page 188
36. Click OK on the window to create the hierarchy.
37. Notice that the Deptcent Desc pill in the Rows shelf
changed to include a minus sign—this indicates that the
hierarchy has been expanded. Click the minus sign to
collapse the hierarchy.
38. Take a screenshot of your results (label it 4-3B).

With this much data loaded into Tableau, there is a
tremendous amount of analysis and visualization that you
can do.
Q6. Based on what you have seen of the average
transaction amounts for different departments
and products in the Maumelle store, what
would you recommend to the Maumelle store
manager who is trying to maximize profits?
Advertise certain products more? Advertise
certain products less? Open an additional
store nearby? Close this store, etc.?
We can re-create the process that you completed in
steps 13–38 in Power BI to get a feel for how that tool
interacts with data differently than Tableau. You may wish
to visit Appendix I for a general introduction to Power BI if
you have not used the tool before. Keep in mind that
Power BI has regular updates, so some screens and
requirements may change from what’s described in this
text.
Two of the most significant differences from how we
worked with this data in Tableau and how we will work
with it in Power BI are the following:
Sheet view versus Dashboard view:
Tableau defaults to creating your visualizations one

sheet at a time and displays them on a large scale
while you are developing them. Creating a Dashboard
is an additional step, and you can drag and drop the
sheets that you have already created onto a
dashboard.
Power BI defaults to a dashboard view. Your
visualizations are created in smaller tiles that you can
immediately re-size and re-arrange around the
canvas.
Because of this difference, instead of manipulating
the same sheet throughout this analysis, we will
create copies of tiles and manipulate the copies as
we work through the steps in Power BI.
page 189
Filtering:
Tableau offers two main methods for filtering: dragging

a measure or dimension into the Filters shelf to create
a Filter and using the “Keep Only” button on a portion
of a visualization.
Power BI does not have a “Keep Only” function, but it
defaults to providing filters for every dimension or
measures that you are interacting with throughout the
dashboard.
Because of this difference, we will use the Filter
fields instead of using “Keep Only” as we drill
down through our data in Power BI.
1. Open the Power BI Desktop application.
2. Click Get Data to connect to the database.
3. Select SQL Server database, then click Connect.
The next steps are similar to how you connect to the

database using Excel’s PowerQuery tools.
4. Enter the server name (provided to you by your
instructor and the database name (WCOB_DILLARDS)
in the fields for Server and Database.
5. Select DirectQuery for the Data Connectivity mode.
This option is not available in the PowerQuery tool in
Excel, but it is similar to the “Live” connection option in
Tableau. Both Power BI’s DirectQuery and Tableau’s
Live options allow you to connect to the data without
importing all of the rows into the report. This often helps
you start working with the data faster because you do
not have to wait for large datasets to load.
page 190
image
6. Select the Department, SKU, Store, and Transact

tables, then click Load.
image
It may take some time for the data to load. Once it

does, you will see the tables populated in the Fields list.
7. Expand the Store table to see its attributes and select
State.
page 191
image
Power BI will create a symbol map showing each state

that is listed in the Store table.
8. To compare the average Transaction Amount across
each state, expand the Transact table and select
Tran_Amt.
image
page 192
As the data loads, we can take a moment to look at

the visualizations pane. The State field that you placed a
checkmark in appears in the Location portion of the
visualizations pane, and the Tran_Amt field appears in
the Size portion. Both fields also appear in the Filters
section of the visualization pane. By immediately creating
filters for each field that you have selected for your visual,
you can easily slice and drill-down your data. We’ll work
with these filters throughout the rest of this lab.
If you don’t see any fields in Location, Size, or Filters,
you just need to click on the visual tile that you have
created first. If you click outside of the tile, you won’t see
the fields listed because you can add tiles to the report.
image
page 193
9. Similar to what Tableau does with Measures and what

Excel PivotTables does with Values, Power BI typically
will default to summing data if you add numerical data
to a visualization. For this report, we would rather work
with average transaction amounts instead of the sum.
To change it, click the drop-down on the Tran_Amt
field in the Size section of the Visualizations Pane and
select Average.
image
Your tile should now show a visualization with a map

that includes symbols that vary in size based on the
average transaction amount across states.
Take a screenshot and label it 4-3C.
10. To drill down into the average transaction amount in the
cities of Arkansas, we will start by duplicating the
existing visual. Right click the map visual and select
Copy (you can also select the visual and use the
keyboard shortcut Ctrl + C to copy).
image
page 194
11. Use the keyboard shortcut Ctrl + V to
paste the visual (or Command + V on a Mac), then drag
the copy beneath the original visual.
12. Change the visual type to a Bar Chart.
image
13. Add a filter for Arkansas by clicking the drop-down next

to State in the Filters field and selecting AR.
image
14. The single bar representing Arkansas in this chart isn’t

very meaningful – from here we want to drill down to
see the cities within Arkansas that have the highest
average transaction amounts. Place a checkmark next
to the City field (in the Store table) to add City to the
Bar Chart to see which city in Arkansas has the highest
returns.
page 195
image
Just like it did in Tableau, the resulting bar chart shows

that Maumelle had the highest average transaction
amount across all Arkansas cities.
You might want to re-size the bar chart to make it
easier to read.
Take a screenshot and label it 4-3D.
We’ll dig into which products are sold most frequently
in Maumelle next to learn more about why their average
transaction amount is so high.
15. To work with just Maumelle’s data, first copy and paste
the Arkansas City Bar Chart. Place it to the right of the
Arkansas City bar chart.
16. First, narrow this visual down to just view Maumelle.
Place a check-mark in the box next to Maumelle in the
city filter.
17. Next, we need to create a product description hierarchy
in Power BI similar to how we did in Tableau. Expand
the Department table field list and drag and drop
DEPTDEC_DESC on top of DEPTCENT_DESC.
image
page 196
18. Place a check mark next to the
DEPTCENT_DESC Hierarchy to add it to your new tile
to add details to the Maumelle bar chart.
19. To view the Department Centuries, right-click the bar
and select Drill Down.
image
20. To view Department Decades, repeat the same step by

right-clicking on any of the Century bars in the chart to
drill down again. This will show the decades beneath
the specific century that you selected.
21. You can expand all centuries at once by clicking the
Drill Down button in the top right of the tile.
image
Take a screenshot and label it 4-3E.

Q7. After working through these steps in Tableau
and in Power BI, what do you think are the
pros and cons of creating data visualizations
and exploring data each of the tools?

Dillard’s Store Data: Visualizing
Regression in Tableau
Company summary
Data
You will use an Excel file with data transformed and ready
for regression analysis, Lab 4-4 Dummy.xlsx.
page 197
Software needed
Tableau

Learn how to visualize regressions in Tableau.

In Chapter 3, you ran a variety of regression and other
analyses addressing the following questions:
Do customers in the state with the largest total
transaction amount have a significantly higher
transaction balance during September 2016 than all
other states?
Are online transaction amounts statistically greater than
or lesser than non-online transactions during the period
September 1, 2016, to September 15, 2016?
Do customers who charge their purchases to a Dillard’s
credit card spend more on each transaction during the
time period September 1, 2016, to September 15, 2016?
In this lab, we will work to visualize these data in a
way that helps users grasp the information needed to
make decisions.

To complete the ETL process, we will need to extract the
data from the Excel spreadsheet that you saved in the
Chapter 3 comprehensive labs and transform and load it
into Tableau.
1. Open a new Tableau workbook and connect to
Microsoft Excel.
image
page 198
2. Browse to Lab 4-4 Dummy.xlsx (or you can
use the Excel output you created with the dummy
variables from Lab 3-5) and click Open. This will extract
the data.
3. When running a regression in Tableau, you will want to
place your explanatory variables on the columns and
your dependent variables on the rows. To do so, drag
and drop the Arkansas-dummy measure to the
Columns shelf and the Tran Amt Measure to the
Rows shelf.
image
4. Tableau defaults to aggregating the measures, but we

are interested in each individual observation. To
disaggregate the variables, navigate to the Analysis
tab and click Aggregate Measures to disaggregate the
values.
image
5. It may take some time for the data to disaggregate.

Once they do, navigate back to the Analysis tab and
click Lines, and then select Show All Trend Lines.
image

Data
6. Hover over the trend line to see the regression formula
and the p-value.
7. Each variable can be meaningful in explaining the total
spent on each transaction (or Tran Amt), but when
working with data visualization, it can be even more
meaningful to compare models. Compare models by
adding the Online-dummy and DLRD- page 199
dummy variables (where DLRD
represents the use of a Dillard’s credit card in the
transaction) to the columns. Note these are univariate
analyses in that they compare only Arkansas, online,
and the use of a Dillard’s credit card one-by-one and
not altogether.
image

Q1. Which of these three variables has a
noticeable trend as compared to the others,
suggesting greater explanatory power?
Q2. Which of these three variables best explains
the average transaction amount? (Hint:
Consider the r-squared in each or the p-
values among the three models.)
Q3. The coefficient on the DLRD-dummy is
negative here. What does that suggest? Is
that consistent with the results of Lab 3-5?
Q4. In the trend line looking at the Arkansas-
dummy, what is the base level of Transaction
Amount (the y-intercept) before considering
the Arkansas-dummy?
Q5. Let’s suppose we could capture the net worth
of each Dillard’s customer. Would you expect
net worth to have higher explanatory variables
than either of the transaction took place in
Arkansas, was an online purchase, or was
paid for using a Dillard’s credit card? Why or
why not?
1S. Berinato, Good Charts: The HBR Guide to Making Smarter, More Persuasive Data
Visualizations (Boston: Harvard Business Review Press, 2016).
2Source: Zobel, Justin. Writing for Computer Science (Singapore: Springer-Verlag, 1997).
3Source: Justin Zobel
page 200
Chapter 5
The Modern Accounting
Environment

Most of the focus of Data Analytics in accounting is focused on
auditing, managerial accounting, financial statement analysis, and
tax. This is partly due to the demand for high-quality data and the
need for enhancing trust in the assurance process, informing
management for decisions, and aiding investors as they select their
portfolios. In this chapter, we look at how both auditors and
managers are using technology in general to improve the decisions
being made. We also introduce how Data Analytics helps facilitate
continuous auditing and reporting.
A Look Back
Chapter 4 completed our discussion of the IMPACT model by
explaining how to communicate your results through data
visualization and through written reports. We discussed how to
choose the best chart for your dataset and your purpose. We also
helped you learn how to refine your chart so that it communicates as
efficiently and effectively as possible. The chapter wrapped up by
describing how to provide a written report tailored to specific
audiences who will be interested in the results of your data analysis
project.
A Look Ahead
In Chapter 6, you will learn how to use audit software to perform
substantive audit tests, including when and how to select samples
and how to confirm account balances. Specifically, we discuss the
use of different types of descriptive, diagnostic, predictive, and
prescriptive analytics as they are used to generate computer-
assisted auditing techniques.
page 201
fizkes/Shutterstock
The large public accounting firms offer a variety of analytical tools to their customers.
Take PwC’s Halo, for example. This tool allows auditors to interrogate a client’s data
and identify patterns and relationships within the data in a user-friendly dashboard.
By mapping the data, auditors and managers can identify inefficiencies in business
processes, discover areas of risk exposure, and correct data quality issues by drilling
down into the individual users, dates and times, and amounts of the entries. Tools
like Halo allow auditors to develop their audit plan by narrowing their focus and audit
scope to unusual and infrequent issues that represent high audit risk.
Source: http://halo.pwc.com
OBJECTIVES
LO 5-1 Understand how automation has created a data-rich

environment.
LO 5-2 Understand different approaches to organizing
enterprise data and common data models.
LO 5-3 Describe the appropriate tasks and approaches to
automating procedures.
LO 5-4 Evaluate continuous monitoring techniques and
alarms.
LO 5-5 Understand cloud-based collaboration platforms.
page 202
THE MODERN DATA ENVIRONMENT

LO 5-
1
Understa
nd how
automati
on has
created
a data-
rich
environ
ment.
As businesses have embraced automation over the past several

decades, more information about financial transactions is captured in
large databases. In addition to details of transaction data, these
systems capture metadata (e.g., timestamps, user details, and
contents of unstructured data) that provide insight into the workings
of the company. Sensors provide details on movement through a
building to identify optimal location of resources, track the health of
employees to help control health insurance costs, and allow detailed
analysis of the volume of everyday events to help managers manage
robotic scripts and tasks. Even traditional data-entry tasks are now
performed by specialized software that recognizes text from
documents and maps it into database fields. Once you have an
understanding of the IMPACT model, understanding the needs and
structure of the business will help your perform meaningful analyses.
Automation can include routine tasks, such as combining data
from different sources for analysis, and more complex actions, such
as responding to natural language queries. In the past, analytics and
automation were performed by hobbyists and consultants within a
firm. In a modern environment, companies form centers of expertise
where they concentrate specialists in a single geographic location
and use information and communication technologies to work in
remote teams. Because the data are network-accessible, multiple
users interact with the data and complete workflows of tasks with the
assistance of remote team members and bots, or automated robotic
scripts commonly called robotics process automation. The
specialists manage the bots like they would normal employees,
continuously evaluating their performance and contribution to the
company.
You’ll recall from your auditing course that assurance services
are crucial to building and maintaining trust within the capital
markets. In response to increasing regulation in the United States,
the European Union, and other jurisdictions, both internal and
external auditors have been tasked with providing enhanced
assurance while also attempting to reduce (or at least maintain) the
audit fees. This has spurred demand for more audit automation
along with an increased reliance on auditors to use their judgment
and decision-making skills to effectively interpret and support their
audit findings with managers, shareholders, and other stakeholders.
Auditors have been applying simple Data Analytics for decades in
evaluating risk within companies. Think about how an evaluation of
inventory turnover can spur a discussion on inventory obsolescence
or how working capital ratios are used to identify significant issues
with a firm’s liquidity. From an internal audit perspective, evaluating
cost variances can help identify operational inefficiencies or
unfavorable contracts with suppliers.
The audit concepts of professional skepticism and reasonable
assurance are as much a part of the modern audit as in the past.
There has been a shift, however, of simply providing reasonable
assurance on the processes to the additional assurance of the
robots that are performing a lot of the menial audit work. Where,
before, an auditor may have looked at samples and gathered
evidence to make inferences to the population, now that same
auditor must understand the controls and parameters that have been
programmed into the robot. In other words, as these automated bots
do more of the routine analytics, auditors will be free to exercise
more judgment to interpret the alarms and data while refocusing their
effort on testing the parameters used by the robots.
Auditors use Data Analytics to improve audit quality by more
accurately assessing risk and selecting better substantive
procedures and tests of controls. While the exercises the auditors
conduct are fairly routine, the models can be complex and require
auditor judgment and interpretation. For example, if an auditor
receives 1,000 notifications of a control violation during the day, does
that mean there is a control weakness or that the settings on the
automated control are too precise? Are all those notifications actual
control violations that require immediate attention, or are most of
them false positives—transactions that are flagged as exceptions but
are normal and acceptable?
page 203
The auditors’ role is to make sure that the appropriate analytics

are used and that the output of those analytics—whether a
dashboard, notifications of exceptions, or accuracy of predictive
models—correspond to management’s expectations and assertions.
The Increasing Importance of the Internal

Audit
If you look at the assurance market, there are many trends that are
affecting the profession. First, the major applications of Data
Analytics in auditing are not solely focused on the financial
statements as evaluated by public accounting firms. Rather, these
tend to focus on data quality, internal controls, and the complex
information systems that support the business process—areas
typically reserved for the internal audit department at a firm. Second,
the risk and advisory practices of the public accounting firms are
experiencing greater growth, in large part due to firms’ outsourcing
or co-sourcing of the internal audit function. Third, external auditors
are permitted to rely on the work of internal auditors to provide
support for their opinion of financial statements.
For these reasons, most of the innovations in Data Analytics
have originated in internal audit departments, where there is
constant pressure to enhance business value while minimizing costs.
In the recent past, many companies’ experience with Data Analytics
in the internal audit department have come from internal auditors
who have investigated Data Analytics on their own. These
individuals then find a champion with management and are
encouraged to continue their work. Under the guidance of the chief
audit executive (CAE) or another manager, these individuals build
teams to develop and implement analytical techniques to aid the
following audits:
1. Process efficiency and effectiveness.
2. Governance, risk, and compliance, including internal controls
effectiveness.
3. Information technology and information systems audits.
4. Forensic audits in the case of fraud.
5. Support for the financial statement audit.
Internal auditors are also more likely to have working knowledge of
the different types of systems implemented at their companies. They
are familiar with how the general journals from a product like
Oracle’s JD Edwards actually reconcile to the general ledger in SAP
to generate financial reports and drill down into the data. Because
implementation of these systems varies across organizations (and
even within organizations), internal auditors can understand how
analytics are not simply a one-size-fits-all type of strategy.
PROGRES CHECK
1. What types of sensors do businesses use to track activity?
2. Make the case for why an internal audit is increasingly
important in the modern audit. Why is it also important for
external auditors and the scope of their work?
ENTERPRISE DATA
LO 5-
2
Understa
nd
different
approac
hes to
organizin
g
enterpris
e data
and
common
data
models.
While organizations have become more data-centric as they have

adopted enterprise systems (ES) over the past few decades, these
systems can vary greatly among organizations. Some companies will
take a homogeneous systems approach for their data structure by
ensuring that all of its divisions and subsidiaries use a uniform
installation of a common ES system, such as SAP. Homogenous
systems enable management to consolidate the information from
various locations and roll them up management reports, audit
support, and financial statements with minimal additional effort.
Other companies that grow through acquisition, take a page 204
heterogeneous systems approach, where they
attempt to integrate the existing systems (such as SAP, Oracle,
PeopleSoft, JD Edwards, and others) of the companies they acquire
and use a series of translators to convert the output of those
systems into usable financial information. Systems translator
software attempts to map the various tables and fields from these
varied enterprise systems into a data warehouse, where all of the
data can be analyzed centrally, as shown in Exhibit 5-1. The data
warehouse is updated periodically, typically on a daily basis, to
reflect recent firm activity.
EXHIBIT 5-1
Homogeneous Systems, Heterogeneous Systems, and Software Translators
One of the primary obstacles that managers and auditors face is
access to appropriate data. As noted in Chapter 2, managers and
auditors may request flat files or extracts from an IT manager.
Frequently, these files may be incomplete, unrelated, limited in
scope, or delayed when they are not considered a priority by IT
managers. Increasingly, managers and auditors request read-only
access to the data warehouse so they can evaluate transaction data,
such as purchases and sales, and the related master data, such as
employees and vendors, in a timely manner. By avoiding a data
broker, they get more relevant data for their analyses and analyze
multiple relationships and explore other patterns in a more
meaningful way. In either case, the managers and auditors work with
duplicated data, rather than querying the production or live
systems directly.
Common Data Models

As automation of data analytics procedures becomes more common,
working within different data environments can present challenges,
especially for auditors. To minimize some of the effort required to
interact with data, analysts adopt a common data model, which is a
tool used to map existing database tables and fields from various
systems to a standardized set of tables and fields for use with
analytics. When the underlying systems change, the model is
updated to pull data from the new tables and fields. Similar to the
translation software mentioned previously, a common data model
makes it easier to perform and automate routine analytical
procedures even when the underlying systems change.
page 205
The AICPA’s Audit Data Standards (ADS) provide one example

of a common data model with tables and fields that are needed by
auditors to perform routine audit tasks. The AICPA recommends that
ES vendors standardize the output of data that auditors are likely to
use. The goal of the standards is to reduce the data loading and
transformation effort required by the auditor, so they can focus on
the analytics more quickly as well as define real-time or continuous
analytics via the data warehouse. These standards are voluntary,
and actual implementation is currently limited, but they provide a
good basis for data needed to audit specific company functions.
The current set of audit data standards defines the following
standards:
Base: defines the formats for files and fields as well as master data
requirements for users, business units, segments, and tax tables.
General Ledger: defines the chart of accounts, source listings, trial
balance, and general ledger or journal entry detail.
Order to Cash Subledger: defines sales orders and line items,
shipments and line items, invoices and line items, open accounts
receivable and adjustments, cash receipts, and customer master
data, shown in Exhibit 5-2.
Procure to Pay Subledger: defines purchase orders and line items,
goods received and line items, invoices received and line items,
open accounts payable and adjustments, payments, and supplier
master data.
page 206
Inventory Subledger: defines inventory location master
data, product master data, inventory on hand data, and inventory
movement transactions as well as physical inventory and material
cost.
Fixed Asset Subledger: defines fixed asset master data, additions,
removal, and depreciation calculations.
EXHIBIT 5-2
Audit Data Standards
The audit data standards define common elements needed to audit the order-to-cash or
sales process.
Source:
https://www.aicpa.org/InterestAreas/FRC/AssuranceAdvisoryServices/DownloadableDocum
ents/AuditDataStandards/AuditDataStandards.O2C.July2015.pdf
*if receivable balances are tracked by customer only (not by invoice), then
Customer_Account_ID is used as a key to join tables to the Open_Accounts_Receivable
table instead of both Customer_Account_ID and Invoice_ID
**The Uer_Listing table can be joined to three fields, all of which contain a user ID—
Entered_by, Approved_By, Last_Modified_By
With standard data elements in place, not only will internal
auditors streamline their access to data, but they also will be able to
build analytical tools that they can share with others within their
company or professional organizations. This can foster greater
collaboration among auditors and increased use of Data Analytics
across organizations. These data elements will be useful when
performing substantive testing in Chapter 6.
Even if the standard is never adopted by data suppliers, auditors
can still take advantage of the audit data standards as a common
data model. For example, Exhibit 5-3 shows the mapping of a set of
Purchase Card data to the Procure to Pay Subledger Standard.
Once the mapping algorithm has been generating using SQL or
other tool, any new data can be analyzed quickly and easily.
EXHIBIT 5-3
Mapping Purchase Card Data to the Procure to Pay Subledger Audit Data Standard
PROGRESS CHECK
3. What are the advantages of the use of homogeneous
systems? Would a merger target be more attractive if it
used a similar financial reporting system as the potential
parent company?
4. How does the use of audit data standards facilitate data
transfer between auditors and companies? How does it
save time for both parties?
AUTOMATING DATA ANALYTICS

LO 5-
3
Describe
the
appropri
ate tasks
and
approac
hes to
automati
ng
procedur
es.
Most of the effort in Data Analytics is preparing the analysis for the
first time. This involves identifying the data, mapping the tables and
fields through ETL, and developing the visualization if needed. Once
that task is complete, automation of the procedure involves
identifying the timing or schedule of how often the procedure should
run, any parameters that might change, and what should happen if a
new observation appears as an outlier.
page 207
The steps you follow to perform the analysis are part of the
algorithm, and they can be recorded using a scripting language,
such as Python or R, or using off-the-shelf monitoring software. That
process is outside the scope of this textbook, but there are many
resources online to help you with this next step.
The main impact of automation and Data Analytics on the
accounting profession comes through optimization of the
management dashboard and audit plan. When beginning an
engagement—whether to audit the financial statements, certify the
enterprise system, or make a recommendation to improve a
business process—auditors generally follow a standardized audit
plan. The benefit of a standardized audit plan is that newer members
of the audit team can jump into an audit and contribute. Audit plans
also identify the priorities of the audit.
An audit plan consists of the one or more of the following
elements:
A methodology that directs that audit work.
The scope of the audit, defining the time period, level of materiality,
accounts and subsidiaries being audited and expected completion
time for the audit.
Potential risk within the area being audited.
Procedures and specific tasks that the audit team will execute to
collect and analyze evidence. These typically include tests of
controls and substantive tests of transaction details.
Formal evaluation by the auditor and supervisors.
Because audit plans are formalized and standardized, they lend
themselves to the use of Data Analytics and, consequently,
automation. For example,
The methodology may be framed by specific standards, such as the
Public Company Accounting Oversight Board’s (PCAOB) auditing
standards, the Committee of Sponsoring Organizations’s (COSO)
Enterprise Risk Management framework, or the Information
Systems Audit and Control Association’s (ISACA) Control
Objectives for Information and Related Technologies (COBIT)
framework. Data Analytics may be used to analyze the standards
and determine which requirements apply to the organization being
audited.
The scope of the audit defines parameters that will be used to filter
the records or transactions being evaluated.
Simple to complex Data Analytics can be applied to a client’s data
during the planning stage of the audit to identify which areas the
auditor should focus on. This may include outlier detection or other
substantive tests of suspicious or risky transactions.
Audit procedures themselves typically identify data, locations, and
attributes that the auditors will evaluate. These are the variables
that will provide the input for many of the substantive analytical
procedures discussed in Chapter 6.
The evaluation of audit data may be distilled into a risk score. This
may be a function of the volume of exceptional records or level of
exposure for the functional area. If the judgment and decision
making can be easily defined, a rule-based analytic could
automatically assign a score for the auditor to review. For more
complex judgments, the increasing prevalence of artificial
intelligence and machine learning discussed in Chapter 3 may be of
assistance. Historical observations of the scores auditors assign to
specific cases and outcomes, may assist the creation of an
automated scoring model.
Typically, internal audit organizations that have adopted Data
Analytics to enhance their audit have done so when an individual on
the team has begun tinkering with Data Analytics. They convince
their managers that there is value in using the data to direct the audit
and the manager may become a champion in the process. Once
they show the value proposition of Data Analytics, they are given
more resources to build the program and adapt the existing audit
program to include more data-centric evaluation where appropriate.
Because of the potential disruption to the organization, it is more
likely that an internal auditor will adapt an existing audit plan than
develop a new system from scratch. Automating the page 208
audit plan and incorporating data analytics involve
the following steps, which are similar to the IMPACT model:
1. Identify the questions or requirements in the existing audit plan.
2. Master the data by identifying attributes and elements that are
automatable.
3. Perform the test plan, in this case by developing analytics (in the
form of rules or models) for those attributes identified in step 2.
4. Address and refine results. List expected exceptions to these
analytics and expected remedial action by the auditor, if any.
5. Communicate insight by testing the rules and comparing the output
of the analytics to manual audit procedures.
6. Track outcomes by following up on alarms and refining the models
as needed.
Let’s assume that an internal auditor has been tasked with
implementing data analytics to automate the evaluation of a
segregation of duties control within SAP. The auditor evaluates the
audit plan and identifies a procedure for testing this control. The
audit plan identifies which tables and fields contain relevant data,
such as an authorization matrix, and the specific roles or
permissions that would be incompatible. The auditor would use that
information to build a model that would search for users with
incompatible roles and notify the auditors.
CONTINUOUS MONITORING
TECHNIQUES
LO 5-
4
Evaluate
continuo
us
monitori
ng
techniqu
es and
alarms.
Data Analytics and automation allow management and internal

auditors to continuously monitor and audit the systems and
processes within their companies. Whereas a traditional audit may
have the internal auditors perform a routine audit plan once every 12
to 36 months or so, the continuous audit evaluates data in a form
that matches the pulse of the business. For example, purchase
orders can be monitored for unauthorized activity in real time, while
month-end adjusting entries would be evaluated once a month.
When exceptions occur—for example, a purchase order is created
with a customer whose address matches an employee’s—the
auditors are alerted immediately and given the option to respond
right away to resolve the issue.
Continuous auditing is a process that provides real-time
assurance over business processes and systems. It involves the
application of rules or analytics that perform a continuous
monitoring function that constantly evaluates internal controls and
transactions and is the chief responsibility of management. It also
generates continuous reporting on the status of the system so that
an auditor can know at any given time whether the system is
operating within the parameters set by management or not. In the
future, continuous reporting may also enable firms to publish real-
time financial accounting data for public analysis, but in practical use
this may cause more problems for firms if they are unable to validate
or provide assurance on the data being reported.
Implementing continuous auditing procedures is similar to
automating an audit plan with the additional step of scheduling the
automated procedures to match the timing and frequency of the data
being evaluated and notifying the auditor when exceptions occur.
Alarms and Exceptions

Whenever an automated or continuous auditing rule is violated, an
exception occurs. The record is flagged and systems generate an
exception report that typically identifies the record and the date of
the exception.
Alarms are essentially a classification problem. A data value is
sent through a simple decision tree based on a series of rules and
classified as a positive event (alarm) or a negative page 209
event (no alarm). Remember we talked about
accuracy of models in Chapter 3: These alarms will not always be
correct.
Once the notification of the alarm or exception arrives, auditors
follow a set of procedures to resolve the issue. First, they must
determine whether the alarm represents a true positive, a transaction
that is problematic, such as an error or fraud, or a false positive,
where a normal transaction is classified as problematic. When too
many alarms are false positive, auditors face information overload,
where there are too many incorrect alarms that distract them from
adequately evaluating the system. Because auditors are mostly
concerned with true positives, they should attempt to train or refine
the models to minimize the potential flood of alarms that occurs
when too many alarms are false positives. This is summarized in
Table 5-1.
TABLE 5-1
Four Types of Alarms That an Auditor Must Evaluate
Normal Event Abnormal Event
Alarm False positive True positive
No Alarm True negative False negative
WORKING PAPERS AND AUDIT

WORKFLOW
LO 5-
5
Understa
nd
cloud-
based
collabora
tion
platform
s.
As audit procedures become increasingly technical, documentation

continues to be an essential way for internal and external auditors to
increase their reliance on automated controls and procedures. The
idea of a black-box audit is no longer sufficient; rather, auditors must
have a better understanding of the tools they use and the output of
those tools. This is where working papers come into play.
Working papers are essential to audit planning, performance, and
evaluation. They provide the documentation for the procedures the
auditors follow, evidence they collect, and communication with the
audit client. As they relate to Data Analytics, working papers should
contain the following items:
Work programs used to document the audit procedures to collect,
manipulate, model, and evaluate data.
IT-related documentation, including flowchart and process maps
that provide system understanding.
Database maps (such as UML diagrams) and data dictionaries that
define the location and types of data auditors will analyze.
Documentation about existing automated controls, including
parameters and variables used for analysis.
Evidence, including data extracts, transformed data, and model
output, that provides support for the functioning controls and
management assertions.
Policies and procedures that help provide consistent quality work
are essential to maintaining a complete and consistent audit. The
audit firm or chief audit executive is responsible for providing
guidance and standardization so that different auditors and audit
teams produce clear results. These standardizations include
consistent use of symbols or tick marks and a uniform mechanism
for cross-referencing output to source documents or data.
Electronic Working Papers and Remote Audit
Work
As audit teams embrace a variety of information and communication
technologies to enable collaboration from different locations, audit
firms have done so, as well. Increasingly, internal and external audit
teams consist of more specialized onsite auditors who interact with a
team of experts and data scientists remotely at page 210
locations around the world. Many of the routine tasks
are offloaded to the remote or seasonal workers, freeing up onsite
auditors to use more professional judgment and expertise during the
engagement. This results in cost savings for the firm through
increased efficiency at the firm level.The glue that holds the audit
team together is the electronic workpaper platform as well as other
collaboration tools, such as Microsoft Teams or Slack. The electronic
workpaper platforms, such as TeamMate or Xero, automate the
workflow of evidence collection, evaluation, and opinion generation
on the part of the audit teams. The large accounting firms have
proprietary systems that accomplish a similar purpose. For example,
PwC uses three systems to automate its audit process. Aura is used
to direct the audit by identifying which evidence to collect and
analyze, Halo performs Data Analytics on the collected evidence,
and Connect provides the workflow process that allows managers
and partners to review and sign off on the work. Most of these
platforms are hosted in the cloud, so members of the audit team can
participate in the various functions from any location. Smaller audit
shops can build ad hoc workpaper repositories using OneDrive with
Office 365, though there are fewer controls over the documents.
PROGRESS CHECK
5. Continuous audit uses alarms to identify exceptions that
might indicate an audit issue and require additional
investigation. If there are too many alarms and exceptions
based on the parameters of the continuous audit system,
will continuous auditing actually help or hurt the overall
audit effectiveness?
6. PwC uses three systems to automate its audit process.
Aura is used to direct the audit by identifying which
evidence to collect and analyze, Halo performs Data
Analytics on the collected evidence, and Connect provides
the workflow process that allows managers and partners to
review and sign off on the work. How does that line up with
the steps of the IMPACT model we’ve discussed throughout
the text?
Summary
As accounting has evolved over the past few decades,
automation has driven many of the changes in turn
enabling additional Data Analytics. Enterprise data
appears in many forms and the adoption of a common
data model makes it easier to analyze data from a variety
of systems with ease. Data Analytics has improved
management’s and auditors’ ability to understand their
business, assess risk, inform their opinions, and improve
assurance over the processes and controls in their
organizations.
Key Words
audit data standards (ADS) (205) A set of standards
developed by the AICPA that defines common tables and
fields that are needed by auditors to perform common
audit tasks.
common data model (204) A tool used to map existing
database tables and fields from various systems to a
standardized set of tables and fields for use with
analytics.
page 211
continuous auditing (208) A process that provides real-

time assurance over business processes and systems.
continuous monitoring (208) A process that constantly

evaluates internal controls and transactions and is the
chief responsibility of management.
continuous reporting (208) A process that provides real-

time access to the system status and accounting
information.
data warehouse (204) A repository of data accumulated

from internal and external data sources, including
financial data, to help management decision making.
flat file (204) A single table of data with user-defined

attributes that is stored separately from any application.
homogeneous systems approach (203) Homogeneous

systems represent one single installation or instance of a
system. It would be considered the opposite of a
heterogeneous system.
heterogeneous systems approach (204)

Heterogeneous systems represent multiple installations or
instances of a system. It would be considered the
opposite of a homogeneous system.
production or live systems (204) Active enterprise

systems that collect and report and are directly affected
by current transactions.
systems translator software (204) Software used to

map the various tables and fields from varied enterprise
systems into a consistent format.
ANSWERS TO PROGRESS
CHECKS
1. Sensors can include door sensors to track movement in
a building, health sensors to track employee health, and
metadata to track transaction activity to name a few.
2. There are many reasons for this trend, with perhaps the
most important being that external auditors are
permitted to rely on the work of internal auditors to
provide support for their opinion of financial statements.
3. A homogeneous system allows effortless transmission
of accounting and auditing data across company units
and international borders. It also allows company
executives (including the chief executive officer, chief
financial officer, and chief information officer),
accounting staff, and the internal audit team to
intimately know the system. In the case of a merger,
integration of the two systems will require less effort
than if they were heterogeneous.
4. The use of audit data standards allows an efficient data
transfer of data in a standardized format that auditors
can use in their audit testing programs. It can also save
the company time and effort in providing its transaction
data in a usable fashion to auditors.
5. If there are too many alarms and exceptions,
particularly with false negatives and false positives,
continuous auditing becomes more of a burden than a
blessing. Work must be done to ensure more true
positives and negatives to be valuable to the auditor.
6. PwC’s Aura system would help identify the questions
and master the data, the first two steps of the IMPACT
model. PwC’s Halo system would help perform the test
plan and address and refine results, the middle two
steps of the IMPACT model. Finally, PwC’s Connect
system would help communicate insights and track
outcomes, the final two steps of the IMPACT model.
page 212

1. Under the guidance of the chief audit executive (CAE)
or another manager, internal auditors build teams to
develop and implement analytical techniques to aid all
of the following audits except:
a. process efficiency and effectiveness.
b. governance, risk, and compliance, including internal
controls effectiveness.
c. tax compliance.
d. support for the financial statement audit.
2. Which audit data standards ledger defines product
master data, location data, inventory on hand data, and
inventory movement?
a. Order to Cash Subledger
b. Procure to Pay Subledger
c. Inventory Subledger
d. Base Subledger
3. Which audit data standards ledger identifies data
needed for purchase orders, goods received, invoices,
payments, and adjustments to accounts?
a. Order to Cash Subledger
b. Procure to Pay Subledger
c. Inventory Subledger
d. Base Subledger
4. A company has two divisions, one in the United States
and the other in China. One uses Oracle and the other
uses SAP for its basic accounting system. What would
we call this?
a. Homogeneous systems
b. Heterogeneous systems
c. Dual data warehouse systems
d. Dual lingo accounting systems
5. Which of the following defines the time period, the level
of materiality, and the expected time for an audit?
a. Audit scope
b. Potential risk
c. Methodology
d. Procedures and specific tasks
6. All of the following may serve as standards for the audit
methodology except:
a. PCAOB’s auditing standards
b. COSO’s ERM framework
c. ISACA’s COBIT framework
d. FASB’s accounting standards
7. When there is an alarm in a continuous audit, but it is
associated with a normal event, we would call that a:
a. false negative.
b. true negative.
c. true positive.
d. false positive.
page 213
8. When there is no alarm in a continuous
audit, but there is an abnormal event, we would call that
a:
a. false negative.
b. true negative.
c. true positive.
d. false positive.
9. If purchase orders are monitored for unauthorized
activity in real time while month-end adjusting entries
are evaluated once a month, those transactions
monitored in real time would be an example of a:
a. traditional audit.
b. periodic test of internal controls.
c. continuous audit.
d. continuous monitoring.
10. Who is most likely to have a working knowledge of the
various enterprise systems that are in use in the
company?
a. Chief executive officer
b. External auditor
c. Internal auditor
d. IT staff
1. Why has most innovation in Data Analytics originated
more in an internal audit than an external audit? Or if
not, why not?
2. Is it possible for a firm to have general journals from a
product like JD Edwards actually reconcile to the
general ledger in SAP to generate financial reports or
drill-down to see underlying transactions? Why or why
not?
3. Is it possible for multinational firms to have many
different financial reporting systems and enterprise
systems packages all in use at the same time?
4. How does the systems translator software work? How
does it store the merged data into a data warehouse?
5. Why is it better to extract data from a data warehouse
than a production or live system directly?
6. Would an auditor view heterogeneous systems as an
audit risk? Why or why not?
7. Why would audit firms prefer to use proprietary
workpapers rather than just storing working papers on
the cloud?
Problems
1. What are the advantages of the use of homogeneous
systems? Would a merger target be more attractive if it
used a similar financial reporting system as the
potential parent company?
2. Consider Exhibit 5-2. Looking at the audit data
standards order-to-cash process, what function is there
for the AR_Adjustments transaction table—that is,
adjustments to the Accounts Receivable? Why is this
an audit data standard, and why is it important for an
auditor to see?
3. Who developed the audit data standards? In your
opinion, why is it the right group to develop and
maintain them rather than, say, the Big 4 firms or a
small practitioner?
page 214
4. Simple to complex Data Analytics can be
applied to a client’s data during the planning stage of
the audit to identify which areas the auditor should
focus on. Which types of techniques or tests might be
used in this stage?
5. What approach should a company make if its
continuous audit system has too many alarms that are
false positives? How would that approach change if
there are too many missed abnormal events (such as
false negatives)?
6. Implementing continuous auditing procedures is similar
to automating an audit plan with the additional step of
scheduling the automated procedures to match the
timing and frequency of the data being evaluated and
the notification to the auditor when exceptions occur. In
your opinion, will the traditional audit be replaced by
continuous auditing?
page 215
Lab 5-1 Create a Common Data

Model
Company summary
The State of Oklahoma captures purchase card
transaction information for each of the state agencies to
determine where resources are used. The comptroller has
asked you to prepare the purchase card transactions
using a common data model based on the audit data
standards so they can be analyzed. The Fiscal Year runs
from July 1 to June 30. FY2018 data includes 420,595
purchase transactions totaling $166,285,071.06.
Technique
Use SQL to transform purchase card data into a common
data model
Software needed
DB Browser for SQLite
Part 1: Import and validate the data.

Part 2: Write a SQL query to transform the data.
Part 1: Import and Validate the Data

In most cases, you will be able to begin working with the
data in the form it is given to you. However, when you are
preparing an analysis that you will use across multiple
data sets and periods, there is some benefit to doing an
additional transformation step to map the data to a
common data model before you load it into the tool. For
example, if you build analyses based on the AICPA’s
Audit Data Standard, you can simply map any new data
table to the common fields without having to rewrite your
analysis. This illustrates the Master the Data portion of the
IMPACT model.
Note: If you’re interested in getting more hands-on
with the programming a data mapping model, search the
Internet for object-relational mapping (ORM) as a method
for querying and transforming data using more advanced
programming tools.
This file uses PCARD_FY2018_SQL data from
Connect.
1. Open DB Browser for SQLite.
2. Click the New Database button, name your file
“PCARD_FY2018.db”, and click Save. When the Edit
table definition window appears, click Cancel.
3. Go to File > Import > Table from CSV file. . .
4. Browse to the PCARD_FY2018_SQL on your computer
and click Open.
5. Name the table PCARD_FY2018 and click OK. Wait for
it to import.
6. Click the Browse Data tab to review the data and verify
that it imported properly.
7. Click the Execute SQL tab and run the following two
queries one at a time to check the data integrity by
comparing the numbers with those at the beginning of
the lab:
SELECT COUNT(Amount)
FROM PCARD_FY2018;
SELECT SUM(Amount)
FROM PCARD_FY2018;
8. Take a screenshot of either query (label it 5-1A).
9. Save your database by going to File > Write changes.
Q1. What was the purpose of executing the two

queries?
page 216
Part 2: Write a SQL Query to

Transform Your Data
Before we transform the data, we need to match the
original attributes to the ADS attributes. In the table below,
the attributes from PCard match up with the ADS with the
exception of four attributes listed in italics. We may want
to keep the attributes as an extension of the ADS so we
don’t lose some potentially useful data.
ADS Purchase_Order
PCard (Source)
(Destination)
“” Purchase_Order_ID
TransactionDate Purchase_Order_Date
“2018” Purchase_Order_Fiscal_Year
BusinessUnit Business_Unit_Code
MerchantName Supplier_Account_ID
CardholderFirstInitial Entered_By
+
CardholderLastName
PostedDate Entered_Date
Amount Purchase_Order_Amount_Local
“USD” Purchase_Order_Local_Currency
SourceYearMonth SourceDate
SourceType SourceType
ItemDescription ItemDescription
MCCDescription MerchantCategory
10. Open your PCARD_FY2018 database from Part 1 in
DB Browser for SQLite.
11. Click the Execute SQL tab.
12. Use the SQL code below as a template to remap the
common elements. As you write the SQL code, replace
[Table] with the name of your table and add a line for
each mapping from the table above where [Field/Value]
elements with either the fields or values from your
source data table. These may include:
a. The name of the field/attribute you’re referencing
(e.g., [TransactionDate]);
b. Double quotes (“” or ’’ depending on the version of
SQL) to leave the field blank;
c. A specific value in quotes (e.g., “2020” AS
Purchase_Order_Fiscal_Year to add the value 2020
for the fiscal year or “USD” to add a local currency); or
d. An expression to combine values (e.g.,
[First_Name]&” “&[Last_Name] AS Entered_By or
[First_Name] ||’ ’|| [Last_Name]AS Entered_By
SELECT
[Field/Value] AS Purchase_Order_ID,
[Field/Value] AS Purchase_Order_Date,
[Field/Value] AS Purchase_Order_Fiscal_Year,
...
FROM[TABLE];
13. Run the query by clicking the Play icon. Correct any
errors in your SQL code if it didn’t run correctly.
page 217
15. Click the Save Results button and choose
Save as view.
16. Name it PCARD_FY2018_ADS and click OK.

17. Save your database by going to File > Write changes.
18. Click the Save Results button and choose Export to
CSV followed by Save.
19. Name your CSV file PCARD_FY2018_ADS and click
OK.
20. Close DB Browser for SQLite.
Q2. How does mapping the data to a common data
model make it easier to perform analysis in the
future?
End of Lab
Lab 5-2 Create a Dashboard Based

on a Common Data Model
Company summary
The State of Oklahoma captures purchase card
transaction information for each of the state agencies to
determine where resources are used. The comptroller has
asked you to prepare the purchase card transactions
using a common data model based on the audit data
standards so they can be analyzed. The Fiscal Year runs
from July 1 to June 30. FY2018 data includes 420,595
purchase transactions totaling $166,285,071.06.
Technique
Use Tableau visualize purchase card data based on the
Audit Data Standard
Software needed
Tableau
Part 1: Import and validate the data.

Part 2: Create four visualizations.
Part 3: Combine the visualizations into a dashboard.
page 218
Part 1: Import and Validate the Data

Start by mastering the data and loading into Tableau:
1. Open Tableau.
2. Click Connect to a File and choose Text file.
3. Navigate to your PCARD_FY2018_ADS.csv from Lab
5-1 or Connect and click Open.
4. On the Data Source tab, click Update Now to preview
your data and verify that it loaded correctly.
5. Important! Adjust your data types so they will be
correctly interpreted by Tableau. Click the #, calendar,
or Abc icon above each field and choose the following:
1. Dates: Purchase Order Date, Entered Date, Source
Date
2. Number (decimal): Purchase Order Amount Local
3. String: everything else.
7. Save your workbook as PCARD_FY2018
Q1. What do you think would happen if you didn’t

update the data types in Tableau before you
perform your analysis?
Part 2: Create Four Visualizations

Next, perform the test plan, address results, and
communicate your findings:
8. Starting on Sheet1, create the following visualizations
(each on a separate sheet):
a. Show a distribution of total purchase amount by
cardholder showing the business unit as a color,
sorted in descending order by purchase amount:
I. Columns: SUM(Purchase Order Amount Local)
II. Rows: Entered By
III. Marks: Business Unit Code (color)
b. Show a bar chart with the total purchases for each
month showing the category as a color:
I. Columns: YEAR(Purchase Order Date)
MONTH(Purchase Order Date)
II. Rows: SUM(Purchase Order Amount Local)
III. Marks: Merchant Category (color)
c. Show a tree map of total purchases by business unit:
I. Marks: SUM(Purchase Order Amount Local) (size);
SUM(Purchase Order Amount Local) (color);
Business Unit Code (label)
d. Show a distribution of total purchase amount by
supplier showing the month of purchase as a color,
sorted in descending order by purchase amount:
I. Columns: SUM(Purchase Order Amount Local)
II. Rows: Supplier Account ID
III. Marks: MONTH(Purchase Order Date) (color)
9. Take a screenshot of any one of the four visualizations
(label it Lab 5-2B).
Part 3: Combine the Visualizations

into a Dashboard
10. In your Tableau workbook, create a Dashboard tab and
drag each of the four visualizations into it from the pane
on the left.
11. Take a screenshot (label it Lab 5-2C).
page 219
Q2. Which five suppliers receive the most

purchases?
Q3. Which month has the highest total purchase

amount? The lowest?
Q4. What would happen if you change the data
source in Tableau to another file that is
formatted following the ADS for purchase
orders?
End of Lab
Lab 5-3 Set Up a Cloud Folder
Auditors collect evidence in electronic workpapers that
include a permanent file with information about policies
and procedures and a temporary file with evidence related
to the current audit. These files could be stored locally on
a laptop, but the increased use of remote communication
makes collaboration through the cloud more necessary.
There are a number of commercial workpaper
applications, but we can simulate some of those features
with consumer cloud platforms, like Microsoft OneDrive.
Company summary
You have rotated into the internal audit department at a
mid-sized manufacturing company. Your team is still using
company email to send evidence back and forth, usually
in the form of documents and spreadsheets. There is a lot
of duplication of these files, and no one is quite sure
which version is the latest. You see an opportunity to
streamline this process using OneDrive.
Technique
Gather documents, explore document history and
revisions
Software needed
A modern web browser
Part 1: Create a shared folder.

Part 2: Upload files.
Part 1: Create a Shared Folder

Note: These instructions are specific to the free consumer
version of Microsoft OneDrive. The approach is similar for
competing products, such as Box, Dropbox, Google Drive,
or other commercial products.
1. Go to OneDrive.com.
2. Click Sign in in the top right corner.
3. Sign in with your Microsoft account. (If your
organization subscribes to Office 365, use your school
or work account here.)
4. On the main OneDrive screen, click New > Folder.
5. Name your folder DA Audit Working Papers and click
Create.
6. Open your new folder and click Share from the bar at
the top of the screen.
page 220
7. Add the email address of one of your
classmates or your instructor, as directed. Choose
Anyone with a link can edit from the sharing options,
then click Send.
Q1. What advantage is there to sharing files in one

location rather than emailing copies back and
forth?
Part 2: Upload Files

Now that you have a folder, you can upload some
documents that will be useful for labs in this chapter and
the next.
9. From Connect, download the Audit Analytics Lab
Files, as directed by your instructor.
10. Unzip the file you downloaded to your computer. You
should see two folders: Master Audit File and Current
Audit File.
11. Return to your OneDrive DA Audit Working Papers
folder, and upload the two folders:
a. Click Upload > Folder in OneDrive and navigate to
the folder where you unzipped the lab files.
b. Or drag and drop the two folders from your desktop to
the OneDrive window in your browser.
12. You should see two new folders in your OneDrive.
Because you added them to a shared folder, the people
you shared the folder with can now see these as well.
Q2. Explore the two folders you just uploaded.
What kinds of documents and files do you
see?
Q3. How do you think these files can be used for
data analysis?
End of Lab
Lab 5-4 Review Changes to

Working Papers
See Lab 5-3 for background information on this lab. The
goal of a shared folder is that other members of the audit
team can contribute and edit the documents. Commercial
software provides an approval workflow and additional
internal controls over the documents to reduce
manipulation of audit evidence, for example. For
consumer cloud platforms, one control appears in the
versioning of documents. As revisions are made, old
copies of the documents are kept so that they can be
reverted to, if needed. Note: You must complete Lab 5-3
before beginning this lab.
Part 1: Upload revised documents.

Part 2: Review document revision history.
Part 1: Upload Revised Documents

Let’s start by making changes to files in your DA Working
Papers.
1. From Connect, download Audit Analytics Lab Files
Revised, as directed by your instructor.
2. Unzip the file you downloaded to your computer. You
should see two files: Audit Plan and User_Listing.
page 221
3. Return to your OneDrive DA Audit
Working Papers folder, and upload the Audit Plan into
your Master Audit File and the User_Listing into your
Current Audit File. You will be prompted to Replace or
Keep Both files. Click Replace for each.
Part 2: Review Document Revision

History
Now let’s look at the history of the document.
5. Right-click on one of the newly uploaded files, and
choose Version history from the menu that appears.
The document will open with a version pane appearing
on the left.
6. Click the older version of the file from the Versions
pane. Newer versions are at the top.
8. Move between the old version of the file and the current
version by clicking the time stamp in the panel on the
left.
Q1. What has changed between these two
versions?
End of Lab
Lab 5-5 Identify Audit Data

Requirements
As the new member of the internal audit team, you have
introduced your team to the shared folder and are in the
process of modernizing the internal audit at your firm. The
chief audit executive is interested in using Data Analytics
to make the audit more efficient. Your internal audit
manager agrees and has tasked you with reviewing the
audit plan. She has provided three “audit action sheets”
with procedures that they have been using for the past
three years to evaluate the procure-to-pay (purchasing)
process and is interested in your thoughts for modernizing
them. Note: You should complete Lab 5-3 before
beginning this lab.
Technique
Review the audit plan, look for procedures involving data,
and identify the locations of the data.
Software needed
Part 1: Look for audit procedures that evaluate data.

Part 2: Identify the location of the data.
Part 1: Look for Audit Procedures

That Evaluate Data
1. Open your DA Audit Working Papers folder on
OneDrive or download the Audit Action Sheets
document from Connect.
2. Look inside the Master Audit File for the document
titled Audit Action Sheets and open it to edit it.
page 222
3. Use the Yellow highlighter to identify any
master or transaction tables, such as “Vendors” or
“Purchase Orders.”
4. Use the Green highlighter to identify any fields or
attributes, such as “Name” or “Date.”
5. Use the Blue highlighter to identify any specific values
or rules, such as “TRUE,” “January 1st,” “Greater than .
. .”
6. Create a new spreadsheet called Audit Automation
Summary in your Master Audit File and summarize
your highlighted data elements from the three audit
action sheets. Use the following headers:

Q1. Read the first audit action sheet. What other
data elements that are not listed in the
procedures do you think would be useful in
analyzing this account?
Part 2: Identify the Location of the

Data
Now that you have analyzed the action sheets, look
through the systems documentation to see where those
elements exist.
8. In the Master Audit File, open the UML System
Diagram and Data Dictionary files.
9. Using the data elements you identified in your Audit
Automation Summary file, locate the actual names of
tables and attributes and acceptable data values. Add
them in three new columns in your summary:
Database Database Acceptable

Table Attribute Values
Q2. Which attributes were difficult to locate or in

unexpected places in the database?
11. Save and close your file.
End of Lab
Lab 5-6 Prepare Audit Plan

With the data elements identified, you can formalize your
internal audit plan. In the past, your internal audit
department performed each of the three action sheets
once every 24 months. You have shared how increasing
the frequency of some of the tests would provide a better
control for the process and allow the auditor to respond
quickly to the exceptions. Your internal audit manager has
asked you to propose a new schedule for the three audit
action sheets. Note: You should complete Lab 5-5 before
beginning this lab.
Technique
Review the audit plan, identify procedures that must be
completed manually, and identify those that can be
automated and scheduled.
Also determine when the procedures should occur.
page 223
Software needed

Evaluate the timing and scheduling of audit procedures.
Set the Frequency of Your Audit

Procedures
1. Open your Audit Automation Summary created in
Lab 5-5.
2. Add two new columns:
Auto/Manual Frequency
3. For each element and rule, determine whether it

requires manual review or can be performed
automatically and alert auditors when exceptions occur.
Add either “Auto” or “Manual” to that column.
4. Finally, determine how frequently the data should be
evaluated. Indicate “Daily,” “Weekly,” “Monthly,”
“Annually,” or “During Audit.” Think about when the data
are being generated. For example, transactions occur
every day, but new employees are added every few
months.
6. Save and close your file.
End of Lab
page 224
Chapter 6
Audit Data Analytics

In this chapter, we focus on substantive testing within the audit setting.
We identify when to use descriptive analytics to understand the
business environment, diagnostic analytics to compare expectation with
reality, including Benford’s analysis, and how predictive and prescriptive
analytics are used to address future audit concerns.
A Look Back
In Chapter 5, we introduced Data Analytics in auditing by considering
how both internal and external auditors are using technology in general,
and audit analytics specifically, to evaluate firm data and generate
support for management assertions. We emphasized audit planning,
audit data standards, continuous auditing, and audit working papers.
A Look Ahead
Chapter 7 explains how to apply Data Analytics to measure
performance for management accountants. By measuring past
performance and comparing it to targeted goals, we are able to assess
how well a company is working toward a goal and recommend actions
to correct unexpected patterns.
page 225
ra2studio/Shutterstock
Internal auditors at Hewlett-Packard Co. (HP) understand how Data Analytics can
improve processes and controls. Management identified abnormal behavior with manual
journal entries, and the internal audit department responded by working with various
governance and compliance teams to develop dashboards that would allow them to
monitor accounting activity. The dashboard made it easier for management and the
auditors to follow trends, identify spikes in activity, and drill down to identify the individuals
posting entries. Leveraging accounting data allows the internal audit function to focus on
the risks facing HP and act on data in real time by implementing better controls. Audit
data analytics provides an enhanced level of control that is missing from a traditional
periodic audit.
OBJECTIVES
LO 6-1 Understand different types of analysis for auditing and

when to use them.
LO 6-2 Understand basic descriptive audit analyses.
LO 6-3 Understand more complex statistical analyses, including
Benford’s law.
LO 6-4 Understand advanced predictive and prescriptive
analytics.
page 226
WHEN TO USE AUDIT DATA ANALYTICS

LO 6-
1
Understa
nd
different
types of
analysis
for
auditing
and
when to
use
them.
As discussed in Chapter 5, Data Analytics can be applied to the auditing

function to increase coverage of the audit, while reducing the time the
auditor dedicates to the audit tasks. Think about the nature, extent, and
timing of audit procedures.
Nature represents why auditors perform audit procedures. In other
words, nature helps determine the objectives of the audit and the
outputs generated by the business processes.
Extent indicates how much auditors can test. The prevalence of data
has expanded the extent of audit testing.
Timing tells us how often the procedure should be run. Automation
allows auditors to run analytics on a schedule and receive real-time
alerts when exceptions occur.
All three of these elements help auditors identify when to apply Data
Analytics to the audit process. Auditors should evaluate current
capabilities within their department and identify the goal of Data
Analytics. Does it add value? Does it enhance the process? Does it help
the auditor be more efficient and effective? Applying Data Analytics, in
theory, should add value. In reality, it is easy to overpromise on the
expected benefits of Data Analytics and underdeliver with the results.
Without clear objectives and expected outcomes, audit departments will
fail with their use of Data Analytics. Here we refer once again to the
IMPACT model.
Identify the Problem

What is the audit department trying to achieve using data analytics? Do
auditors need to analyze the segregation of duties to test whether
internal controls are operating effectively? Are auditors looking for
operational inefficiencies, such as duplicate payments of invoices? Are
auditors trying to identify phantom employees or vendors? Are auditors
trying to collect evidence that the company is complying with specific
regulations? Are auditors trying to test account balances to tie them to
the financial statements?
These activities support the functional areas of compliance, fraud
detection and investigation, operational performance, and internal
controls for internal audit departments as well as the financial reporting
and risk assessment functions of external audit.
Master the Data

In theory, auditors should have read-only access to enterprise data
through a nonproduction data warehouse. In practice, they make
multiple requests for flat files or data extractions from the IT manager
that they then analyze with a software tool, such as Excel or Tableau.
Most audit data are provided in structured or tabular form, such as a
spreadsheet file.
Regardless of the source or type, the audit data standards provide a
general overview of the basic data that auditors will evaluate. For
example, consider the Sales_Orders table from the standards shown in
Table 6-1. An auditor interested in user activity would want to focus on
the Sales_Order_ID, Sales_Order_Date, Entered_By, Entered_Date,
Entered_Time, Approved_By, Approved_Date, Approved_Time, and
Sales_Order_Amount_Local attributes. These may give insight into
transactions on unusual dates, such as weekends, or unusually high
volume by specific users.
TABLE 6-1
Elements in the Sales_Order Table from the Audit Data Standards
Field Name Description

Sales_Order_ID Unique identifier for each sales order.
This ID may need to be created by
concatenating fields (e.g., document
number, document type, and year) to
uniquely identify each sales order.
Sales_Order_Document_ID Identification number or code on the
sales order.
Sales_Order_Date The date of the sales order, regardless
of the date the order is entered.
Sales_Order_Fiscal_Year Fiscal year in which the
Sales_Order_Date occurs: YYYY for
delimited, CCYYMMDD fiscal year-end
(ISO 8601) for XBRL-GL.
Sales_Order_Period Fiscal period in which the
Sales_Order_Date occurs. Examples
include W1–W53 for weekly periods,
M1–M12 for monthly periods, and Q1–
Q4 for quarterly periods.
Business_Unit_Code Used to identify the business unit,
region, branch, and so on at the level
that financial statements are being
audited. Must match a
Business_Unit_Code in the
Business_Unit_Listing file.
Customer_Account_ID Identifier of the customer from whom
payment is expected or to whom
unused credits have been applied. Must
match a Customer_Account_ID in the
Customer_Master_Listing_YYYYMMDD
file.
Entered_By User_ID (from User_Listing file) for
person who created the record.
Entered_Date Date the order was entered into the
system. This is sometimes referred to
as the creation date. This should be a
system-generated date (rather than
user-entered date), when possible. This
date does not necessarily correspond
with the date of the transaction itself.
Entered_Time The time this transaction was entered
into the system. ISO 8601 representing
time in 24-hour time (hhmm) (e.g., 1:00
p.m. = 1300).
Approved_By User ID (from User_Listing file) for
person who approved customer master
additions or changes.
Approved_Date Date the entry was approved.
Approved_Time The time the entry was approved. ISO
8601 representing time in 24-hour time
(hhmm) (e.g., 1:00 p.m. = 1300).
Last_Modified_By User_ID (from User_Listing file) for the
last person modifying this entry.
Last_Modified_Date The date the entry was last modified.
Last_Modified_Time The time the entry was last modified.
ISO 8601 representing time in 24-hour
time (hhmm) (e.g., 1:00 p.m. = 1300).
Sales_Order_Amount_Local Sales monetary amount recorded in the
local currency.
Sales_Order_Local_Currency The currency for local reporting
requirements. See ISO 4217 coding.
Segment01 Reserved segment field that can be
used for profit center, division, fund,
program, branch, project, and so on.
Segment02 See above.
Source: Adapted from

https://www.aicpa.org/content/dam/aicpa/interestareas/frc/assuranceadvisoryservices/download
abledocuments/auditdatastandards/auditdatastandards.o2c.july2015.pdf, accessed January 1,
2018
There are also many pieces of data that have traditionally evaded
scrutiny, including handwritten logs, manuals and handbooks, and other
paper or text-heavy documentation. Essentially, manual tasks including
observation and inspection are generally areas where Data Analytics
may not apply. While there have been significant advancements in
artificial intelligence, there is still a need for auditors to exercise their
judgment, and data cannot always supersede the auditor’s reading of
human behavior or a sense that something may not be quite right even
when the data say it is. At least not yet.
page 227
Data may also be found in unlikely places. An auditor may be tasked
with determining whether the steps of a process are being followed.
Traditional evaluation would involve the auditor observing or
interviewing the employee performing the work. Now that most
processes are handled through online systems, an auditor can perform
Data Analytics on the time stamps of the tasks and determine the
sequence of approvals in a workflow along with the page 228
amount of time spent on each task. This form of
process mining enables insight into areas where greater efficiency can
be applied. Likewise, data stored in paper documents, such as invoices
received from vendors, can be scanned and converted to tabular data
using specialized software. These new pieces of data can be joined to
other transactional data to enable new, thoughtful analytics.
There is an increasing opportunity to work with unstructured Big
Data to provide additional insight into the economic events being
evaluated by the auditors, such as surveillance video or text from e-
mail, but those are still outside the scope of current Data Analytics that
an auditor would develop.
Perform the Test Plan

While there are many different tests or models that auditors can
incorporate into their audit procedures, Data Analytics procedures in
auditing traditionally are found in computer-assisted audit techniques
(CAATs). CAATs are automated scripts that can be used to validate
data, test controls, and enable substantive testing of transaction details
or account balances and generate supporting evidence for the audit.
They are especially useful for re-performing calculations, identifying
high-risk samples, and performing other analytical reviews to identify
unusual patterns of behavior or unusual items.
Most CAATs are designed to summarize and describe the data being
evaluated based on a predetermined expected outcome. For example,
an auditor evaluating an incentive plan that gives employees bonuses
for opening new accounts would evaluate the number of new accounts
by employee and the amount of bonus paid to see if they were aligned.
The auditor could look for a count of new accounts by account type,
count the number of customers, evaluate the opening date, and sort the
data by employee to show the top-performing employees. These
descriptive analytics summarize activity or master data elements
based on certain attributes. The auditor may select a sample of the
accounts to verify that they were opened and the documentation exists.
Once an auditor has a basic understanding of the data, he or she
can then perform diagnostic analytics, which look for correlations or
patterns of interest in the data. For example, the auditor may look for
commonalities between the customers’ demographic data and the
employees’ data to see if employees are creating new accounts for fake
customers to inflate their performance numbers. They may also focus
on customers who have common attributes like location or account age.
Outliers may warrant further investigation by the auditor as they
represent increased risk and/or exposure.
An auditor then performs predictive analytics, where he or she
attempts to find hidden patterns or variables that are linked to abnormal
behavior. The auditor uses the variables to build models that can be
used to predict a likely value or classification. In our example, the
predictive model might flag an employee or customer with similar
characteristics to other high-risk employees or customers whenever a
new account is opened.
Finally, the auditor may generate prescriptive analytics that identify
a course of action for him or her to take based on the actions taken in
similar situations in the past. These analytics can assist future auditors
who encounter similar behavior. Using artificial intelligence and machine
learning, these analytics become decision support tools for auditors who
may lack experience to find potential audit issues. For example, when a
new account is created for a customer who has been inactive for more
than 12 months, a prescriptive analytic would allow an auditor to ask
questions about the transaction to learn whether this new account is
potentially fake, whether the employee is likely to create other fake
accounts, and whether the account and/or employee should be
suspended or not. The auditor would take the output, apply judgment,
and proceed with what he or she felt was the appropriate action.
page 229
Most auditors will perform descriptive and diagnostic analytics as
part of their audit plan. On rare occasions, they may experiment with
predictive and prescriptive analytics directly. More likely, they may
identify opportunities for the latter analytics and work with data scientists
to build those for future use.
Some examples of CAATs and audit procedures related to the
descriptive, diagnostic, predictive, and prescriptive analytics can be
found in Table 6-2.
TABLE 6-2
Examples of Audit Data Analytics
Example
Analytic Type Example Audit Procedures
CAATs
Descriptive— Age analysis— Analysis of new accounts opened
summarizes groups balances and employee bonuses by
activity or by date employee and location.
master data Sorting— Count the number/dollar amount of
based on identifies largest transactions that occur outside
certain or smallest normal business hours or at the
attributes values and helps end/beginning of the period.
identify patterns
Summary
statistics—
mean, median,
min, max, count,
sum
Sampling—
random and
monetary unit
Example
CAATs
Diagnostic— Z-score—outlier Analysis of new accounts reveals
detects detection that an agent has an unusual
correlations Benford’s law— number of new accounts opened
and patterns of identifies for customers who have been
interest and transactions or inactive for more than 12 months.
compares users with non- An auditor assigns an expected
them to a typical activity Benford’s value to purchase
benchmark based on the transactions, then averages them
distribution of by employee to identify employees
digits with unusually large purchases.
Drill-down— An auditor filters out transactions
explores the that are below a materiality
details behind threshold.
the values
Exact and fuzzy
matching—joins
tables and
identifies
plausible
relationships
Sequence check
—detects gaps
in records and
duplicates
entries
Stratification—
groups data by
categories
Clustering—
groups records
by non-obvious
similarities
Example
CAATs
Predictive— Regression— Analysis of new accounts opened
identifies predicts specific for customers who have been
common dependent inactive for more than 12 months
attributes or values based on collects data that are common to
patterns that independent new account opening, such as
may be used variable inputs account type, demographics, and
to identify Classification— employee incentives.
similar activity predicts a
category for a
record
Probability—
uses a rank
score to
evaluate the
strength of
classification
Sentiment
analysis—
evaluates text
for positive or
negative
sentiment to
predict positive
or negative
outcomes
Example
CAATs
Prescriptive— What-if analysis Analysis determines procedures to
recommends —decision follow when new accounts are
action based support systems opened for inactive customers,
on previously Applied statistics such as requiring approval.
observed —predicts a
actions specific outcome
or class
Artificial
intelligence—
uses
observations of
past actions to
predict future
actions for
similar events
page 230
While many of these analyses can be performed using Excel, most

CAATs are built on generalized audit software (GAS), such as IDEA,
ACL, or TeamMate Analytics. The GAS software has two main
advantages over traditional spreadsheet software. First, it enables
analysis of very large datasets. Second, it automates several common
analytical routines, so an auditor can click a few buttons to get to the
results rather than writing a complex set of formulas. GAS is also
scriptable and enables auditors to record or program common analyses
that may be reused on future engagements.

The models selected by the auditors will generate various results. A
sample selection may give auditors a list of high-risk transactions to
evaluate. A segregation of duties analysis may spit out a list of users
with too much access. In every case, the auditors should develop
procedures in the audit plan for handling these lists, exceptions, and
anomalies. The process may be to evaluate documentation related to
the sample, review employees engaging in risky activity, or simply notify
the audit committee of irregular behavior.
Many analytics can be adapted to create an audit dashboard for
measuring risk in transactions or exceptions to control rules, particularly
if the firm has adopted continuous auditing. The primary output of
CAATs is evidence that may be used to test management assertions
about the processes, controls, and data quality. This evidence is
included in the audit workpapers.
Track Outcomes
The detection and resolution of audit exceptions may be a valuable
measure of the efficiency and effectiveness of the internal audit function
itself. Additional analytics may track the number of exceptions over time
and the time taken to report and resolve the issues. For the CAATs
involved, a periodic validation process should occur to ensure that they
continue to function as expected.
PROGRESS CHECK
1. Using Table 6-2 as a guide, compare and contrast descriptive
and diagnostic analytics. How might these be used in an audit?
2. In a continuous audit, how would a dashboard help to
communicate audit findings and spur a response?
DESCRIPTIVE ANALYTICS
LO 6-
2
Understa
nd basic
descriptiv
e audit
analyses.
Now that you’ve been given an overview of the types of CAATs and
analytics that are commonly used in an audit, we’ll dive a little deeper
into how these analytics work and what they generate. Remember that
descriptive analytics are useful for sorting and summarizing data to
create a baseline for more advanced analytics. These analytics enable
auditors to set a baseline or point of reference for their evaluation. For
example, if an auditor can identify the median value of a series of
transactions, he or she can make a judgment as to how much higher the
larger transactions are and whether they represent outliers or
exceptions.
page 231
In this and the next few sections, we’ll present some examples of
procedures that auditors commonly use to evaluate enterprise data. In
these examples, we show the basic process for Excel, including
formulas, and IDEA. Note that in the Excel formulas, we identify data
elements in [brackets]. To use these formulas, replace the bracketed
[data element] with a value or range of values as appropriate. For
example, [Aging date] would be replaced with C3 if the data are in
column C, row 3.
Age Analysis
Aging of accounts receivable and accounts payable helps determine the
likelihood that a balance will be paid. This substantive test of account
balances evaluates the date of an order and groups it into buckets
based on how old it is, typically in 0–30, 31–60, 61–90, and >90 days,
or similar. See Table 6-3 for an example. Extremely old accounts that
haven’t been resolved or written off should be flagged for follow-up by
the auditor. It could mean that (1) the data are bad, (2) a process is
broken, (3) there’s a reason someone is holding that account open, or
(4) it was simply never resolved.
TABLE 6-3
Aging of Accounts Receivable
There are many ways to calculate aging in Excel, including using

pivot tables. If you have a simple list of accounts and balances, you can
calculate a simple age of accounts in Excel using the following
procedure.
Data
Customer/vendor name
Unpaid order number
Order date
Amount
In Excel
1. Open your worksheet.
2. Add a cell with the aging date.
3. Add a calculated column for the days outstanding: =[Aging date]–
[Order date].
4. Add four new calculated columns for the buckets:
a. 0–30 days: =IF([Aging date]–[Order date]<=30,[Amount],0).
b. 31–60 days: =IF(AND([Aging date]–[Order date]<=60, [Aging date]–
[Order date]>30),[Amount],0).
c. 61–90 days: =IF(AND([Aging date]–[Order date]<=90, [Aging date]–
[Order date]>60),[Amount],0).
d. >90 days: =IF([Aging date]–[Order date]>90),[Amount],0).
5. Copy the formulas for all records.
6. Add a total to the bottom of each bucket: =SUM([bucket column]).
In IDEA
2. Go to Analysis > Categorize > Aging.
Source: IDCaseWare IDEA
page 232
3. Select aging date, transaction date for the Aging field to
use, and amount for the Amount field to total field.
4. Click OK.
Sorting
Sometimes, simply viewing the largest or smallest values can provide
meaningful insight. Sorting in ascending order shows the smallest
number values first. Sorting in descending order shows the largest
values first.
Data
Any numerical, date, or text data of interest
In Excel
2. Select the data you wish to sort.
3. Go to Home > Styles > Format as Table.
4. Click the drop-down arrow next to the header or the column you want
to sort.
5. Click Sort A to Z for ascending order or Sort Z to A for descending
order.
page 233
In IDEA
1. Open your data table.
2. Go to Data > Order > Sort.
3. Choose your desired field(s) and direction, Ascending or Descending.
4. Click OK.
Summary Statistics
Summary statistics provide insight into the relative size of a number
compared with the population. The mean indicates the average value,
while the median produces the middle value when all the transactions
lined up in a row. The min shows the smallest value, while the max
shows the largest. Finally, a count tells how many records exist, where
the sum adds up the values to find a total. Once summary statistics are
calculated, you have a reference point for an individual record. Is the
amount above or below average? What percentage of the total does a
group of transactions make up?
Data
Any numerical data, such as a dollar amount or quantity
In Excel
1. Open your workbook.
2. Add the following calculated values:
Mean: =AVERAGE([range]).
Median: =MEDIAN([range]).
Minimum: =MIN([range]).
Maximum: =MAX([range]).
Count: =COUNT([range]).
Sum: =SUM([range]).
3. Alternatively, format your data as a table and show the total row at the
bottom:
a. Select your data.
b. Go to Home > Styles > Format as Table.
c. Select a table style and click OK.
d. Go to Table Tools > Design > Table Style Options and click the
Total Row box.
e. Go to the bottom of the table and click the drop-down arrow next to
the column total value that appears. Choose an appropriate statistic.
In IDEA
2. In the Properties pane on the right, click Field Statistics.
3. Allow IDEA to create statistics for fields without statistics, if prompted.
4. In the output screen, you can click any blue number to locate those
transactions.
Sampling
Sampling is useful when you have manual audit procedures, such as
testing transaction details or evaluating source documents. The idea is
that if the sample is an appropriate size, the features of the sample can
be confidently generalized to the population. So, if the sample has no
errors (misstatement), then the population is unlikely to have errors as
well. Of course, sampling has its limitations. The confidence level is not
a guarantee that you won’t miss something critical like fraud. But it does
limit the scope of the work the auditor must perform.
page 234
There are three determinants for sample size: confidence level,

tolerable misstatement, and estimated misstatement.
Data
Any list of transactions or master data
In Excel
1. Enable Analysis ToolPak:
a. Go to File > Options > Add-ins > Excel Add-ins > Go.
b. Select Analysis ToolPak, and click OK.
2. Go to Data > Analysis > Data Analysis.
3. Click Sampling, then OK.
a. Select your input range, usually the transaction number.
b. Choose Random, and input the number of items you want to appear
in your sample.
c. Click OK.
4. A new worksheet will appear with a list of your randomly selected
transactions.
In IDEA
2. Go to Analysis > Sample > Random.
a. Input number of records to select for your sample size.
b. Change other values as needed.
c. Click OK.
3. A new worksheet will be created with your random sample.
Monetary unit sampling (MUS) allows auditors to evaluate account
balances. MUS is more likely to pull accounts with large balances
(higher risk and exposure) because it focuses on dollars, not account
numbers.
Data
The book value of the financial accounts you’re evaluating
The sample size
In Excel
1. Find the sampling interval. Divide the book value by sample size. In
this example, 1,000,000 is the total rounded book value of a GL
account and 132 is the desired sample size.
a. 1,000,000/132 = 7,575
2. Choose one of the following options to sort the line items or individual
accounts. Then add a column to calculate the cumulative balance.
a. Alphabetically by name.
b. Numerically by number.
c. By date.
3. Pick a random number between 1 and your sampling interval.
a. This will be the starting value. For example, 1,243.
4. Go down the list of cumulative balances until you pass your random
number.
a. For example, test the first account that passes 1,243.
5. Continue down the list of cumulative balances until you pass the next
sampling interval.
a. For example, test the second account that passes 1,243 + 7,575 =
8,818.
6. Repeat step 5 until you run out of accounts.
a. 8,818 + 7,575 = 16,393; 16,393 + 7,575 = 23,968 . . .
page 235
In IDEA
1. Open your data table.
2. Go to Analysis > Sample > Monetary Unit > Plan.
a. Choose your monetary value field.
b. Set your confidence level, tolerable error, and expected error.
c. Click Estimate to calculate your sample size.
d. Adjust other values as needed, then click Accept.
e. Click OK.
3. A new worksheet will appear with your sample transactions.
PROGRESS CHECK
3. What type of descriptive analytics would you use to find
negative numbers that were entered in error?
4. How does monetary unit sampling help you isolate the items of
greatest potential significance to an auditor in evaluating
materiality?
DIAGNOSTIC ANALYTICS AND

BENFORD’S LAW
LO 6-
3
Understa
nd more
complex
statistical
analyses,
including
Benford’s
law.
Diagnostic analytics provide more details into not just the records, but
also records or groups of records that have some standout features.
They may be significantly larger than other values, may not match a
pattern within the population, or may be a little too similar to other
records for an auditor’s liking. Here we’ll identify some common
diagnostic analytics and how to use them.
Z-Score
A standard score or Z-score is a concept from statistics that assigns a
value to a number based on how many standard deviations it stands
from the mean, shown in Exhibit 6-1. By setting the mean to 0, you can
see how far a point of interest is above or below it. For example, a point
with a Z-score of 2.5 is two-and-a-half standard deviations above the
mean. Because most values that come from a large population tend to
be normally distributed (frequently skewed toward smaller values in the
case of financial transactions), nearly all (98 percent) of the values
should be within plus-or-minus three standard deviations. If a value has
a Z-score of 3.9, it is very likely an outlier that warrants scrutiny.
EXHIBIT 6-1
Z-Scores
The Z-score shows the relative position of a point of interest to the population.
Source: http://www.dmaictools.com/wp-content/uploads/2012/02/z-definition.jpg
page 236
In Excel
1. Calculate the average: =AVERAGE([range]).
2. Calculate the standard deviation: =STDEVPA([range]).
3. Add a new column called “Z-score” next to your number range.
4. Calculate the Z-score: =STANDARDIZE([value],[mean],[standard
deviation]).
a. Alternatively: =([value]–[mean])/[standard deviation].
5. Sort your values by Z-score in descending order.
In IDEA
Z-score calculation is not a default feature of IDEA.
Benford’s Law
Benford’s law states that when you have a large set of naturally
occurring numbers, the leading digit(s) are more likely to be small. The
economic intuition behind it is that people are more likely to make $10,
$100, or $1,000 purchases than $90, $900, or $9,000 purchases. This
law has been shown in many settings, such as the amount of electricity
bills, street addresses, and GDP figures from around the world (as
shown in Exhibit 6-2).
EXHIBIT 6-2
Benford’s Law
Benford’s law predicts the distribution of first digits.
In auditing, we can use Benford’s law to identify transactions or

users with nontypical activity based on the distribution of the first digits
of the number. For example, assume that purchases over $500 require
manager approval. A cunning employee might try to make large
purchases that are just under the approval limit to avoid suspicion. She
will even be clever and make the numbers look random: $495, $463,
$488, etc. What she doesn’t realize is that the frequency of the leading
digit 4 is going to be much higher than it should be, shown in Exhibit 6-
3. Benford’s law can also detect random computer-generated numbers
because those will have equally distributed first digits. Adding additional
leading digits refines the analysis.
EXHIBIT 6-3
Using Benford’s Law
Structured purchases may look normal, but they alter the distribution under Benford’s law.
We show an illustration of how to evaluate data and their frequency
with respect to Benford’s law in both Excel and IDEA.
page 237
Data
Large set of numerical data, such as monetary amounts or quantities
In Excel
1. Open your spreadsheet.
2. Add a new column and extract the leading digit: =LEFT([Amount],1).
3. Create a frequency distribution:
a. Create a list on your sheet and copy the Digit, Actual Count, Actual
%, and Expected % columns with their corresponding values and
formulas, shown in Table 6-4.
TABLE 6-4
Illustration of Benford’s Law
Digit Actual Count Actual % Expected %

Digit Actual Count Actual % Expected %
1 =COUNTIF([Range], (=[Actual 30.1%

[Digit]) Count]/SUM[Actual
Count])
2 ... ... 17.6%
3 ... ... 12.5%
4 ... ... 9.6%
5 ... ... 7.9%
6 ... ... 6.7%
7 ... ... 5.8%
8 ... ... 5.1%
9 ... ... 4.6%
=SUM([Actual =SUM([Actual %]) =SUM([Expected

Count]) %])
4. Create a combo chart to plot your actual and expected percentages:

a. Highlight the Actual % and Expected % columns.
b. Go to Insert > Charts > Recommended Charts.
c. Click the All Charts tab.
d. Choose Combo from the list on the left.
e. Click Custom Combination.
page 238
f. For the Actual %, choose Clustered Column.
g. For the Expected %, choose Scatter.
h. Click OK.
i. Adjust and format your chart as needed.
In IDEA
2. Go to Analysis > Explore > Benford’s Law.
a. Choose the numerical field to analyze.
b. Only check First digit. Uncheck everything else.
c. Click OK.
3. A graph will appear with the Benford’s expected amount and the actual
frequency of the dataset.
4. Click any digits that are significantly above the bounds and choose
Extract Records.
Bonus: Use the average expected Benford’s law value to identify
specific employees with abnormally large transactions. In this case, a
user with lots of transactions should have an average expected
Benford’s law percentage of 11.1 percent or above. Employees whose
average purchases are closer to 8 or 5 percent have a lot of 7, 8, and 9
values that are skewing their average.
In Excel
1. Open your spreadsheet with financial data that contain an employee
name and transaction amount.
2. Add a new column and extract the leading digit. Note: the LEFT()
function in Excel creates a text value. Use VALUE() to convert it to a
number.=VALUE(LEFT([Amount],1))
3. Add the expected Benford’s law percentages to your sheet similar to
Table 6-5.
TABLE 6-5
Expected Benford’s Law Percentages
Digit Benford Expected %
1 30.1%
Digit Benford Expected %
2 17.6%
3 12.5%
4 9.6%
5 7.9%
6 6.7%
7 5.8%
8 5.1%
9 4.6%
4. Add a new column next to your data to look up the expected

Benford’s law percentage for your value: =INDEX([Benford Expected
%], MATCH([Amount],[Digit],0)).
5. Create a PivotTable to see the average % by user:
a. Select your data.
b. Go to Insert > Tables > PivotTable.
c. Click OK to add the PivotTable to a new sheet.
d. Drag [Employee Name] to Rows.
e. Drag [Benford Expected] to Values.
page 239
f. Click Sum of [Benford Expected] and choose Value
Field Settings.
g. Change the summarize value field by to Average, and click OK.
h. Select the [Average of Benford Expected] column in your
PivotTable, and sort it in ascending order: Go to Data > Sort & Filter
> Sort Smallest to Largest.
In IDEA
This is not possible using the built-in tool.
Drill-Down
The most modern Data Analytics software allows auditors to drill down
into specific values by simply double-clicking a value. This lets you see
the underlying transactions that gave you the summary amount. For
example, you might click the total sales amount in an income statement
to see the sales general ledger summarizing the daily totals. Click a
daily amount to see the individual transactions from that day.
Exact and Fuzzy Matching

Matching in CAAT is used to link records, join tables, and find potential
issues. Auditors use exact matching to join database tables with a
foreign key from one table to the primary key of another. In cases where
the data are inconsistent or contain user-generated information, such as
addresses, exact matches may not be sufficient. For example, “234
Second Avenue” and “234 Second Ave” are not the same value. To join
tables on these values auditors will use a fuzzy match based on the
similarity of the values. The auditor defines a threshold, such as 50
percent, and if the values share enough common characters, they will
be matched. The threshold can be higher to reduce the number of
potential matches or lower to increase the likelihood of a match.
Note that not all matches are the same. Using queries and other
database management tools, auditors may want only certain records,
such as those that match or those that don’t match. These matches
require the use of certain join types. Inner Join will show only the
records from both tables that match and exclude everything that doesn’t
match. Left Join will show all records from the first table and only
records from the second table that match. Right Join will show all
records from the second table and records from the first table that
match. Outer Join will show all nonmatching ones. Full Outer Join will
show all records, including matching and nonmatching ones. Fuzzy
matching finds matches that may be less than 100 percent matching by
finding correspondences between portions of the text or other entries.
Data needed
Two tables/sheets with a common attribute, such as a primary
key/foreign key, name, or address
In Excel
1. Search the Internet for Fuzzy Lookup Add-In for Excel, then
download and install it to your computer.
2. Open your spreadsheet with two sheets you’d like to join using a fuzzy
match. For example, employees and vendors. If your data is not
already formatted as tables in Excel, select your data and choose
Home > Styles > Format as Table. You can name the table in the
Table Design tab.
3. Go to Fuzzy Lookup > Fuzzy Lookup (Go to File > Options > Add-
ins > COM Add-ins > Go. . . and check Fuzzy Lookup Add-in for
Excel if you don’t see the bar).
page 240
page 241
a. Select the sheet you want for the Left Table and a sheet
that has similar values for the Right Table.
b. Choose the columns that you expect to find matching values in the
Left and Right Columns pane. Note: For addresses, choose
Address AND Zip Code for more likely matches.
c. Select your output columns, if needed.
d. Adjust the similarity threshold, if needed.
e. Open a new worksheet.
f. Click Go.
4. Evaluate the similarity.
In IDEA
To do a fuzzy address match, you must first join the two tables on
numbers in the address, then compare the similarity in addresses.
1. Open your workbook with two data files you’d like to join.
2. Open the first table and double click any piece of data to show the
Field Manipulation box.
3. Click Append and add a new field called “Table1 address numbers.”
4. Click the empty box in the Parameter column to open the equation
editor.
5. Use the IDEA function @JustNumbers([ADDRESS]+[ZIP]).
6. Click the green check mark. Click OK, then click Yes.
7. Repeat steps 2–6 for the second table.
8. Click on Table 1. Go to Analysis > Visual Connector to create a
relationship.
9. Add the two tables and drag new “Table1 Address Numbers” field to
the “Table2 Address Numbers” field to create a join.
0. Click OK, select Matches only, and name your dataset.
1. Click OK to show your matching data between the two tables. This only
shows exact matches based on the numbers from the addresses.
2. Double-click the data to show the Field Manipulation box.
3. Click Append and add a new field called “Distance.”
4. Set the Dec to 6, and click the Parameter box to open the equation
editor.
5. Use the IDEA function @SimilarPhrase([ADDRESS1],[ADDRESS2]).
6. Click the green check mark. Click OK, then click Yes.
7. Sort the “Distance” column in descending order. The higher the
percentage, the more similar they match.
8. Save your table.
Sequence Check
Another substantive procedure is the sequence check. This is used to
validate data integrity and test the completeness assertion, making sure
that all relevant transactions are accounted for. Simply put, sequence
checks are useful for finding gaps, such as a missing check in the cash
disbursements journal, or duplicate transactions, such as duplicate
payments to vendors. This is a fairly simple procedure that can be
deployed quickly and easily with great success. Begin by sorting your
data by identification number.
In Excel
=IF([Second Value]–[First Value]=1,"","Missing")
page 242
Stratification and Clustering

There are several approaches to grouping transactions or individuals. In
most cases, the items can be grouped by similar characteristics or
strata. With stratification, the auditor identifies specific groups, such as
geographic location or functional area, that can be used to simplify their
analysis. When similarities are less obvious, such as personal
preference or expressed behavior, clustering may be used to infer these
groupings. Both stratification and clustering are generally used for data
exploration, rather than substantive testing. The identification of these
groupings, whether obvious or not, help narrow the scope of the audit
and focus on risk. Clustering is discussed in depth in Chapter 3.
PROGRESS CHECK
5. A sequence check will help us to see if there is a duplicate
payment to vendors. Why is that important for the auditor to
find?
6. Let’s say a company has nine divisions, and each division has
a different check number based on its division—so one starts
with “1,” another with “2,” etc. Would Benford’s law work in this
situation?
CREATING ADVANCED PREDICTIVE

AND PRESCRIPTIVE ANALYTICS
LO 6-
4
Understa
nd
advance
d
predictiv
e and
prescripti
ve
analytics.
Predictive and prescriptive analytics provide less deterministic output

than the previous analytics. This is because we’re moving away from
deterministic values to more probabilistic models, judging things like
likelihood and possibility. Here we’ll briefly discuss some applications of
these different concepts, but we refer you back to Chapter 3 for
background information.
Regression
Regression allows an auditor to predict a specific dependent value
based on independent variable inputs. In other words, what would we
expect behavior to be given some inputs and does that match reality? In
auditing, we could evaluate overtime booked for workers against
productivity or the value of inventory shrinkage given environmental
factors.
Classification
Classification in auditing is going to be mainly focused on risk
assessment. The predicted classes may be low risk or high risk, where
an individual transaction is classified in either group. In the case of
known fraud, auditors would classify those cases or transactions as
fraud/not fraud and develop a classification model that could predict
whether similar transactions might also be potentially fraudulent.
There is a longstanding classification method used to predict
whether a company is expected to go bankrupt or not. Altman’s Z is a
calculated score that helps predict bankruptcy and might be useful for
auditors to evaluate a company’s ability to continue as a going concern.
When using classification models, it is important to remember that
large training sets are needed to generate relatively accurate models.
Initially, this requires significant manual classification by the auditors or
business process owner so that the model can be useful for the audit.
Probability
When talking about classification, the strength of the class can be
important to the auditor, especially when trying to limit the scope (e.g.,
evaluate only the 10 riskiest transactions). Classifiers that use a rank
score can identify the strength of classification by measuring the
distance from the mean. That rank order focuses the page 243
auditor’s efforts on the items of potentially greatest
significance.
Sentiment Analysis
Evaluate text (e.g., 10-K or annual report) for positive or negative
sentiment to predict positive or negative outcomes or to look for
potential bias on management’s part. There is more discussion on
sentiment analysis in Chapter 8.
Applied Statistics
Additional mixed distributions and nontraditional statistics may also
provide insight to the auditor. For example, an audit of inventory may
reveal errors in the amount recorded in the system. The difference
between the error amounts and the actual amounts may provide some
valuable insight into how significant or material the problem may be.
Auditors can plot the frequency distribution of errors and use Z-scores
to hone in on the cause of the most significant or outlier errors.
Artificial Intelligence
As the audit team generates more data and takes specific action, the
action itself can be modeled in a way that allows an algorithm to predict
expected behavior. Artificial intelligence is designed around the idea that
computers can learn about action or behavior from the past and predict
the course of action for the future. Assume that an experienced auditor
questions management about the estimate of allowance for doubtful
accounts. The human auditor evaluates a number of inputs, such as the
estimate calculation, market factors, and the possibility of income
smoothing by management. Given these inputs, the auditor decides to
challenge management’s estimate. If the auditor consistently takes this
action and it is recorded by the computer, the computer learns from this
action and makes a recommendation when a new inexperienced auditor
faces a similar situation.
Decision support systems that accountants have relied upon for
years (e.g., TurboTax) are based on a formal set of rules and then
updated based on what the user decides given several choices. Artificial
intelligence can be used as a helpful assistant to auditors and may
potentially be called upon to make judgment decisions itself.
Additional Analyses
The list of Data Analytics presented in this chapter is not exhaustive by
any means. There are many other approaches to identifying interesting
patterns and anomalies in enterprise data. Many ingenious auditors
have developed automated scripts that can simplify several of the audit
tasks presented here. Excel add-ins like TeamMate Analytics provide
many different techniques that apply specifically to the audit of fixed
assets, inventory, sales and purchase transactions, etc. Auditors will
combine these tools with other techniques, such as periodically testing
the effectiveness of automated tools by adding erroneous or fraudulent
transactions, to enhance their audit process.
PROGRESS CHECK
7. Why would a bankruptcy prediction be considered
classification? And why would it be useful to auditors?
8. If sentiment analysis is used on a product advertisement,
would you guess the overall sentiment would be positive or
negative?
page 244
Summary
This chapter discusses a number of analytical techniques
that auditors use to gather insight about controls and
transaction data. These include descriptive analytics that are
used to summarize and gain insight into the data, diagnostic
analytics that identify patterns in the data that may not be
immediately obvious, predictive analytics that look for
common attributes of problematic data to help identify similar
events in the future, and prescriptive analytics that provide
decision support to auditors as they work to resolve issues
with the processes and controls.
Key Words
computer-assisted audit techniques (CAATs) (228)
Automated scripts that can be used to validate data, test
controls, and enable substantive testing of transaction details
or account balances and generate supporting evidence for
the audit.
descriptive analytics (228) Procedures that summarize

existing data to determine what has happened in the past.
Some examples include summary statistics (e.g. Count, Min,
Max, Average, Median), distributions, and proportions.
diagnostic analytics (228) Procedures that explore the

current data to determine why something has happened the
way it has, typically comparing the data to a benchmark. As
an example, these allow users to see how values compare
those from a budget, a competitor, or trend.
fuzzy matching (229) Process that finds matches that may

be less than 100 percent matching by finding
correspondences between portions of the text or other
entries.
monetary unit sampling (MUS) (234) Allows auditors to

evaluate account balances. MUS is more likely to pull
accounts with large balances (higher risk and exposure)
because it focuses on dollars, not account numbers.
predictive analytics (228) Procedures used to generate a

model that can be used to determine what is likely to happen
in the future. Examples include regression analysis,
forecasting, classification, and other predictive modeling.
prescriptive analytics (228) Procedures that model data to
enable recommendations for what should be done in the
future. These typically include developing more advanced
machine learning and artificial intelligence models to
recommend a course of action based on a current problem.
ANSWERS TO PROGRESS CHECKS

1. Descriptive activity summarizes activity by computing basic
descriptive statistics like means, medians, minimums,
maximums, and standard deviations. Diagnostic analytics
compares variables or data items to each other and tries to
find co-occurrence or correlation to find patterns of interest.
Both of these approaches look at historic data. An auditor
might use descriptive analytics to understand what they
are auditing and diagnostic analytics to determine whether
there is risk of misstatement based on the expected value
or why the numbers are they way they are.
2. Use of a dashboard to highlight and communicate findings
will help identify alarms for issues that are occurring on a
real-time basis. This will allow issues to be addressed
immediately.
3. By computing minimum values or by sorting, you can find
the lowest reported value and, thus, potential negative
numbers that might have been entered erroneously into
the system and require further investigation.
4. Monetary unit sampling is more likely to pull accounts with
large balances (higher risk and exposure) because it
focuses on the amount of the transaction rather than giving
each transaction an equal chance. The larger dollar value
of the transaction, the more likely it is to affect materiality
thresholds.
page 245
5. Duplicate payments to vendors suggest that there is a gap
in the internal controls around payments. After the first
payment was made, why did the accounting system allow a
second payment? Were both transactions authorized?
Who signed the checks or authorized payments? How can
we prevent this from happening in the future?
6. Benford’s law works best on naturally occurring numbers. If
the company dictates the first number of its check
sequence, Benford’s law will not work the same way and
thus would not be effective in finding potential issues with
the check numbers.
7. Bankruptcy prediction predicts two conditions for a
company: bankrupt or not bankrupt. Thus, it would be
considered a classification activity. Auditors are required to
assess a client’s ability to continue as a going concern and
the bankruptcy prediction helps with that.
8. Most product advertisements are very positive in nature
and would have positive sentiment.

1. Which items would be currently out of scope for Data
Analytics?
a. Direct observation of processes
b. Evaluation of time stamps to evaluate workflow
c. Evaluation of phantom vendors
d. Duplicate payment of invoices
2. What would be the sampling interval if we are using a
manual approach to monetary unit sampling for a book
value of $2,000,000 and a sample size of 200?
a. 10,000
b. 1,000
c. 100,000
d. Cannot be determined
3. Monetary unit sampling is more likely to:
a. sample accounts with smaller balances.
b. sample accounts with less risk.
c. sample accounts with larger balances.
d. sample accounts with more risk.
4. The determinants for sample size include all of the
following except:
a. confidence level.
b. tolerable misstatement.
c. potential risk of account.
d. estimated misstatement.
5. CAATs are automated scripts that can be used to validate
data, test controls, and enable substantive testing of
transaction details or account balances and generate
supporting evidence for the audit. What does CAAT stand
for?
a. Computer-aided audit techniques
b. Computer-assisted audit techniques
c. Computerized audit and accounting techniques
d. Computerized audit aids and tests
page 246
6. Which type of audit analytics might be used to
find hidden patterns or variables linked to abnormal
behavior?
a. Prescriptive analytics
b. Predictive analytics
c. Diagnostic analytics
d. Descriptive analytics
7. What describes finding correspondences between at least
two types of text or entries that may not match perfectly?
a. Incomplete linkages
b. Algorithmic matching
c. Fuzzy matching
d. Incomplete matching
8. Which testing approach would be used to predict whether
certain cases should be evaluated as having fraud or no
fraud?
a. Classification
b. Probability
c. Sentiment analysis
d. Artificial intelligence
9. Which testing approach would be useful in assessing the
value of inventory shrinkage given multiple environmental
factors?
a. Probability
b. Sentiment analysis
c. Regression
d. Applied statistics
10. What type of analysis would help auditors find missing
checks?
a. Sequence check
b. Benford’s law analysis
c. Fuzzy matching
d. Decision support systems
1. How do nature, extent, and timing of audit procedures help
us identify when to apply Data Analytics to the audit
process?
2. When do you believe that Data Analytics will add value to
the audit process? How can it most help?
3. Using Table 6-2 as a guide, compare and contrast
predictive and prescriptive analytics. How might these be
used in an audit? Or a continuous audit?
4. Prescriptive analytics rely on models based on past actions
to suggest recommended actions for new, similar
situations. For example, auditors might review managers’
approval of new credit applications for inactive customers.
If auditors know the variables and values that were
common among past approvals and denials, they could
compare the action recommended by the model with the
response of the manager. How else might this prescriptive
analytics help auditors assess risk or test audit issues?
5. One type of descriptive analytics is simply sorting data.
Why is seeing extreme values helpful (minimums,
maximums, counts, etc.) in evaluating accuracy and
completeness and in potentially finding errors and fraud
and the like?
page 247
Problems
1. One type of descriptive analytics is age analysis. Why are
auditors particularly interested in the aging of accounts
receivable and accounts payable? How does this analysis
help evaluate management judgment on collectability of
receivables and potential payment of payables? Would a
dashboard item reflecting this aging be useful in a
continuous audit?
2. One of the benefits of Data Analytics is the ability to see
and test the full population. In that case, why is sampling
(even monetary sampling) still used, and how is it useful?
3. What does a Z-score greater than 3.0 (or −3.0) suggest?
How is that useful in finding extreme values? What type of
analysis should we do when we find extreme or outlier
values?
4. What are some patterns that could be found using
diagnostic analysis? Between which types of variables?
5. In a certain company, one accountant records most of the
adjusting journal entries at the end of the month. What type
of analysis could be used to identify that this happens and
the cumulative size of the transactions that the one
accountant records? Is this a problem or if not, when would
it be?
6. Which distributions would you recommend be tested using
Benford’s law? What would a Benford’s law evaluation of
sales transaction amounts potentially show? What would a
test of vendor numbers or employee numbers show?
Anything different from a test of invoice or check numbers?
Any cases where Benford’s law wouldn’t work?
7. How could artificial intelligence be used to help with the
evaluation of the estimate for the allowance for doubtful
accounts? Could past allowances be tested for their
predictive ability that might be able to help set allowances
in the current period?
8. How do you think sentiment analysis of the 10-K might
assess the level of bias (positive or negative) of the annual
reports? If management is too positive about the results of
the company, can that be viewed as being neutral or
impartial?
page 248
Lab 6-1 Evaluate the Master Data for

Interesting Addresses
You’re starting to make a name for yourself in the internal
audit department. Your manager liked your analysis of the
audit plan and now would like you to see what other ways
data analytics could be applied beyond the existing audit
action sheets.
As you’ve been reading about risk and fraud, you learned
that one common risk is that employees may be tempted to
create fictitious suppliers that they use to embezzle money.
The premise is simple enough. An employee with access to
create master data adds a supplier record for a spouse. She
then submits an invoice for “cleaning services” that were
never performed and is promptly paid, assuming there isn’t
good follow-up from the accounts payable department. The
employee is smart enough to know that an exact address
would raise red flags, so she alters it slightly to avoid
detection. Other suspicious addresses may include PO Box
addresses because they can obscure the identity of a
fictitious supplier.
You know that one way to detect this issue is to look for
fuzzy matches, and you’re eager to show your manager what
you know. Refer to Lab 3-2 for another example. This lab
assumes you have completed Lab 5-3.
Techniques
Data preparation
Filtering
Fuzzy matching
Software needed
Excel

Part 2: Master the employee and vendor data.
Part 3: Perform the analysis.
Part 4: Address the results.

Q1. Given what you know about vendor addresses,
what types of addresses would be the most
suspicious?
Q2. How could a vendor be added to an enterprise

system with a suspicious address?
Part 2: Master the Employee and Vendor

Data
In Excel
1. Open OneDrive and navigate to your Current Audit File
folder.
2. Create a new Excel workbook and call it User-Supplier
Match.
3. Rename the Sheet1 to Users and add a new sheet called
Suppliers.
4. Return to your OneDrive tab and open the User_Listing
and Supplier_Listing files.
5. Copy the data from the User_Listing file to the Users sheet
in your new spreadsheet.
6. Copy the data from the Supplier_Listing file to the
Suppliers sheet in your new spreadsheet.
7. From your User-Supplier Match spreadsheet, click Open
in Desktop App.
In IDEA
1. Download the P2P IDEA Audit Data from Connect, as
directed.
2. Unzip the file on your computer.
page 249
3. Open IDEA and go to Home > Projects >
Select.
4. Click the External Projects tab, then navigate to your
downloaded P2P IDEA Audit Data project folder.
5. Click OK.
6. Click the + to expand the project folder

In Excel
There are a couple ways to look for suspicious addresses.
You could look for specific values or use tools to help you link
records.
1. Begin by narrowing down addresses with the word “box.”
This should include “PO Box,” “P.O. Box,” and “Box.”
a. Select the data in the Supplier sheet, and format it as a
table of your choosing (Home > Styles > Format as
Table).
b. Click the drop-down arrow next to the
Supplier_Physical_Street_Address1 field to show the
sort and filter menu.
c. Choose Text Filters > Contains. . .
d. Enter box and click OK.
In IDEA
1. Open your Supplier_Listing table.
2. Go to Data > Search > Search.
a. Text to find: box
b. Fields to look in:
SUPPLIER_PHYSICAL_STREET_ADDRESS1
c. Click OK.
Q3. How many PO Box addresses appear?

Q4. Why should you follow up on PO Box addresses?
Now let’s look for fuzzy matches.
In Excel
1. Click the drop-down arrow next to the Address field, and
choose Clear Filter From
“Supplier_Physical_Street_Address1”.
2. Perform a fuzzy match on the
Supplier_Physical_Street_Address1, and
Supplier_Physical_ZipPostalCode from the Suppliers
sheet and the User_Physical_Street_Address1 and
User_Physical_Street_ZipPostalCode from the Users
sheet. Refer to the example in Chapter 6 or Lab 3-1 for
specific step-by-step instructions.
3. Take a screenshot (label it 6-1E).
In IDEA
IDEA doesn’t support fuzzy matching directly, but this works
with a few steps by merging the supplier and user tables and
then looking for fuzzy duplicate records. The resulting table
will show duplicate records that will match despite not being
exact.
1. Open the Supplier_Listing table.
2. Click Data > Fields > Append.
a. Field name: TYPE
b. Field type: Virtual Character
c. Length: 20
page 250
d. Parameter: “Supplier”
e. Click OK.
3. Open the User_Listing table.
4. Click Data > Fields > Append.
a. Field name: TYPE
c. Length: 20
d. Parameter: “Employee”
e. Click OK.
5. Go to Analysis > Relate > Append.
a. Click the + to expand the project folder.
b. Double-click SUPPLIER_LISTING.
c. Click OK.
6. Go to Data > Fields > Append.
a. Field name: COMBO_ADDRESS
c. Length: 100
d. Parameter: =
SUPPLIER_PHYSICAL_STREET_ADDRESS1 +
USER_PHYSICAL_STREET_ADDRESS1
e. Click OK.
7. In your new Append Databases table, click Analysis >
Explore > Duplicate Key > Fuzzy.
a. Output: Fuzzy matches
b. Similarity degree (%): Adjust as needed
c. Key: COMBO_ADDRESS
d. Click OK.
8. Take a screenshot (label it 6-1F).
Q5. How many fuzzy matches appeared?
Q6. Which of the matches are suspicious?
Q7. Which of the matches are normal?
Part 4: Address the Results

Q8. Are there any limitations to the way you just
evaluated addresses?
Q9. What other data values would indicate that there
may be fictitious suppliers in the system?
End of Lab
Lab 6-2 Perform Substantive Tests of

Account Balances
Account balances do not exist in databases. Rather, they are
the combination of data elements that are added together to
come up with a total through queries and formulas. The
balance for accounts receivable, for example, is the
combination of sales orders, cash receipts, and credit
memos.
As an internal auditor, you have been tasked with
validating the balance in accounts receivable. Your audit
manager has given you a list of receivables for comparison.
Additionally, company policy states that accounts receivables
should be collected within 60 days of the sale. To test this
policy, you have been asked to perform an aging of
outstanding accounts.
Techniques
Use Excel tools to calculate account balances and to group
accounts by age
page 251
Software
Excel

Part 2: Master the sales order and receipt data.

Q1. What data do you need to calculate the account
balances?
Q2. What is the formula needed to compute the

balance in accounts receivable?
Q3. How would you compute and group the age of
each receivable?

1. To address the question of the data needed, we will
compute the accounts receivable for each customer. That
is, as of 9/30/2019, how many customers have yet to pay
the amount they owe?
2. Open the File SlainteAging-Sept.xlsx.
3. Create a PivotTable using the Sales_Order data. Ensure
that the PivotTable will use the Internal Data Model so that
you can retrieve fields from both of the tables in the
spreadsheet by placing a check mark next to Add this
data to the Data Model in the Create PivotTable window. If
you need a refresher on working with PivotTables in Excel,
refer to Appendix C of the textbook.
4. In the PivotTable Fields window, click All to view both

tables in the workbook.
page 252
5. Create a PivotTable that shows the
Sales_Order_Total and the Receipt_Amount for each
Sales_Order_ID.
6. The data will look odd at first, and you will be prompted to
create relationships. You can allow Excel to auto-detect the
relationships, and it will identify the relationship between
the Primary and Foreign Keys that exist between the two
tables.
Q4. What is the primary and foreign key that relates

the two tables in this workbook?
7. After creating the relationships, the top few records of your
PivotTable output should look like the following:
page 253
8. Copy the data in the PivotTable to a new
spreadsheet to convert the PivotTable data to a range.
Doing so will allow us to be able to identify which of the
invoices have yet to be paid in full yet. You can ensure that
you’re copying only the range by selecting and copying all
of the data in the PivotTable, except for the last row
containing the Grand Total.
9. Add a column to your new range, and calculate the
difference between the Sales_Order_Total and the
Receipt_Amount.
10. Add a filter to the Difference column, and filter out all
values that appear as 0s. This will allow you to view all of
the invoices that haven’t been paid in full yet.
11. This data can be made more interesting by identifying how

late the payments are. Return to the Cash_Received
spreadsheet in your workbook.
12. Add a new column to the Cash_Received table called
Sales_Order_Date. This will allow you to easily compare
the date of the original Sales Order to the date of the
payment.
13. Use a False VLookup formula to look up the date that
corresponds with the Sales_Order_ID that each cash
receipt corresponds to. Hint: Your lookup value is the
Sales_Order_ID (FK), your table_array is the Sales_Order
table, and you want the data in column 2 (the
Sales_Order_Date) to return.
14. Now that you have the Sales_Order_Date easily

accessible, you can create another column to calculate the
difference between the dates. Create a new column
labeled Age, and subtract the Cash_Receipt_Date from
the Sales_Order_Date.
15. Your next step is to create a True VLookup formula to
assign each cash receipt to an aging bucket. Create an
aging table with the following information somewhere on
your spreadsheet:
0 0–30
30 31–60
60 61–90
90 90+
page 254
16. Add another new column to the
Cash_Received table labeled Bucket, and create a True
VLookup formula to identify the bucket for each invoice.
17. We can quickly create a summary of how many invoices

fall into each bucket using Excel’s COUNTIF function. In
the column to the right of your aging table, create a
column labeled Count.
18. In the cell to the right of your 0–30 bucket, type the
COUNTIF function. COUNTIF requires two arguments,
range and criteria. The range in this case is the bucket
column. The criteria is 0–30. COUNTIF will count every
instance of 0–30 in the bucket column.
19. Repeat the steps for the remaining three buckets. The top
two records in the Count column should return the
following data:
20. Return to your PivotTable, and refresh the data so that you
can pull in your new fields for further analysis. You can
refresh your data by clicking the Refresh button in the
Analyze tab from the ribbon.
21. You should now be able to add the bucket field to the
PivotTable. Do so. Remove Sales_Order_Total from the
PivotTable as well.
page 255
22. Collapse the fields so that you do not see the
detail of each invoice within the buckets, but only the
totals. The top two records of the PivotTable will appear as
the following:
23. Save your file.
Q5. Of the not yet collected balances in each of the

four buckets, which bucket is least likely to be
collected? Which bucket is most likely to be
collected? How would this help us come up with
an allowance for doubtful accounts?
24. Now, let’s assume that three more months have passed.
Open up the spreadsheet SlainteAging-Dec.xlsx.
Q6. Based on what you have viewed with the
September data, what do you expect to find as far
as outstanding balances now that the year has
finished at the end of December?
page 256
25. Repeat the same steps as you did above in the new
dataset.
a. Create a PivotTable that shows the Sales_Order_Total
and Receipt_Amount for each Sales_Order_ID.
i. Remember to use the Internal Data Model and to
build relationships so that the data in your PivotTable is
accurate.
b. Create a range from your PivotTable data and calculate
the difference between the Sales_Order_Total and the
Receipt_Amount. Filter the Difference column to show
only the invoices that haven’t been paid in full yet.
c. Return to the Cash_Received table and create the
additional columns so that you can identify the aging
bucket for each invoice.
d. Create a PivotTable to identify which invoices fall into
each bucket.
26. Save your file as Lab6-2December.xslx, ensuring that the
PivotTable with buckets is included in your final
spreadsheet.
End of Lab
Lab 6-3 Finding Duplicate Payments

Companies occasionally make duplicate payments to
suppliers due to lack of internal controls, errors, or fraud. In
this lab you will analyze payment transactions to collect
evidence about whether duplicate payments have been
made to suppliers.
Technique
Search for duplicates
Software needed
Excel or IDEA

Part 2: Master the purchase order and payment data.

Q1. Before computerization or Data Analytics, how
would companies find that they had made
duplicate payments?
Q2. What data items do you need to be able to find
duplicate payments? Would the date of the
duplicate payments usually be the same or
different?

In Excel
1. Open OneDrive and go to the Current Audit Data folder.
2. Open the Payments_Made spreadsheet.
3. Click Open in Desktop App to load it in the desktop
version of Excel.
4. Select the Invoice_Reference column and choose Home
> Styles > Conditional Formatting > Highlight Cell
Rules > Duplicate Values. . ., and click OK.
page 257
5. Select all of the data, choose Home > Styles
> Format as Table, and pick a light, non-banded theme.
6. Click the drop-down next to Invoice_Reference, choose
Filter by color. . ., and select the highlight color used in
step 4.
8. Remove the filter on Invoice_Reference and repeat steps
4–6 on the Payment_Amount column.
In IDEA
1. Open the P2P IDEA Audit Data project in IDEA.
2. Open the Payments_Made table.
3. Go to Analysis > Explore > Duplicate Key > Detection.
a. Click Output duplicate records
b. Click Key, then choose INVOICE_REFERENCE from the
drop-down list and click OK.
c. Click OK.
5. Repeat steps 2-3 on the Payment_Amount column.
Q3. How many duplicate records did you locate?
Q4. What course of action would you recommend?
End of Lab

Dillard’s Store Data: Hypothesis
Testing (Part I)
Company summary
Dillard’s is a department store with approximately 330 stores
in 29 states. Its headquarters is in Little Rock, Arkansas. You
can learn more about Dillard’s by looking at
finance.yahoo.com (Ticker symbol = DDS) and the Wikipedia
site for DDS. You’ll quickly note that William T. Dillard II is an
accounting grad of the University of Arkansas and the Walton
College of Business, which may be why he shared
transaction data with us to make available for this lab and
labs throughout this text.
Data
Desktop. Directions for accessing the Remote Desktop can
be found at www.mhhe.com/richardsondaa2e. See your
instructor for login credentials. The 2016 Dillard’s data
covers all transactions over the period 1/1/2014 to
10/17/2016.
Software needed
PowerPivot add-in for Excel (available on the Remote
Desktop at the University of Arkansas). If you do not see the
PowerPivot tab on the Excel ribbon, you will need to enable
the add-in.

Test a hypothesis in Excel. Specifically, we will see if the
returns in January are greater than the rest of the year.
page 258

January returns are associated with Christmas. Most retail
establishments have fairly generous return policies in case a
gift received was the wrong size or just not the desired item.
Do retail companies have the same generous policies
throughout the year, and do customers take advantage of
them throughout the year?
Therefore, our specific question that we hope to test is
whether there a significant difference in the amount of returns
in January compared to the rest of the year.

1. Extract data from SQL Server into Excel using Excel’s Get
& Transform functionality using the following query. If you
need a refresher on how to do so, refer to Appendix F of
the textbook. In this case, we will Edit the data when you
see a preview of it.
Select Tran_Date, Tran_Type, SUM(Tran_Amt) AS Amount
From Transact
Group By Tran_Date, Tran_Type
Order By Tran_Date
This query will load all of the transactional history for
both sales and refunds, grouped by day, as well as
transactional type. The way the data are organized, all of
the dollar amounts for sales and for refunds are in the
same attribute, Tran_Amt, and the transaction type (i.e.,
Sale or Return) is differentiated with the attribute
Tran_Type. In order to create a measure based on sales
and refunds as separate values, we need to split the
Tran_Amt data into two columns, one dedicated to sales,
and one for the refund amounts. To do so, we will use the
Query Editor to transform the data by “pivoting” the
Tran_Type column.
2. From the Query Editor tool, select the Tran_Type column
and click Pivot Column from the Transform tab.
3. Change the Values Column drop-down to Amount in the

Pivot Column window, then click OK.
4. Now that the data have been transformed, you can load
them into Excel. Click Close & Load from the Home tab. It
will take a moment for all of the data (1,014 rows) to load
into Excel.
page 259
5. Create a PivotTable by clicking PivotTable
from the Insert tab on the Excel ribbon.
6. Even though you have loaded the data into Excel, you
have not added it to Excel’s Internal Data Model. To do so,
place a check mark in the box next to Add this data to the
Data Model in the Create PivotTable window.

7. To create a measure for Refunds over Purchases, select
Measures > New Measure. . . from the PowerPivot tab in
the Excel ribbon. If the PowerPivot tab is not available, you
can add it from Excel Options > Add-ins > COM Add-ins,
then place a check mark next to Microsoft Power Pivot for
Excel.
8. The new measure’s name defaults to measure 1, which

isn’t very descriptive. Because we’ll be measuring average
Transaction amount, we’ll change the name to R/P. Type
R/P over the default text.
9. The formula will auto-populate as you type; begin typing

SUM, then fill in the remainder of the measure to divide the
purchasing transactions by the refund transactions:
=sum(Query1[R])/SUM(Query1[P]).
10. At the bottom of the Measure window is an option to select
a category. The Category has no bearing on how the
measure or the KPI will work. For this measure, we’ll leave
it on the default of General. Click OK to create the
measure.
page 260
11. Now that the measure is created, it has been
added to the PivotTable Fields window. Create a
PivotTable to view only the January dates (place
Tran_Date(Month) in the filter) and days along the rows.
Use the new measure you created, R/P, as the value.
page 261
Parsing out month and day will require
placing Tran_Date in the rows column first, then removing
the Year and Quarter attributes that automatically
populate. Drag and Drop Tran_Date(month) to the filter,
and keep the Tran_Date attribute in the rows.
12. Since we want to work with data from only January, we
need to filter the PivotTable. Click the drop-down next to
the Tran_Date (Month) filter, select All to expand the
options, and select Jan. Click OK to save the filter.
This PivotTable will provide the data we need for one
part of our hypothesis test—the values from all January
dates in the database. Now we need to separate the
values from all non-January dates in the database. We’ll
do this by copying the PivotTable you just created, and
modifying the filter.
13. Select the entire PivotTable (including the Filter cells), and
copy the selection.
14. Place your cursor in cell D1, and paste the PivotTable
there.
15. Now you can modify the filter. Place a check mark in the
box next to Select Multiple Items, then scroll to the top of
the filter options to select All. Finally, scroll down to take
the check mark out of the box next to January. This will
provide the data for all transactions, except for the items
that are from January.
page 262
16. Take a screenshot of your results (label it 6-
4A).
17. To clarify the difference between the two PivotTables, you
can rename the labels that say sum of R/P in each table.
Place your cursor inside the cell with the sum of R/P label,
and type in January and Rest of the Year in its place:

To run a hypothesis test in Excel, you need to first enable the
Data Analysis ToolPak add-in. To do so, follow this menu
path: File > Options > Add-ins. From this window, select the
Go. . . button, and then place a check mark in the box next to
Analysis ToolPak. Once you click OK you will be able to
access the ToolPak from the Data tab on the Excel ribbon.
18. Click the Data Analysis button from the Excel ribbon and
select t-Test: Two-Sample Assuming Unequal
Variances. This will allow us to run a hypothesis test to
see if there are significant differences between the January
transactions and the rest of the year.
19. In the t-Test window, you will need to input your variable
ranges. For Variable 1 Range, select all of the values
except for the Grand Total that correspond with the
January PivotTable (just the values—you do not need to
select the corresponding dates).
20. Follow the same pattern for Variable 2 by selecting all of

the data that correspond with the second PivotTable’s
values, except for the Grand Total.
page 263
21. Place a check mark in the box next to Labels to ensure

that the labels for the data (January and Rest of the Year)
show up in the resulting output, and click OK.
Part 4: Address and Refine the Results

Q1. Using the p-values (or the t-statistic and critical
values), are the returns as a percentage of sales
in January greater than, less than, or the same as
the returns as a percentage of sales for the rest of
the year?
Q2. What can we conclude about returns?
Q3. Do you think most Christmas sales are returned in

January, or do they also occur in early January?
How would you modify your tests to take this into
account?
Part 5: Communicate Insights and Track

Outcomes
In Chapter 7, we’ll learn more about dashboards and ways to
communicate these results to management.
Q4. Do you think knowing the level of returns is
important to management?
Q5. Assuming management want returns information,
do you think they need this information on a daily,
weekly, or monthly basis? Due to information
overload, they can’t track everything on a daily
basis, but some information is important to
disclose frequently.
End of Lab
page 264

Dillard’s Store Data: Hypothesis
Testing (Part II—Data Visualization)
Company summary
Dillard’s is a department store with approximately 330 stores
in 29 states. Its headquarters is in Little Rock, Arkansas. You
can learn more about Dillard’s by looking at
finance.yahoo.com (Ticker symbol = DDS) and the Wikipedia
site for DDS. You’ll quickly note that William T. Dillard II is an
accounting grad of the University of Arkansas and the Walton
College of Business, which may be why he shared
transaction data with us to make available for this lab and
labs throughout this text.
Data
Desktop. Directions for accessing the Remote Desktop can
be found at www.mhhe.com/richardsondaa2e. See your
instructor for login credentials. The 2016 Dillard’s data
covers all transactions over the period 1/1/2014 to
10/17/2016.
Software needed
Tableau (available on the Remote Desktop at the University
of Arkansas)

Develop a dashboard to display returns percentages across
months and across states.
Prerequisite
Lab 6-4. This lab requires some of the skills covered in Lab
6-4 for steps 1–4. If you haven’t completed Lab 6-4, then
you can still read through the steps in that lab to see the
screenshots of the ETL process in Excel.
Lab 4-2. Some Tableau skills from Lab 4-2 are also
expected. If you haven’t completed Lab 4-2, you can still
read through the steps in that lab to learn the basics of how
to build a map and a dashboard in Tableau.

After performing a hypothesis test to determine that there is a
significant difference between January’s returns percentage
and the rest of the months in a given year, you would like to
dig further into the data to visualize that difference across
months, and also across stores and in comparison to sales
data.

1. Loading the data into Tableau from the original SQL Server
database first requires some transformation in Excel.
Extract and load the transactional and store data into
Excel’s Query Editor using the following query:
Select Tran_Date, Tran_Type, State, Store.Store,
SUM(Tran_amt) AS Amount
From Transact
Inner Join Store
On Transact.Store = Store.Store
Group By Tran_Date, Tran_Type, State, Store.Store
Order by Tran_Date
page 265
2. Pivot the Tran_Type column on the Amount
values in the Query Editor window.
3. Close and load the data into Excel.
4. Once the data have loaded (298,527 rows), save the
spreadsheet as Lab 6-6.xlsx.
5. Open Tableau, and connect to an Excel Data source.
Browse and open the file you just saved. Now that the data
are loaded into their final destination for analysis, Tableau,
you have one more step to prepare the data. You need to
create the Returns Percentage measure, just like it had to
be created in Lab 6-4 in Excel.
6. On Sheet 1, create a Calculated Field. Right-click in the
Measures Area, and select Create Calculated Field.
7. Name your new field R/P, and create the calculation

SUM([R])/SUM([P]), then click OK.

8. We will create three data visualizations to display on a
returns dashboard. First, we will create a map displaying
the sales dollars per state.
9. Drag and drop the state dimension into the middle of the
Tableau drawing to start your map.
10. Double-click the P measure to display customer
purchases. Tableau will default to a symbol map, so
change this to a filled map using the Show Me window.
page 266
11. Name the sheet Sales by State and take a
screenshot (label it 6-5A).
12. Right-click your new Sales by State sheet, and click
Duplicate to start a new sheet with this map as a base.
13. In the new sheet, drag the P measure out of the Marks
card, and replace it with the calculated measure that you
created, R/P. Drag R/P onto color in the Marks card so
that the map has the correct shading.
14. Name the sheet Returns Percentage by State and take a
screenshot (label it 6-5B).
15. Open a new sheet. Drag Tran_Date to the rows. It will
default to Years. but you can expand the pill twice to see
Quarters, then Months. Remove the Years and Quarters
pills so that only the Months remains.
16. Double-click R/P so that it appears in the Marks card.
17. In the Show Me tab, replace the tabular data that Tableau
defaults to with a highlight table. This may have caused
Tableau to change your Months pill from the Rows to the
Columns. You can just drag the Months pill back down to
Rows.
18. Create a new dashboard. Arrange the three visualizations
in whichever way you find most visually pleasing and
easiest to read and take a screenshot (label it 6-5C).
19. Using the small filter button in the top right of the Returns
Percentage by State visual on the dashboard, designate
that visual as a filter for the entire dashboard. Now, you
can click any of the states on that map to focus on that
state in the sales map, as well as to see how the Returns
Percentages differ monthly for that particular state.
Part 4: Address and Refine the Results

Q1. What does getting the detail data (or drilled-down
data) allow you to test and see? Which of these
detail data would be most useful for
management?
Q2. Why would it be useful to get return data by
product code or product category? Would that
cause the company to change its return policy for
certain items?
Q3. What other data visualizations would be
meaningful to drill down into these data?
End of Lab
page 267
page 268
Chapter 7
Managerial Analytics

This chapter explains how to apply Data Analytics to measure
performance and answer managerial accounting questions. By
measuring past performance and comparing it to targeted goals, we
are able to assess how well a company is working toward a goal.
Also, we can determine required adjustments to how decisions are
made or how business processes are run, if any.
A Look Back
In Chapter 6, we focused on substantive testing within the audit
setting. We highlighted discussion of the audit plan, and account
balances were checked. We also highlighted the use of statistical
analysis to find errors or fraud in the audit setting. In addition, we
discussed the use of clustering to detect outliers and the use of
Benford’s analysis.
A Look Ahead
In Chapter 8, we will focus on how to access and analyze financial
statement data. Through analysis of ratios and trends we identify
how companies appear to stakeholders. We also discuss how to
analyze financial performance, and how visualizations help find
insight into the data. Finally, we discuss the use of text mining to
analyze the sentiment in financial reporting data.
page 269
For years, Kenya Red Cross had attempted to refine its strategy and align its daily
activities with its overall strategic goals. It had annual strategic planning meetings
with external consultants that always resulted in the consultants presenting a new
strategy to the organization that the Red Cross didn’t have a particularly strong buy-
in to, and the Red Cross never felt confident in what was developed or what it would
mean for its future. When Kenya Red Cross went through a Data Analytics–backed
Balanced Scorecard planning process for the first time, though, it immediately felt like
its organization’s mission and vision was involved in the strategic planning and that
“strategy” was no longer so vague. The Balanced Scorecard approach helped the
Kenya Red Cross align its goals into measurable metrics. The organization prided
itself on being “first in and last out” but hadn’t actively measured its success in that
goal, nor had the organization fully analyzed how being the first in and last out of
disaster scenarios affected other goals and areas of its organization. Using Data
Analytics to refine its strategy and assign measurable performance metrics to its
goals, Kenya Red Cross felt confident that its everyday activities were linked to
measurable goals that would help the organization reach its goals and maintain a
strong positive reputation and impact through its service. Exhibit 7-1 gives an
illustration of the Balanced Scorecard at the Kenya Red Cross.
EXHIBIT 7-1
The Kenya Red Cross Balanced Scorecard
Source: Reprinted with permission from Balanced Scorecard Institute, a Strategy

Management Group company. Copyright 2008–2017.
OBJECTIVES
LO 7-1 Understand management accounting questions.

LO 7-2 Evaluate management requirements and identify
useful KPIs.
LO 7-3 Evaluate underlying management data quality.
LO 7-4 Understand how to drill-down and follow up with
results.
page 270
IDENTIFYING MANAGEMENT
ACCOUNTING QUESTIONS
LO 7-
1
Understa
nd
manage
ment
accounti
ng
question
s.
In the past six chapters, you learned how to apply the IMPACT
model to data analysis projects in general and, specifically, to
internal and external auditing and financial statement analysis. The
same accounting information used in internal and external auditing
and financial statement analysis can also be used to determine how
closely an organization is meeting its strategic objectives. Together
with operational and performance measurement data, we can better
determine the gaps in actual company performance and targeted
strategic objectives, data should be condensed into easily digestible
and useful digital dashboards, providing precisely the information
needed to help make operational decisions that support a company’s
strategic direction.
This chapter brings us to how to apply Data Analytics to measure
performance. More specifically, we measure past performance and
compare it to targeted goals to assess how well a company is
working toward a goal. In addition, we can determine required
adjustments to how decisions are made or how business processes
are run, if any.
Management accounting is one of the primary areas where Data
Analytics helps the decision-making process. From assigning costs
to jobs, processes, and activities; to understanding cost behavior
and relevant costs in decisions; and to forecasting and performance
evaluation, managers rely on real-time data to evaluate the
effectiveness of their strategies. These data help with the planning,
management, and controlling of firm resources.
Managers rely on a combination of descriptive analytics to
compute the results of an initiative, diagnostic analytics to compare
those results to a benchmark (such as a budget), predictive analytics
to plan for future periods, and prescriptive analytics to guide the
controlling process.
Relevant Costs
Most other management decisions rely on the interpretation of cost
classification and which costs are relevant or not. Aggregating the
total costs of, say, the cost to produce and item versus the cost to
purchase them in a make-or-buy or outsourcing decision may be an
appropriate use of descriptive analytics, as would determining
capacity to accept special orders or processing further.
Relevant costs relates to relevant data, similar to the scope of
and audit. Managers understand that companies are collecting a lot
of data, and there is a push to find patterns in the data that help
identify opportunities to connect with customers and better evaluate
performance. However, not all data are relevant to the decision-
making process. The more relevant data that are available to inform
the decision and include in the relevant costs, the more confident
management can be of the answer. Of course, there is always a
trade-off between the cost of collecting that information and the
incremental value of the analysis. Be careful not to include the sunk
cost of data that has already been collected while considering the
opportunity cost of not utilizing data to make profitable business
decisions.
Key Performance Indicators and Variance
Analysis
Because data are increasingly available and affordable for
companies to access and store, and because the growth in
technology has created robust and affordable business intelligence
tools, data and information are becoming the key components for
decision making, replacing limited analysis and complementing
management’s intuition. Specifically, various measures and metrics
are defined, compiled from the data, and used for decision making.
Performance metrics are, rather simply, any number used to
measure performance at a company. The amount of inventory on
hand is a metric, and that metric gains meaning when compared to a
baseline (e.g., how much inventory was on hand page 271
yesterday?). A specific type of performance metric is a
key performance indicator (KPI). Just like any performance metric,
a KPI should help managers keep track of performance and strategic
objectives, but the KPIs are performance metrics that stand out as
the most important—that is, “key” metrics that influence decision
making and strategy. Nearly every organization can use data to
create the same performance metrics (although, of course, with
different results), but it is dependent upon each organization’s
particular strategy which performance metrics that organization
would deem to be a KPI.
Variance analysis allows managers to evaluate the KPIs and how
far they vary from the expected outcome. For example, managers
compare actual results to budgeted results to determine whether a
variance is favorable or unfavorable, similar to that shown in Exhibit
7-2. The ability to use these types of bullet charts to not only identify
the benchmark, but also to see the relative distance from the goal
helps managers identify root causes of the variance (e.g., the price
we pay for a raw material or the increased volume of sales) and drill-
down to determine the good performance to replicate and the poor
performance to eliminate.
EXHIBIT 7-2
Variance Analysis Identifies Favorable and Unfavorable Variances
Cost Behavior
Managers must also understand what is driving the costs and profits
to plan for the future and apply to budgets or use as input for lean
accounting processes. For example, they must evaluate mixed costs
to predict the portion of fixed and variable costs for a given period.
Predictive analytics, such as regression analysis, might evaluate
actual production volume and total costs to estimate the mixed cost
line equation, such as the one shown in Exhibit 7-3.
EXHIBIT 7-3
Regression Analysis of Mixed Costs
This example was calculated using a scatter plot chart over a 12-
month period in Excel. The mixed costs can be interpreted as
consisting of fixed costs of approximately $181,480 per month (the
intercept) and variable costs of approximately $13.30 per unit
produced. The R2 value of 0.84 tells us that this line fits the data
pretty well and will predict the correct value 84 percent of the time.
page 272
Regression and other predictive techniques help managers

identify outliers, anomalies, and poor performers so they can act
accordingly. They also rely on more observations so the prediction is
much more accurate than other rudimentary accounting calculations,
such as the High-Low method. These same trend analyses inform
the master budget from sales to cash and can be combined with
sensitivity or what-if analyses to predict a range of values.
PROGRESS CHECK
1. If a manager is trying to decide whether to discontinue a
product or division, he or she would look at the contribution
margin of that object. What are some examples of relevant
data that would be useful in this calculation? Irrelevant
data?
2. A bullet chart (as shown in Exhibit 7-2) uses a reference
line to show actual performance relative to a benchmark.
What advantages does a bullet graph have over a gauge,
such as a fan with red, yellow, and green zones and a
needle pointing to the current value?
BALANCED SCORECARD AND KEY

PERFORMANCE INDICATORS
LO 7-
2
Evaluate
manage
ment
requirem
ents and
identify
useful
KPIs.
As you will recall from Chapter 4, the most effective way to

communicate the results of any data analysis project is through data
visualization. A project in which you are determining the right KPIs
and communicating them to the appropriate stakeholders is no
different. One of the most common ways to communicate a variety of
KPIs is through a digital dashboard. A digital dashboard is an
interactive report showing the most important metrics to help users
understand how a company or an organization is performing. There
are many public digital dashboards available; for example, the
Walton College of Business at the University of Arkansas has an
interactive dashboard to showcase enrollment, where students are
from, where students study abroad, student retention and graduation
rates, and where alumni work after graduation
(https://walton.uark.edu/osie/reports/data-dashboard.php). The
public dashboard detailing student diversity at the Walton College
can be used by prospective students to learn more about the
university and by the university itself to assess how it is doing in
meeting goals. If the university has a goal of increasing gender
balance in enrollment, for example, then monitoring the “Diverse
Walton” metrics, pictured in Exhibit 7-4, can help the university
understand how it is doing at reaching that goal.
page 273
EXHIBIT 7-4
Walton College Digital Dashboard—Diverse Walton
Digital dashboards provide interesting information, but their value
is maximized when the metrics provided on the dashboard are used
to affect decision making and action. One iteration of a digital
dashboard is the Balanced Scorecard. The Balanced Scorecard
was created by Robert S. Kaplan and David P. Norton in 1996 to
help companies turn their strategic goals into action by identifying
the most important metrics to measure, as well as identifying target
goals to compare metrics against.
The Balanced Scorecard is comprised of four components:
financial (or stewardship), customer (or stakeholder), internal
process, and organizational capacity (or learning and growth). As
depicted in Exhibit 7-5, the measures in each category affect other
categories, and all four should be directly related to the strategic
objectives of an organization.
EXHIBIT 7-5
Components of the Balanced Scorecard
For each of the four components, objectives, measures, targets,
and initiatives are identified. Objectives should be aligned with
strategic goals of the organization, measures are the KPIs that show
how well the organization is doing at meeting its objective, and
targets should be achievable goals toward which to move the metric.
Initiatives should be the actions that an organization can take to
move its specified metrics in the direction of their stated target goal.
Exhibit 7-6 is an example of different objectives that an organization
might identify for each component. You can see how certain
objectives relate to other objectives—for example, if the organization
increases process efficiency (in the internal process component
row), that should help with the objective of lowering cost in the
financial component row.
EXHIBIT 7-6
An Example of a Balanced Scorecard
Reprinted with permission from Balanced Scorecard Institute, a
Strategy Management Group Company. Copyright 2008–2017.
Understanding how the four components interact to answer

different types of questions and meet different strategic goals is
critical when it comes to identifying the right measures to include in
the dashboard, as well as using those measures to help with
decision making. Creating a Balanced Scorecard or page 274
any type of digital dashboard to present KPIs for
decision making follows the IMPACT model.
Bernard Marr identified 75 KPIs to measure performance in the
different components that he considers the most important for
decision makers to know, and these 75 KPIs are compiled in Exhibit
7-7. In a Balanced Scorecard, each component should focus on 3 or
4 KPIs. Including all 75 of these metrics in a given dashboard would
be overwhelming and difficult to manage, but depending on the
strategy of the company and the initiatives that are chosen as focal
points, any of the KPIs in Exhibit 7-7 may be optimal for measuring
(and ultimately improving) performance.
page 275
EXHIBIT 7-7
Suggested KPIs That Every Manager Needs to Know1
Source: https://www.linkedin.com/pulse/20130905053105-64875646-the-75-kpis-every-
manager-needs-to-know
Financial Performance KPIs Operational KPIs

1. Net Profit 38. Six Sigma Level
2. Net Profit Margin 39. Capacity Utilization Rate
3. Gross Profit Margin (CUR)
4. Operating Profit Margin 40. Process Waste Level
5. EBITDA 41. Order Fulfillment Cycle Time
6. Revenue Growth Rate 42. Delivery in Full, on Time
(DIFOT) Rate
7. Total Shareholder Return
(TSR) 43. Inventory Shrinkage Rate
(ISR)
8. Economic Value Added (EVA)
44. Project Schedule Variance
9. Return on Investment (ROI)
(PSV)
10. Return on Capital Employed 45. Project Cost Variance (PCV)
(ROCE) 46. Earned Value (EV) Metric
11. Return on Assets (ROA) 47. Innovation Pipeline Strength
12. Return on Equity (ROE) (IPS)
13. Debt-to-Equity (D/E) Ratio 48. Return on Innovation
14. Cash Conversion Cycle (CCC) Investment (ROI2)
15. Working Capital Ratio 49. Time to Market
16. Operating Expense Ratio 50. First-Pass Yield (FPY)
(OER) 51. Rework Level
17. CAPEX to Sales Ratio 52. Quality Index
18. Price-to-Earnings Ratio (P/E 53. Overall Equipment
Ratio) Effectiveness (OEE)
54. Process or Machine
Downtime Level
55. First Contact Resolution
(FCR)
Customer KPIs Employee Performance KPIs

19. Net Promoter Score (NPS) 56. Human Capital Value Added
20. Customer Retention Rate (HCVA)
21. Customer Satisfaction Index 57. Revenue per Employee
22. Customer Profitability Score 58. Employee Satisfaction Index
23. Customer Lifetime Value 59. Employee Engagement Level
24. Customer Turnover Rate 60. Staff Advocacy Score
25. Customer Engagement 61. Employee Churn Rate
26. Customer Complaints 62. Average Employee Tenure
63. Absenteeism Bradford Factor
64. 360-Degree Feedback Score
65. Salary Competitiveness Ratio
(SCR)
66. Time to Hire
67. Training Return on
Investment
Marketing KPIs Environmental and Social

27. Market Growth Rate Sustainability KPIs
28. Market Share 68. Carbon Footprint
29. Brand Equity 69. Water Footprint
30. Cost per Lead 70. Energy Consumption
31. Conversion Rate 71. Saving Levels Due to
Conservation and
32. Search Engine Rankings (by
Improvement Efforts
keyword) and Click-Through
Rate 72. Supply Chain Miles
33. Page Views and Bounce Rate 73. Waste Reduction Rate
34. Customer Online Engagement 74. Waste Recycling Rate
Level 75. Product Recycling Rate
35. Online Share of Voice (OSOV)
36. Social Networking Footprint
37. Klout Score
page 276
The Balanced Scorecard is based around a company’s strategy.

A well-defined mission, vision, and set of values are integral in
creating and maintaining a successful culture. In many cases, when
tradition appears to stifle an organization, the two concepts of culture
and tradition must be separated. An established sense of purpose
and a robust tradition of service can serve as a catalyst to facilitate
successful organizational changes. A proper strategy for growth
considers what a firm does well and how it achieves it. With a proper
strategy, an organization is less likely to be hamstrung by a “this is
how we’ve always done it” mentality.
If a strategy is already developed, or after the strategy has been
fully defined, it needs to be broken down into goals that can be
measured. Identifying the pieces of the strategy that can be
measured is critical. Without tracking performance and measuring
results, the strategy is only symbolic. The adage “what gets
measured, gets done” shows the motivation behind aligning strategy
statements with KPIs—people are more inclined to focus their work
and their projects on initiatives that are being paid attention to and
measured. Of course, simply measuring something doesn’t imply
that anything will be done to improve the measure—the attainable
initiative attached to a metric indicating how it can be improved is a
key piece to ensuring that people will work to improve the measure.
PROGRESS CHECK
3. To illustrate what KPIs emphasize in “what gets measured,
gets done,” Walmart has a goal of a “zero waste future.”2
How does reporting Walmart’s waste recycling rate help the
organization figure out if it is getting closer to its goal? Do
you believe it helps the organization accomplish its goals?
4. How can management identify useful KPIs? How could
Data Analytics help with that?
MASTER THE DATA AND PERFORM

THE TEST PLAN
LO 7-
3
Evaluate
underlyin
g
manage
ment
data
quality.
Once the measures have been determined, the data that are
necessary to showcase those measures need to be identified. You
were first introduced to how to identify and obtain necessary data in
Chapter 2 through the ETL (extract, transform, and load) process. In
addition to working through the same data request process that is
detailed in Chapter 2, there are two other questions to consider
when obtaining data and evaluating their quality:
1. How often do the data get updated in the system? This will help you
be aware of how up-to-date your metrics are so that you interpret
the changes over time appropriately.
2. Additionally, how often do you need to see updated data? If the data
in the system are updated on a near-real-time basis, it may not be
necessary for you to have new updates pushed to your scorecard
as frequently. For example, if your team will assess their progress
only in a once-a-week meeting, there is no need to have a
constantly updating scorecard.
page 277
While the data for calculating KPIs are likely stored in the
company’s enterprise system or accounting information system, the
digital dashboard containing the KPIs for data analysis should be
created in a data visualization tool, such as Excel or Tableau.
Loading the data into these tools should be done with precision and
should be validated to ensure the data imported were complete and
accurate.
Designing data visualizations and selecting the right way to
express data (as whole numbers, percentages, or absolute values,
etc.) was discussed in Chapter 4. Specifically for digital dashboards,
the format of your dashboard can follow the pattern of a Balanced
Scorecard with a strategy map, or it can take on a different format.
Exhibit 7-8 shows a template for building out the objectives,
measures, targets, and initiatives into a Balanced Scorecard format.
EXHIBIT 7-8
Balanced Scorecard Strategy Map Template with Measures, Targets, and Initiatives
If the dashboard is not following the strategy map template, the

most important KPIs should be placed in the top left corner, as our
eyes are most naturally drawn to that part of any page that we are
reading.
PROGRESS CHECK
5. How often would you need to see the KPI of Waste
Recycling Rate to know if you are making progress? Any
different for the KPI of ROA?
6. Why do you think that the most important KPIs should be
shown in the top left corner of a digital dashboard?
ADDRESS AND REFINE RESULTS
LO 7-
4
Understa
nd how
to drill-
down
and
follow up
with
results.
Once the dashboard is in use, an active communication plan should

be implemented to ensure that the dashboard’s metrics are meeting
the needs of the business and the users. If there are multiple
audiences who use dashboards, then either different dashboards
should be created, or the dashboard should provide different views
and ways to filter the information so users can customize their
experience and see exactly the metrics they need for page 278
decision making and monitoring. Because dashboards
tend to be monitored on a daily (or even more frequent) basis,
communication with all of the users is imperative to ensure that the
identified metrics are appropriate and useful.
Some questions that would be helpful in determining how the
dashboard could be refined are the following:
1. Which metric are you using most frequently to help you make
decisions?
2. Are you downloading the data to do any additional analysis after
working with the dashboard, and if so, can the dashboard be
improved to save those extra steps?
3. Are there any metrics that you do not use? If so, why aren’t they
helpful?
4. Are there any metrics that should be available on the dashboard to
help you with decision making?
Checking in with the users will help to address any potential
issues of missing or unnecessary data and refine the dashboard so
that it is meeting the needs of the organization and the users
appropriately.
After the results have been refined and each user of the
dashboard is receiving the right information for decision making, the
dashboard should enter regular use across the organization. Recall
that the purpose of creating a digital dashboard is to communicate
how the organization is performing so decision makers can improve
their judgment and decisions and so workers can understand where
to place their priority in their day-to-day jobs and projects. Ensuring
that all of the appropriate stakeholders continue to be involved in
using the dashboard and continually improving it is key to the
success of the dashboard. The creation of a Balanced Scorecard or
any type of digital dashboard is iterative—just as the entire IMPACT
cycle should be iterative throughout any data analysis project—so it
will be imperative to continually check in with the users of the
dashboard to learn how to continually improve it and its usefulness.
PROGRESS CHECK
7. Why are digital dashboards for KPIs an effective way to
address and refine results, as well as communicate insights
and track outcomes?
8. Consider the opening vignette of the Kenya Red Cross.
How do KPIs help the organization prepare and carry out its
goal of being the “first in and last out”?
Summary
Management accountants must use descriptive
analytics to understand and direct activity, diagnostic
analytics to compare with a benchmark and control
costs, predictive analytics to plan for the future, and
prescriptive analytics to guide their decision process.
Relevant costs and data help inform decisions, variance
analysis and bullet graphs help determine where the
company is, and regression helps managers understand
and predict costs.
Because data are increasingly available and affordable
for companies to access and store, and because the
growth in technology has created robust and affordable
business intelligence tools, data and page 279
information are becoming the key
components for decision making, replacing gut
response.
Performance metrics are defined, compiled from the
data, and used for decision making. A specific type of
performance metrics, key performance indicators—or
“key” metrics that influence decision making and
strategy—are the most important.
One of the most common ways to communicate a
variety of KPIs is through a digital dashboard. A digital
dashboard is an interactive report showing the most
important metrics to help users understand how a
company or an organization is performing. Their value is
maximized when the metrics provided on the dashboard
are used to affect decision making and action.
One iteration of a digital dashboard is the Balanced
Scorecard, which is used to help companies turn their
strategic goals into action by identifying the most
important metrics to measure, as well as identifying
target goals to compare metrics against. The Balanced
Scorecard is comprised of four components: financial (or
stewardship), customer (or stakeholder), internal
process, and organizational capacity (or learning and
growth).
For each of the four components, objectives, measures,
targets, and initiatives are identified. Objectives should
be aligned with strategic goals of the organization,
measures are the KPIs that show how well the
organization is doing at meeting its objective, and
targets should be achievable goals toward which to
move the metric. Initiatives should be the actions that an
organization can take to move its specified metrics in
the direction of its stated target goal.
Regardless of whether you are creating a Balanced
Scorecard or another type of digital dashboard to
showcase performance metrics and KPIs, the IMPACT
model should be used to complete the project.
Key Words
Balanced Scorecard (273) A particular type of digital
dashboard that is made up of strategic objectives, as well
as KPIs, target measures, and initiatives, to help the
organization reach its target measures in line with
strategic goals.
digital dashboard (272) An interactive report showing

the most important metrics to help users understand how
a company or an organization is performing. Often
created using Excel or Tableau.
key performance indicator (KPI) (271) A particular
type of performance metric that an organization deems
the most important and influential on decision making.
performance metric (270) Any number measuring how

an organization is performing, particularly when that
measure is compared to a baseline.
ANSWERS TO PROGRESS
CHECKS
1. The contribution margin includes the revenues and
variable costs that are traceable to that division or
product. That data would be relevant. Other relevant
data may be the types of customers and sentiment
toward the product, products that are sold in
conjunction with that product, or market size. Shared or
allocated costs would not be.
2. A bullet graph uses a small amount of space to
evaluate a large number of metrics. Gauges are more
visually engaging and easier to understand, but waste a
lot of space.
3. If waste reduction is an important goal for page 280
Walmart, having a KPI and, potentially, a
digital dashboard that reports how well the organization
is doing will likely be useful in helping it accomplish its
goal. Using a digital dashboard helps an organization to
see if, indeed, it is making progress.
4. The KPIs that are the most helpful are those that are
consistent with the company’s strategy and measure
how well the company is doing in meeting its goals.
Data Analytics will help gather and report the necessary
data to report on the KPIs. The Data Analytics IMPACT
model introduced in Chapter 1—from identifying the
question to tracking outcomes—will be helpful in getting
the necessary data.
5. The frequency of updating KPIs is always a good
question. One determinant will be how often the data
get updated in the system, and the second determinant
is how often the data will be considered by those
looking at the data. Whichever of those two
determinants takes longer is probably correct frequency
for updating KPIs.
6. Because our eyes are most naturally drawn to that part
of any page that we are reading, the most important
KPIs should be placed in the top left corner.
7. By identifying the KPIs that are most important to
corporate strategy and finding the necessary data to
support them and then reporting on them in a digital
dashboard, decision makers will have the necessary
information to make effective decisions and track
outcomes.
8. As noted in the opening vignette, using Data Analytics
to refine its strategy and assign measurable
performance metrics to its goals, Kenya Red Cross felt
confident that its everyday activities were linked to
measurable goals that would help the organization
reach its goals and maintain a strong positive reputation
and impact through its service.

1. What would you consider to be financial performance
KPIs?
a. Total Shareholder Return
b. Customer Profitability Score
c. Market Growth Rate
d. Klout Score
2. What would you consider to be an operational KPI?
a. Inventory Shrinkage Rate
b. Brand Equity
c. CAPEX to Sales Ratio
d. Revenue per Employee
3. What does KPI stand for?
a. Key performance index
b. Key performance indicator
c. Key paired index
d. Key paired indicator
4. The most important KPIs should be placed in the _____
corner of the page even if we are not following a
strategy map template.
a. bottom right
b. bottom left
c. top left
d. top right
5. According to the text, which of these are page 281
not helpful in refining a dashboard?
a. Which metric are you using most frequently to help
you make decisions?
b. Are you downloading the data to do any additional
analysis after working with the dashboard, and if so,
can the dashboard be improved to save those extra
steps?
c. Are there any metrics that you do not use? If so, why
aren’t they helpful?
d. Which data are the easiest to access or least costly to
collect?
6. On a Balanced Scorecard, which is not included as a
component?
a. Financial Performance
b. Customer/Stakeholder
c. Internal Process
d. Employee Capacity
7. On a Balanced Scorecard, which is not included as a
component?
a. Financial Performance
b. Customer/Stakeholder
c. Order Process
d. Organizational Capacity
8. What is defined as an interactive report showing the
most important metrics to help users understand how a
company or an organization is performing?
a. KPI
b. Performance metric
c. Digital dashboard
d. Balanced Scorecard
9. What is defined as any calculation measuring how an
organization is performing, particularly when that
measure is compared to a baseline?
a. KPI
b. Performance metric
c. Digital dashboard
d. Balanced Scorecard
10. What would you consider to be marketing KPIs?
a. Conversion Rate
b. Six Sigma Level
c. Employee Churn Rate
d. Time to Market
1. We know that a Balanced Scorecard is comprised of
four components: financial (or stewardship), customer
(or stakeholder), internal process, and organizational
capacity (or learning and growth). What would you
include in a dashboard for the financial and customer
components?
2. We know that a Balanced Scorecard is comprised of
four components: financial (or stewardship), customer
(or stakeholder), internal process, and organizational
capacity (or learning and growth). What would you
include in a dashboard for the internal process and
organizational capacity components? How do digital
dashboards make KPIs easier to track?
3. Amazon, in the author’s opinion, has cared less about
profitability in the short run but has cared about gaining
market share. Arguably Amazon gains market share by
taking care of the customer. Given the “Suggested 75
KPIs That Every Manager Needs to Know” page 282
from Exhibit 7-7, what would be a natural
KPI for the customer aspect for Amazon?
4. For an accounting firm like PwC, how would the
Balanced Scorecard help balance the desire to be
profitable for its partners with keeping the focus on its
customers?
5. For a company like Walmart, how would the Balanced
Scorecard help balance the desire to be profitable for
its shareholders with continuing to develop
organizational capacity to compete with Amazon (and
other online retailers)?
6. Why is Customer Retention Rate a great KPI for
understanding your Tesla customers?
7. If the data underlying your digital dashboard are
updated in real time, why would you want to update
your digital dashboard in real time? Are there situations
when you would not want to update your digital
dashboard in real time? Why or why not?
8. In which of the four components of a Balanced
Scorecard would you put the Walton College’s diversity
initiative? Why do you think this is important for a public
institution of higher learning?
Problems
1. From Exhibit 7-7, choose five financial performance
KPIs to answer the following three questions. This URL
(https://www.linkedin.com/pulse/20130905053105-
64875646-the-75-kpis-every-manager-needs-to-know)
provides background information for each individual KPI
that may be helpful in understanding the individual KPIs
and answering the questions.
a. Identify the equation/relationship/data needed to
calculate the KPI. If you need data, how frequently
would the data need to be incorporated to be most
useful?
b. Describe a simple visualization that would help a
manager track the KPI.
c. Identify a benchmark for the KPI from the Internet.
Choose an industry and find the average, if possible.
This is for context only.
2. From Exhibit 7-7, choose 10 employee performance
KPIs to answer the following three questions. This URL
calculate the KPI. How frequently would it need to be
incorporated to be most useful?
3. From Exhibit 7-5, choose 10 marketing KPIs to answer
the following three questions. This URL
calculate the KPI. How frequently would it need to be
incorporated to be most useful?
4. How does Data Analytics help facilitate the use of the
Balanced Scorecard and tracking KPIs? Does it make
the data more timely? Are you able to access more
information easier or faster, or what capabilities does it
give? page 283
5. If ROA is considered a key KPI for a
company, what would be an appropriate benchmark?
The industry’s ROA? The average ROA for the
company for the past five years? The competitors’
ROA?
a. How will you know if the company is making
progress?
b. How might Data Analytics help with this?
c. How often would you need a measure of ROA?
Monthly? Quarterly? Annually?
6. If Time to Market is considered a key KPI for a
company, what would be an appropriate benchmark?
The industry’s time to market? The average time to
market for the company for the past five years? The
competitors’ time to market?
a. How will you know if the company is making
progress?
b. How might Data Analytics help with this?
c. How often would you need a measure of Time to
Market? Monthly? Quarterly? Annually?
7. Why is Order Fulfillment Cycle Time an appropriate KPI
for a company like Wayfair (which sells furniture
online)? How long does Wayfair think customers will be
ready to wait if Amazon Prime promises items delivered
to its customers in two business days? Might this be an
important basis for competition?
page 284
Lab 7-1 Evaluate Management

Requirements and Identify Useful
KPIs from a List
Key performance indicators help managers keep track of
performance and strategic objectives.

Learn about many of the key performance indicators.
Evaluate which KPIs best work for Tesla.
Consider the data needed and the desired frequency to
provide each of these KPIs.
Q1. Imagine you work for Tesla. Choose 20 KPIs
that you believe are most important to Tesla’s
management (include at least three from each
category).
The 75 KPIs Every Manager Needs to

Know (Bernard Marr)3
To measure financial 1. Net Profit

performance: 2. Net Profit Margin
3. Gross Profit Margin
4. Operating Profit
Margin
5. EBITDA
6. Revenue Growth Rate
7. Total Shareholder
Return (TSR)
8. Economic Value
Added (EVA)
9. Return on Investment
(ROI)
10. Return on Capital
Employed (ROCE)
11. Return on Assets
(ROA)
12. Return on Equity
(ROE)
13. Debt-to-Equity (D/E)
Ratio
14. Cash Conversion
Cycle (CCC)
15. Working Capital Ratio
16. Operating Expense
Ratio (OER)
17. CAPEX to Sales Ratio
18. Price-to-Earnings
Ratio (P/E Ratio)
To understand your 19. Net Promoter Score

customers: (NPS)
20. Customer Retention
Rate
21. Customer Satisfaction
Index
22. Customer Profitability
Score
23. Customer Lifetime
Value
24. Customer Turnover
Rate
25. Customer
Engagement
26. Customer Complaints
To gauge your market and 27. Market Growth Rate

marketing efforts: 28. Market page 285
Share
29. Brand Equity
30. Cost per Lead
31. Conversion Rate
32. Search Engine
Rankings (by
keyword) and Click-
Through Rate
33. Page Views and
Bounce Rate
34. Customer Online
Engagement Level
35. Online Share of Voice
(OSOV)
36. Social Networking
Footprint
37. Klout Score
To measure your 38. Six Sigma Level

operational performance: 39. Capacity Utilisation
Rate (CUR)
40. Process Waste Level
41. Order Fulfillment
Cycle Time
42. Delivery in Full, on
Time (DIFOT) Rate
43. Inventory Shrinkage
Rate (ISR)
44. Project Schedule
Variance (PSV)
45. Project Cost Variance
(PCV)
46. Earned Value (EV)
Metric
47. Innovation Pipeline
Strength (IPS)
48. Return on Innovation
Investment (ROI2)
49. Time to Market
50. First-Pass Yield (FPY)
51. Rework Level
52. Quality Index
53. Overall Equipment
Effectiveness (OEE)
54. Process or Machine
Downtime Level
55. First Contact
Resolution (FCR)
To understand your 56. Human Capital Value

employees and their Added (HCVA)
performance: 57. Revenue per
Employee
58. Employee Satisfaction
Index
59. Employee
Engagement Level
60. Staff Advocacy Score
61. Employee Churn Rate
62. Average Employee
Tenure
63. Absenteeism Bradford
Factor
64. 360-Degree Feedback
Score
65. Salary
Competitiveness Ratio
(SCR)
66. Time to Hire
67. Training Return on
Investment
To measure your 68. Carbon Footprint

environmental and social 69. Water Footprint
sustainability
performance: 70. Energy Consumption
71. Saving Levels Due to
Conservation and
Improvement Efforts
72. Supply Chain Miles
73. Waste Reduction Rate
74. Waste Recycling Rate
75. Product Recycling
Rate
page 286

For each of these 20 KPIs:
Q2. Identify the specific equation/relationship/data
needed to calculate the KPI. How frequently
do you expect data for these KPIs to be
updated (e.g., daily, monthly)?
Q3. Describe a simple visualization or dashboard
that would help a manager track the KPI.
Does it use red, yellow, and green indicators,
or do you have something else in mind that
would be better?

Q4. Identify a benchmark for five of these KPIs for
Tesla. Where would you find these
benchmarks? Would you base it on averages
for Tesla or on performance from the prior
week, month or year? For the car industry or
a different industry?
End of Lab
Lab 7-2 Create a Balanced

Scorecard Dashboard in Tableau
Superstore has brought you in to help it develop some
metrics to evaluate performance across different
dimensions of its business, including finance, customers,
process, and employee growth.
Company summary
Superstore is a large seller of retail and wholesale office
supplies, furniture, and technology. It operates in the
United States and has divided its sales regions into North,
South, East, and West. Each region has a regional sales
representative who interacts with the customers to take
orders and deal with returns.
Data
Sales order data are available for a four-year period,
including demographic data about the customers, as well
as main categories and subcategories of products. Note
that depending on the version of Tableau you use, the
screenshots may differ from the examples below.
Technique
In this lab, you will use Tableau to generate a dashboard
to evaluate four key performance indicators.
Software needed
Tableau

Generate some key performance indicators.
Evaluate the data.
Perform analyses and generate visualizations.

Your understanding of key performance indicators has
given you some insight into how management at
Superstore might measure and evaluate performance
across different aspects of the business. They depend on
your expertise to do just that.
page 287
Assuming you’ll have access to sales order and

returns data, as well as the sales representatives
involved, think about different ways you could measure
performance.
Q1. What KPIs would you consider using to
evaluate sales financial performance?
evaluate customer relationships?
evaluate process efficiency?
evaluate employee growth?
Q5. For each KPI, identify a benchmark value or
KPI goal that you think management might
use.
Part 2: Generate a Request for Data

The following data are available:
LAB TABLE 7-2A
Orders Returns People
Row ID Order ID Person
Order ID Region
Order Date
Ship Date
Ship Mode
Customer ID
Customer Name
Segment
Country City
State
Postal Code
Region Product ID
Category
Subcategory
Product Name
Sales
Quantity
Discount
Profit
Q6. Using the available fields, identify some

calculations or relationships that would
support your KPIs from Q1 to Q4.
Q7. Are there any KPIs you selected that don’t
have supporting data fields?

Data
Now you’ll use Tableau to generate some analytics that
will provide visualizations for management to quickly
evaluate some of the KPIs. To simplify the process, here
are four KPIs that management has identified as high
priorities:
Finance: Which product categories provide the

highest amount of profit? The goal is 13 percent return
on sales. Use Profit ratio = Total profit/Total sales.
page 288
Process: How long does it take to ship our product to

each state on average? Management would like to see
four days or less. Use Delivery time in days = Ship
date − Order date.
Customers: Which regions have the highest return
rates? Management says only 30 percent of sales
orders should be returned normally. Return rate =
Number of returned/Number of orders.
Employees: Who are our top-performing employees
by sales each month? Rank the total number of sales
by employee.
Now it’s your turn to build a Balanced Scorecard

dashboard in Tableau for each of these metrics. First,
you’ll create four individual worksheets; then you’ll
combine them into a dashboard for quick review.
Note: To compare actual performance to
management’s goals, you’ll need to set some parameters
and create some additional calculated fields.
Create a new project in Tableau

1. Open Tableau, and create a new Tableau book.
2. Click Data > New Data Source > Excel.
3. Navigate to Documents > My Tableau Repository >
Datasources > XX.X > en_US-US > Sample –
Superstore.xls or choose Sample – Superstore from
the saved data sources on the open data screen.
4. Click Open.
5. In Data Source, drag Orders and People to the top
pane to left join them. Then drag Returns to the
whitespace and create a left join.
6. Click Sheet 1 to begin your creating your visualizations.
Source: Tableau Software, Inc. All rights reserved.
Add parameters for management’s benchmark goals

To create parameters, in the left pane, click the down-
arrow next to Dimensions and choose Create
Parameter. . .
page 289
Tableau Software, Inc. All Rights Reserved
1. Name: KPI Target Return on Sales

a. Datatype: Float
b. Display format: Percentage, 0 decimals
c. Allowable values: Range
d. Minimum: 0.01
e. Maximum: 1
f. Step size: 0.01
g. Current value: 0.13 <- This is management’s 13
percent return on sales goal.
2. Name: KPI Target Delivery Days
a. Datatype: Float
b. Display format: Automatic
d. Minimum: 1
e. Maximum: 10
f. Step size: 0.5
g. Current value: 4 <- This is management’s four-day
shipping goal.
3. Name: KPI Target Return Rate
a. Datatype: Float
b. Display format: Percentage, 0 decimals
d. Minimum: 0
e. Maximum: 1
f. Step size: 0.05
g. Current value: 0.1 <- This is management’s 30
percent order return rate goal.
4. Name: KPI Target Top Salespeople
a. Datatype: Integer
b. Display format: Number (standard)
d. Minimum: 0
e. Maximum: 3
f. Step size: 1
g. Current value: 1 <- This shows the number of top
employees management wants to recognize.
page 290
Create the four worksheets

For simplicity, full instructions are provided for the first
sheet. For subsequent sheets, drag the attributes to the
appropriate places.
1. Create a new worksheet called Finance. In the end, it
will look similar to this:
a. Create calculated fields—click the down-arrow next to

Dimensions in the left pane and choose Create
Calculated Field. Enter the name of the new field,
then type the expression in the box below.
Tableau Software, Inc. All Rights Reserved

page 291
i. Profit Ratio: <- if you get an error that
this already exists, skip this step.
a. Name: Profit Ratio
b. Equation: SUM([Profit])/SUM([Sales])
ii. Actual vs Target
a. Name: Actual vs Target Return on Sales
b. Equation: [Profit Ratio] > [KPI Target Return on
Sales]
b. Drag the following measure to the Columns shelf:
Profit Ratio This will become AGG(Profit Ratio).
c. Drag the following dimensions to the Rows shelf:
Category, Sub-Category.
d. Drag the following dimension to the Filters shelf:
Product Name. Click the value and select Custom
Value List in the window that appears. Then click OK.
e. Drag the following measure to the Marks pane: Actual
vs Target Return on Sales becomes AGG (Actual vs
Target – Return on Sales). Click the icon next to it and
select Color from the list.
f. Click the Analytics tab in the left pane. In the
Custom section, drag Reference Line onto the
Finance table. In the window that appears, choose the
following options:
1. Table
2. Value: KPI Target Return on Sales
g. Click OK
h. Hover over the top of the bar chart and click the sort
icon that appears to place the values in descending
order by profit ratio.
i. Save your project.
j. Take a screenshot (label it 7-2A).
2. Create a new worksheet called Process. In the end,
the visualization will look similar to this.
a. Create calculated fields:

i. Delivery Time Days:
ROUND(FLOAT(DATEDIFF(’day’, [Order Date],
[Ship Date])),2)
ii. Actual vs Target Delivery: AVG([Delivery Time
Days]) < [KPI Target Delivery Days]
page 292
b. Columns: Longitude (generated)
c. Rows: Latitude (generated)
d. Type: Filled Map
e. Marks:
i. Delivery Time Days > Use the drop-down menu on
the attribute to change from Sum to Average >
Change the mark to Color
ii. Country > Detail
iii. State > Detail
f. Double-click AVG(Delivery Time Days) color scale on
the left:
i. Red-Green Diverging
ii. Reversed
iii. Advanced: Center: 4
g. Take a screenshot (label it 7-2B).
3. Create a new worksheet called Customer. In the end,
the visualization will look similar to this.

i. Return Rate: COUNT([Returned])/COUNT([Order
ID])
ii. Actual vs Target Return Rate:[Return Rate] < [KPI
Target Return Rate]
page 293
ii. Columns: YEAR(Order Date)
iii. Rows: AGG(Return Rate)
iv. Type: Line
v. Marks:
i. AGG(Actual vs Target - Return Rate) > Color
ii. Region > Label
vi. Analytics > Reference Line > Entire Table > KPI
Target Return Rate
vii. Take a screenshot (label it 7-2C).
4. Create a new worksheet called Growth. In the end, the
visualization will look similar to this.

i. Rank: Index()
ii. Actual vs Target Seller: [Rank] < = [KPI Target Top
Salespeople]
b. Columns: SUM(Sales)
c. Rows: Person
d. Type: Horizontal Bar
e. Marks:
i. Actual vs Target Seller > Color
ii. SUM(Sales) > Label
f. Pages: YEAR(Order Date) <- This will allow you to
select a year to see the top-performing seller for that
year.
g. Rows > Person > Sort > Descending
h. Take a screenshot (label it 7-2D).
page 294
5. Finally, create a new dashboard sheet
called Balanced Scorecard. In the end, the
visualization will look similar to this.
a. Drag Finance, Customer, Process, and Growth to

main body of your dashboard.
b. To enable management to adjust its goals (and
corresponding reference lines), add the parameters to
the dashboard along the side.
c. From the menu, click Analysis > Parameters, and
add each of the four parameters to the dashboard.
d. Take a screenshot (label it 7-2E).

Now that you’ve created the dashboard, take a moment to
interpret the results.
Q8. Which product categories have fallen below
the profit goal of 13 percent?
Q9. Which states are taking significantly more
than four days to ship to?
Q10. Which region(s) has (have) experienced the
highest percent return rate from customers?
Q11. Which sales representative is leading the rest
for the most recent month?
End of Lab
page 295

Dillard’s Store Data: Creating KPIs in
Excel (Part I)
Company summary
Data
To begin mastering the data, you will need to connect to
SQL Server data through Excel using Data > Get &
Transform. If you need a refresher on how to do so, refer
to Appendix F.
Once you connect to the SQL Server dataset
WCOB_DILLARDS in Excel and expand the options to
input a query, input the following SQL query to extract the
data needed for our analysis.
Select Transact.*, Store.STATE
From Transact
Inner Join Store
On Transact.Store = Store.STORE
Where TRAN_DATE BETWEEN ’20160901’ and ’20160915’
Order By Tran_Date
Software needed
Power Pivot Excel add-in. To create a date table, we’ll
extract and load the data through Power Pivot instead of
through the Get & Transform tab. If you don’t see Power
Pivot as a tab in the Excel ribbon, you will need to
activate the add-in.

Learn to build a KPI. In this case, we are trying to assess
whether sales have improved the same date a year
earlier.
Specifically, create a baseline measure in Excel and set
a target value. These two measures will be used to
create a KPI to compare sales data across two different
periods.
page 296
1. From the Remote Desktop, open Excel. From the File

tab on the ribbon, open Options.
2. Select Add-ins from the left side of the Excel Options

window.
page 297
3. From the drop-down window at the bottom
of the Add-ins screen, select COM add-ins, then click
Go. . .
4. Place a check mark in the box next to Microsoft Power

Pivot for Excel, then click OK.

Our question for this lab is whether sales from September
1 to September 15, 2016, are different (better, worse,
approximately the same) than the average sales from the
same time period in 2015.
Q1. Why would comparing current year sales to
prior year sales be useful?
Part 2: Mastering the Data and

Performing the Analysis
While you loaded the data into the spreadsheet originally
with a query from an external data source, that didn’t
automatically load it into Excel’s Internal Data Model.
Excel has a way to super-charge its conditional formatting
by creating KPIs in Power Pivot. Power Pivot is a plug-in
to Excel 2013 and 2010 and compares pre-prepared as
an add-in to Excel 2016. Because you’ll be using Excel in
Walton College’s virtual lab, you’ll have access to Excel
2016. To create KPIs in Excel, the data must be added to
the Internal Data Model.
Identify a base performance metric, and create a
measure. Measures can be implicit or explicit.
Implicit measures are measures created in a
PivotTable—any time you drag and drop a field into
the values section of the PivotTable, it becomes an
implicit measure. Implicit measures are restricted to
the value field settings’ standard aggregations
(SUM, COUNT, MIN, MAX, DISTINCTCOUNT, or
AVG). These implicit measures cannot be used to
create KPIs.
Explicit measures can be created in the Power Pivot
Data Model window or in the Excel main window
Form the Measure dialog box in the Power Pivot tab
on the Excel ribbon.
Identify a target value to compare the measure to the
baseline.
Create a KPI to signal performance of the measure in
comparison to the baseline.
page 298
5. From the Insert tab on the ribbon, click PivotTable.
6. In the Create PivotTable window, make sure to place a

check mark in the box next to Add this data to the
Data Model. Then click OK.
7. Once the PivotTable has been created (this may take a

few moments as the data are loaded into the data
model), you can create a measure and a KPI. Navigate
to the Power Pivot tab in the ribbon.
Click Measures, then select New Measure. . .
8. The new measure’s name defaults to Measure 1, which

isn’t very descriptive. Because we’ll be measuring
average Transaction amount, we’ll change the name to
AVG(Tran_Amt). Type AVG(Tran_Amt) over the default
text.
page 299
9. The formula will auto-populate as you type. Begin

typing average, and then begin typing the field
Tran_Amt to fill in the formula.
10. The category has no bearing on how the measure or

the KPI will work. For this measure, we’ll leave it on the
default of General. Click OK to create the measure.
11. If you scroll down on the PivotTable Fields window,

you will see that the explicit measure has been added
to the bottom of the field list.
12. Now we will create the KPI. In the Power page 300
Pivot tab of the ribbon, click KPIs and
select New KPI. . .
13. Because you have only one measure added to this

spreadsheet for now, the base field defaults to your
newly created measure. If you had more than one
measure, you would use the drop-down to select the
measure you wanted to use for your base field. The
target value can be defined by another measure or by
an absolute value. For this first KPI, we’ll define it by an
Absolute Value. Let’s assume that Dillard’s has a goal
of averaging at least $23 per Transaction.
Input 23 as the Absolute value for the target value.
Leave the default for the status thresholds.
page 301
Q2. Why might you want to edit the status
thresholds? Does 18.4 seem low for the upper
limit?
14. Now that you have your KPI created, you can see each
of them in the PivotTable Fields list.
If you expand the KPI fields, you see three options:
The Value (2016 Sales) will show the actual sale
totals associated with the year 2016 (or sliced by
month or day, depending on the other values you drill
into in the PivotTable).
The Goal will show 2015 sales totals—this is the
measure that you are using to compare 2016 sales
against. The Goal is for the sales to be at least 2
percent higher than the previous year’s sales.
The Status will show stoplight icons indicating red,
yellow, or green circles based on the thresholds you
selected when setting the KPI. Note: When the KPI
fields are added to the PivotTable, they are designed
to show up as stoplights (red, yellow, or green
circles, depending on the status). It is likely, however,
that the KPI status was automatically added to your
PivotTable and that the stoplight signals show as -1,
0, and 1 instead of stoplights. If you remove the
AVG(Tran_Amt) Status from the PivotTable (you can
do so by unchecking the KPI Status in the PivotTable
field list) and then place it back into the PivotTable by
replacing the check mark, it should correct the issue
and the stoplight icons will show in the PivotTable.
15. Create a PivotTable that shows the KPI status for
average Transaction by each of the 15 days in your
data range. Place TRAN_DATE in the Rows and
AVG(Tran_Amt) and AVG (Tran_Amt) Status in the
Values, as shown in the image below.
page 302
16. Take a screenshot (label 7-3A).
Q3. How did Dillard’s perform in September 2016
compared to September 2015? Do you think
the target is set too high or too low? Which
day(s) performed the worst, compared to the
same date(s) in the previous period? Why do
you think that is?
End of Lab
Excel (Part II)
Company summary
Data
your instructor for login credentials. The 2016 Dillard’s
data cover all transactions over the period 1/1/2014 to
10/17/2016.
Software needed

Compare total sales across all Dillard’s stores year over
year, month over month, and day over day and develop it
as a KPI.

Compare 2014, 2015, and 2016 sales data in parallel
periods.
page 303

1. Before we can create measures and KPIs to analyze
the data, we need to extract the data from SQL Server
and load them into Excel. To begin mastering the data,
you will need to connect to SQL Server data through
Excel using Data > Get & Transform. If you need a
refresher on how to do so, refer to Appendix F. Once
you connect to the SQL Server dataset
WCOB_DILLARDS in Excel and expand the options to
input a query, input the following SQL query to extract
the data needed for our analysis.
Select year(Tran_Date) as year, month(Tran_Date) as
month, day(Tran_Date) as day, sum(Tran_Amt) as
amount
From TRANSACT
Where TRAN_TYPE = 'P'
Group By year(Tran_Date), month(Tran_Date),
day(Tran_Date)
Order By year(Tran_Date), month(Tran_Date),
day(Tran_Date)
2. Once you have input the query and the preview of your
data is presented, click Edit to transform the data in the
Query Editor.
page 304
The data have been fully extracted
from SQL Server into Excel’s Internal Data Model, but
they need to be transformed so that we can more
easily compare daily sales amounts year over year.
Instead of seeing a separate record for each day,
beginning with January 1, 2014, and ending with
October 17, 2016, we would prefer to see only 365
records—one record for each day in a calendar year,
but with separate columns for each year (2014, 2015,
and 2016), each with the transaction amount
associated with that year’s month and day.
3. Select the year column.
4. Select Pivot Column from the Transform tab on the
Query Editor ribbon.
5. Select Amount from the drop-down for the Values

column and click OK.
6. Now that the data have been transformed, we’re ready

to load them into Excel. From the Home button on the
Query Editor’s ribbon, click Close and Load.
7. Insert a PivotTable from your new data. In the Create
PivotTable window, make sure to place a check mark
in the box next to Add this data to the Data Model.
Then click OK.
page 305
Once the PivotTable has been created
(this may take a few moments as the data are loaded
into the data model), you can create a measure and a
KPI using Excel’s Power Pivot capabilities. Excel has a
way to super-charge its conditional formatting by
creating KPIs in Power Pivot. If you do not have this
add-in added to Excel yet, see instructions in Lab 7-3
steps 1-4. KPIs require three decisions:
Identify a base performance metric, and create a
measure. Measures can be implicit or explicit.
Implicit measures are measures created in a
PivotTable—any time you drag and drop a field into
the values section of the PivotTable, it becomes an
implicit measure. Implicit measures are restricted to
the value field settings’ standard aggregations
(SUM, COUNT, MIN, MAX, DISTINCTCOUNT, or
AVG). These implicit measures cannot be used to
create KPIs.
Explicit measures can be created in the Power Pivot
Data Model window or in the Excel main window
from the Measure dialog box in the Power Pivot tab
on the Excel ribbon.
Identify a target value to compare the measure to.
Create a KPI to signal performance of the measure in
comparison to the baseline, and determine the range of
values that indicate poor performance, good
performance, and great performance.
We will need to create three measures, the sums of
each of the year’s sales Transactions.
8. Navigate to the Power Pivot tab in the ribbon. Click
Measures, then Select New Measure. . .
9. The new measure’s name defaults to Measure 1, which

isn’t very descriptive. Because we’ll be measuring
average transaction amount, we’ll change the first KPI’s
name to 2014 Sales. Type 2014 Sales over the default
text.

10. The formula will auto-populate as you type, begin typing
SUM, then fill in the parentheses with the column name
2014.
11. At the bottom of the Measure window is an option to
select a category. The Category has no bearing on how
the measure or the KPI will work. For this measure,
we’ll leave it on the default of General. Click OK to
create the measure.
page 306
12. Repeat the same steps used to create the measure for
2014 sales to create measures for 2015 sales and 2016
sales.
13. Now we will create the KPIs to compare 2015 sales to
2014, and 2016 sales to 2015. In the Power Pivot tab of
the ribbon, click KPIs and select New KPI. . .
14. The first KPI we will create is comparing 2016 sales to

the previous year’s sales. Use the drop-down to select
2016 Sales for your base field. The target value can be
defined by another measure or by an absolute value.
We have already defined the measure to compare 2016
sales to, so select 2015 Sales for the target value
Measure.
We will define excellent performance as a 2 percent
improvement over last year’s sales, so move the upper
range of the target slider to 102%. Poor performance
will be defined as a 2 percent decline from last year’s
sales. Move the lower range of the target slider to 98%.
Q1. Do you think +/– 2 percent is the right
benchmark to set? Would you propose a
different percentage change to track here?
Once all of your settings are correct, click OK to
create the KPI.
page 307
15. Create the KPI comparing 2015 sales to 2014 sales

using the same thresholds for measuring performance.
16. Now that you have your two KPIs created, you can see
each of them in the PivotTable Fields list.
page 308
Occasionally, if the KPI status is
automatically added to your PivotTable, the stoplight
signals show as –1, 0, and 1. If you uncheck and
recheck the status field from the fields list, this will
correct the issue and the stoplight icons will show.
If you expand the KPI fields, you see three options:
The Value (2016 Sales) will show the actual sale totals
associated with the year 2016 (or sliced by month or day,
depending on the other values you drill into in the
PivotTable).
The Goal will show 2015 sales totals—this is the
measure that you are using to compare 2016 sales
against. The Goal is for the sales to be at least 2 percent
higher than the previous year’s sales.
The Status will show stoplight icons indicating red,
yellow, or green circles based on the thresholds you
selected when setting the KPI.
17. Create a PivotTable that shows the KPI status of 2015
and 2016 sales by month. To do so, drag and drop
Months into the Rows and Status for both KPIs into the
Values.
page 309
If you just place a check mark in the
box next to the month field, you will notice that the
PivotTable defaults to reading Month values as
numerical data instead of calendar data, so it places it
as a value and sums the month numbers. You just
need to drag and drop month outside of Values and
into Rows.
18. Take a Screenshot (label it 7-4A).
19. To provide some drill-down capabilities, add the Day
field to the Rows (beneath Month).
Q2. Do you notice a pattern with how frequently
the “bad” (red icon) days appear in 2016 in
relation to 2015?
Q3. What do you think is the potential problem
with comparing days (e.g., comparing
September 1, 2016 to September 1, 2015)?
How could this be improved?
End of Lab

Excel (Part III)
Company summary
Data
10/17/2016.
Software needed
page 310

How do we line up sales periods to be in parallel periods,
by day of week in one period with day of week with
previous period?

1. In this lab, we will connect to the database through
PowerPivot instead of through “Get and Transform” that
we have used in previous labs. In order to do so, you
need to have the PowerPivot add-in added into Excel. If
you have not added it in before, you can follow steps 1-
4 in Lab 7-3.” To extract and load the data into Power
Pivot, click Manage on the Power Pivot tab in the
Excel ribbon.
2. In the Power Pivot for Excel window, click Get

External Data from the Home tab, then navigate
through From Database and From SQL Server.
page 311
3. The Table Import Wizard window will
open. Input the SQL Server name and the Database
name that you received from your instructor, then click
Next.

4. We will import the data with a query, so select the radio
button next to Write a query that will specify the data to
import.
page 312
5. We need to bring in only two attributes. In
Lab 7-3, we had to parse out the different date parts in
order to group our data by month and year, instead of
just by day. In this lab, we will use Excel’s Power Pivot
tool to create a Date table. The tool will be able to
parse out the date parts for us, instead of us having to
do so with our query. This will also allow us to view
more interesting date parts, such as the day of the
week (not just the date).
Input the following query into the Table Import
Wizard window to extract the total amount of
Transactions for each day in the database:
Select Tran_Date, SUM(Tran_Amt) AS Sales
From Transact
Group By Tran_Date
After entering the SQL text, click Validate to ensure
the query will run, and then click Finish.
The table will import. This may take a few moments.

page 313
6. Once the data are loaded, you can close the Table
Import Wizard window. Click Close.
7. After closing the Table Import Wizard, you will see

your data loaded into Power Pivot. This does not mean
the data have been loaded into Excel yet, so you can
transform the data within the Power Pivot tool first.
Creating the date table takes three steps: Select the
Tran_Date column, click Date Table from the Design
tab on the ribbon, then click New.

You have created a Date table. Now it’s time to load
the transformed data into Excel.
8. Return to the Home tab on the Power Pivot ribbon,
and select PivotTable.
page 314
9. Select OK to create the PivotTable in a
New Worksheet.
The PivotTable Fields list contains two tables,

Calendar and Query. The Calendar table contains the
Date Hierarchy for drilling down, but it also contains
attributes beneath the More Fields title. These contain
the same attributes in the hierarchy, as well as different
ways of viewing the data, such as Day of Week. The
Query table contains the data that you extracted with your
SQL query. The valuable field from the query table is
Sales, which you will use as a value (or an implicit
measure).

Data
10. Create a PivotTable to compare sales performance on
different weekdays of each month, year over year. To
do so, drag and drop year (from the Calendar > More
fields drop-down) into Columns, Month and
DayofWeek into Rows, and Sales into Values). The
Sales data will be transformed into a measure, Sum of
Sales, automatically. Note: If a relationship wasn't
created automatically, you may see a warning in the
PivotTable field list that states that "Relationships
between tables may need to be created." You can
select auto-detect for the relationship between the
Query table and the new Date table to be created.
page 315
Q1. Something should seem a bit off with your
numbers. There are some big disparities
month over month for some weekdays. Look
back over our query and the ER Diagram (and
if you completed Lab 7-4, compare the query
you executed in this lab to the query from that
lab). What did we leave out of this query?
How could it cause us to make poor
decisions?
Part 4: Analyze and Refine the

Results
The query can be improved by not simply importing all of
the transaction amount data, but by bringing in only the
sales data. The way the data are organized, all of the
dollar amounts for sales and for refunds are in the same
attribute, Tran_Amt, and the transaction type is
differentiated with the attribute Tran_Type. If we filter out
any record that holds return data, we can load only the
data that hold sales transactions into Excel.
12. To edit our original query, click Manage in the Power
Pivot tab in the Excel ribbon.
13. In the Power Pivot tool, click Table Properties from
the Design tab.
Note: If Table Properties is grayed out, switch the table in
view in the Power Pivot window to Query in the bottom
left corner (instead of Calendar).
14. Add in a WHERE clause to the query, validate the new

query, and save it.
Select Tran_Date, SUM(Tran_Amt) AS Sales
From Transact
Where Tran_Type = ’p’
Group By Tran_Date
page 316
15. The data will be automatically refreshed in
the Power Pivot tool and in the Excel worksheet with
the PivotTable. Close the Power Pivot tool.
Part 5: Communicate Results

The refreshed data in the PivotTable is better for making
decisions with, but it still isn’t easy to read at a glance.
Adding some data visualization or conditional formatting
can make these data more meaningful and easier to
interpret.
Q2. What would be the best way to visualize these
data to ease decision making and insight?
End of Lab

Excel (Part IV—Putting It All
Together)
Company summary
be why he shared Transaction data with us to make
Data
10/17/2016.
Software needed

Develop a dashboard to display a variety of KPIs that
you can drill into for state and store details.
Prerequisite
Labs 7-4 and 7-5. If you haven’t completed these labs,
then you can still read through the steps in Labs 7-4 and
7-5 to see the screenshots of the ETL process in Excel
(Lab 7-5) and the KPI creation process (Lab 7-4) to be
ready for this lab.
page 317

In Lab 7-4, you created KPIs for comparing 2015 sales to
2014 sales, but the date was parsed out from the original
Tran_Date attribute. In Lab 7-5, you created a date table
so that the date fields were more descriptive in the Excel
report, but you didn’t create any KPIs. In this lab, we will
combine those two skills to create a descriptive report
with KPIs. We will also expand the reports capabilities by
extracting and loading state and store data in addition to
date and transaction data.

1. Loading the data into Excel from the original SQL
Server database requires some transformation in
Excel’s Power Pivot tool. Extract and load Dillard’s
transactional and store data into Power Pivot using the
following query:
Select Tran_Date, State, Store.Store, SUM(Tran_Amt)
AS Amount
From Transact
Inner Join Store
On Transact.Store = Store.Store
Where Tran_Type = ’p’
Group By Tran_Date, State, Store.Store
Order By Tran_Date
2. It will take a few minutes for these data to load. Once
they do (298,516 rows), close the Table Import Wizard
window. Locate the Tran_Date attribute and use it to
create a Date Table. (Hint: Look in the Design tab.)
3. Now that you have two tables in your data model, return
to the Home tab to create a PivotTable, and close the
Power Pivot tool.
4. In the Power Pivot tab in the Excel ribbon, create a
new measure for Sum(amount). You can call this
measure Current Year. This measure will be used as a
base measure to compare to previous year’s sales
data. Leave the default category as “General.”
5. Open the window to create a new measure to calculate
the previous year’s sales. To create this measure, you
will use Microsoft’s Data Analysis Expressions
language (DAX), which is a formula language for
creating custom calculations and measures. The
function you will use is the = CALCULATE function,
which allows you to not only create a calculation, but
also filter it.
Enter the following expression in the formula box:
= CALCULATE(sum([Amount]),
SAMEPERIODLASTYEAR (’Calendar’[Date]))
You can name this measure Last Year.
6. Create a new KPI, setting Current Year as the Base
Measure and Last Year as the Target Measure. Change
the Status Thresholds to the following:
Anything below 98 percent of last year’s sales (the
target) should be red.
Anything between 98 percent and 102 percent of the
target should be yellow.
Anything above 102 percent of the target should be
green.
7. This KPI will function only with the Date Hierarchy (not
with the date parts). Create a PivotTable with the Date
Hierarchy on the rows and the KPI Status as the
values (if the KPI status is showing –1, 0, and 1 instead
of the stoplight icons, remove the KPI status from the
value fields and then place it back in).
Note: If a relationship wasn't created automatically, you
may see a warning in the PivotTable field list that states
that "Relationships between tables may need to be
created." You can select auto-detect for the relationship
between the Query table and the new Date table to be
created.
page 318
Create another KPI, this time to compare any

month with the month that precedes it (so instead
of comparing September 2016 to September
2015, you will compare September 2016 to
August 2016).
8. Even though the calculation for current month is
technically the same as the calculation for current year
(Sum(Amount)), we have to create a new measure to
use as the KPI’s base. Each base measure can only
have one KPI assigned to it. Create a new measure
called Current Month to calculate sales (this will be the
exact same as how you created Current Year in step 4,
but with a different Measure Name).
9. Create a new measure to use as the monthly target
measure. The DAX expression for calculating last
month’s sales is:
= CALCULATE(SUM([Amount]),
PREVIOUSMONTH(’Calendar’[Date]))
You can name this measure Previous Month.
10. Create a new KPI comparing current sales (your base
measure) to previous month as your target measure.
Create the same status thresholds as the KPI
comparing years (<98%, 98%–102%, >102%).
11. Add this KPI status to your PivotTable.
Part 3: Address and Refine the

Results
This report may be useful at a very high level, but for
state-level and store-level analysis, the level is too high.
Next, we will add in two slicers to help filter the data
based on state and store.
13. From the PivotTable Analyze tab in the Excel ribbon,
click Slicer to insert an interactive filter.
14. Place a check mark in the boxes next to State and

Store to create the slicers and click OK.
page 319
15. Notice what happens as you select
different states: Not only do the data change to reflect
the KPI status for the state that you selected, but the
stores that are associated with that state shift to the top
of the store slicer, making it easier to drill down.
We can ease drill-down capabilities even more by
creating a hierarchy between state and store.
17. Open the Power Pivot tool by clicking Manage from
the Power Pivot tab in the Excel ribbon.
18. From the Power Pivot Home tab, switch to Diagram
View.
19. Select both the State and the Store attributes from the
Query table, then right-click one of the attributes to
create a hierarchy.
20. You can change the name of the Hierarchy to Store

and State Hierarchy.
21. Close the Power Pivot tool. The PivotTable will have
refreshed automatically.
22. You will see that the hierarchy has been added to your
PivotTable Fields list. Drag and drop the hierarchy to
the Rows (above the Date hierarchy).
page 320
Now you can drill down from State to Store directly in

the PivotTable, or you can filter it via the slicer.
Q1. How does the ability to drill down into the state
and store data give management critical
information and help them to identify issues
that are occurring or opportunities that might
be available?
Q2. What would you get sales changes of certain
products (SKUs) or product categories from
one month to the next? Having this type of
information will help you do what to help plan
future promotions or future purchases?
End of Lab
page 321

Dillard’s Store Data: Advanced
Models in Tableau
Company summary
be why he shared Transaction data with us to make
Data
10/17/2016.
Software needed
Tableau (available on the Remote Desktop at the

Develop visualizations to display a variety of
performance data for cluster analysis and store
comparisons.
We’re now going to look at Dillard’s Department store
data to identify outliers and trends in different stores’
transaction activity.
1. Create a new book in Tableau.
2. Go to Connect > To a Server > Microsoft SQL Server.
3. Enter the following and click Sign In:
a. Server: essql1.walton.uark.edu
b. Database: WCOB_DILLARDS
4. Double click the Transact table.
5. Double click the Store table.
6. Change the join to a Right outer join to include all
stores and transactions they involve.
Note: Tableau will try to query the server after each
change you make and will take a up to a minute. After
each change, click Cancel to stop the query until you’re
ready to prepare the final report.
Part 1: Cluster Analysis of High

Volume Stores
In this analysis, we want to see which stores are doing
well. In other words, stores with high volume of
transactions and high average transaction price. Cluster
analysis will group the stores that share similar
performance.
1. Create a new worksheet called Cluster Transactions.
a. Columns: Transaction ID > Measure > Count
b. Rows: Tran Amt > Measure > Average
c. Marks:
i. Store> Color
d. Let the query run at this point. page 322
e. You’ll notice an outlier in top-right corner.
In this case, it is the online division of Dillard’s.
Because we’re evaluating brick-and-mortar stores, we
want to exclude this one.
f. Right-click on the outlier (Store 698) and click

Exclude.
g. To create clusters, click the Analytics tab and drag
Cluster to the scatter plot.
i. Number of clusters: 8
h. Take a screenshot (label it 7-7A).
page 323
Q1. Write down the store numbers in the bottom-

left cluster (four low performing stores) and
the top-right cluster (seven high performing
stores).
Part 2: Stacked Bar Chart of Monthly

Store Performance
Now we should evaluate the low performing stores and
determine how bad things are.
2. Create a new sheet called Sales By Month.
a. Columns: Tran Amt > Measure > Sum
b. Rows: Store, Tran Date > Year
c. Marks:
i. Type: Bar
ii. Tran Date > Color
A. Discrete
B. Month <- There are two month options in the drop-
down. Choose the top one without a year.
C. Sort. . . > Descending
iii. Tran Date > Year > Label
d. Let the query run at this point.
e. Now filter your results. Right-click outside the work
area and click Filters > Store. Then right-click again
and click Filters > Year of Tran Date.

f. Now let’s narrow in on the high performing stores we
identified in our cluster analysis.
i. Uncheck All in the Store filter list.
ii. Check the stores: 405, 716, 743, 760, 768, 775, 796.
g. Finally let’s make a year-over-year comparison on
sales performance by filtering the years.
i. Uncheck All in the Year(Tran Date)filter list.
ii. Check 2014, and 2015.
page 324
Q2. What do you notice about sales for these stores
from 2014 to 2015? Does anything stand out?
3. Uncheck 2014 and check 2016.
Q3. What do you notice about sales for these stores
from 2015 to 2016?
Q4. How would you expect November and
December 2016 sales to be, given they haven’t
been recorded yet?
Part 3: Tree Map of Sales by

Department
Now we’ll look at sales by department. Because we want
to see department names, we’ll need to adjust our data
model first.
5. Click Data Source.
6. Double-click SKU and Department.
7. Change the joins to a Left Outer Join so we include all
transactions and only departments that sold things.
8. Create a new sheet called Sales By Department.
a. No columns or rows
b. Marks:
i. Tran Amt > SUM > Size
ii. Tran Amt > SUM > Color
iii. Deptcent Desc > Label
page 325
c. Let the query run at this point. Once the
query runs, the output should look like the following
screenshot.
Tableau
d. Which department has the highest dollar amount of

sales? Lowest?
9. To drill down to subcategories, we can create a
hierarchy in Tableau.
a. In the attributes list, drag Dept Desc onto Deptcent
Desc to create a hierarchy and click OK.
b. Now in the Marks list, click the + next to Deptcent
Desc to show the brands in each department.
Q5. Which is the best-selling brand of Shoes? The
top three cosmetics brands?
End of Lab
1https://www.linkedin.com/pulse/20130905053105-64875646-the-75-kpis-every-manager-
needs-to-know.
2http://corporate.walmart.com/2016grr/enhancing-sustainability/moving-toward-a-zero-
waste-future (accessed August 2017).
3https://www.linkedin.com/pulse/20130905053105-64875646-the-75-kpis-every-manager-
needs-to-know/ (accessed October 13, 2017).
page 326
Chapter 8
Financial Statement Analytics

In this chapter, we focus on how to access and analyze financial
statement data. We highlight the use of XBRL to quickly and
efficiently gain computer access to financial statement data while
addressing the data quality and consistency issues of XBRL data in
the United States. Next, we discuss how ratios are used to analyze
financial performance. We also discuss the use of sparklines and
other visualization tools to help users identify trends and points of
interest in the data. Finally, we discuss the use of text mining to
analyze the sentiment in financial reporting data.
A Look Back
Chapter 7 focused on generating and evaluating key performance
metrics that are used primarily in managerial accounting. By
measuring past performance and comparing it to targeted goals, we
are able to assess how well a company is working toward a goal.
Also, we can determine required adjustments to how decisions are
made or how business processes are run, if any.
A Look Ahead
In Chapter 9, we highlight the use of data analytics for the tax
function. First, we consider how tax departments get the data they
need from the financial reporting system. Second, we investigate
how data analytics is used to help with tax compliance issues and
tax planning.
page 327
S Narayan/Dinodia Photo/AGE Fotostock
Sometimes the future is now. The StockSnips app uses sentiment analysis, machine
learning, and artificial intelligence to aggregate and analyze news related to publicly
traded companies on Nasdaq and the New York Stock Exchange to “gain stock
insights and track a company’s financial and business operations.” The use of Data
Analytics helps classify the new to help predict revenue, earnings, and cash flows,
which are in turn used to predict the stock performance. What will Data Analytics do
next?
OBJECTIVES
LO 8-1 Understand different types of financial statement

analysis.
LO 8-2 Explain how to create and read visualizations of
financial statement data.
LO 8-3 Describe the value of text mining and sentiment
analysis of financial reporting.
LO 8-4 Describe how XBRL tags financial reporting data.
page 328
FINANCIAL STATEMENT ANALYSIS

LO 8-
1
Understa
nd
different
types of
financial
stateme
nt
analysis.
Financial statement analysis is used by investors, analysts,

auditors, and other interested stakeholders to review and evaluate a
company’s financial statements and financial performance. Such
analysis allows the stakeholder to gain an understanding of the
financial health of the company to allow more insightful and effective
decision making. Most financial statement users will perform
descriptive and diagnostic analytics to understand the firm and
identify trends and relationships among different accounts. Where
there is sufficient data, predictive analytics provides insight into the
future and helps identify companies that may have abnormal returns
(and would be good investments).
Descriptive Financial Analytics

The primary objective of descriptive analytics for financial statements
is to set a benchmark to compare against a company or portfolio of
interest. We may want to calculate the mean rate of return or
working capital within an industry or set of competitors. We may also
use a series of ratio analyses to learn about the composition of
certain accounts or to identify indicators of risk.
Ratio analysis is a tool used to evaluate relationships among
different financial statement items to help understand a company’s
financial and operating performance. It tells us how much of one
account we get for each dollar of another. For example, the gross
profit ratio (gross profit/revenue) tells us how many cents we have
from every dollar of sales that will cover operating expenses.
Financial ratio analysis is a key tool used by accounting, auditing,
and finance professionals to assess the financial health of a
business organization, to assess the reasonableness of reported
financial results, and to predict future performance. Analytical
procedures, including ratio analysis, are recognized as an essential
component of both planning an audit and carrying out substantive
testing. AS 2305 states:
A basic premise underlying the application of analytical procedures is that plausible

relationships among data may reasonably be expected to exist and continue in the
absence of known conditions to the contrary.1
Knowledge of financial statement analysis using ratios is a
component of several professional certifications, including the CPA
(certified public accountant), CMA (certified management
accountant), and CFA (chartered financial analyst) certifications, so
clearly critical for any accountant.
Vertical and Horizontal Analysis

One way that analysts calculate ratios between different accounts is
by preparing a common size financial statement, like the one shown
in Exhibit 8-1. A common size financial statement is a type of
financial statement that contains only basic accounts that are
common across companies. With this we can perform a vertical
analysis, which is an analysis that shows the proportional value of
accounts to a primary account, such as Revenue. In the page 329
following example, we divide operating income by
revenue to show that Apple earns about $0.27 for every dollar in
sales. This is the operating income margin. On the balance sheet we
would use vertical analysis to identify the proportion of assets or
liabilities, for example dividing accounts receivable by total assets.
EXHIBIT 8-1
Vertical Analysis of a Common Size Financial Statement
Other Classes of Ratios
For other indicators of financial health, there are four main types of
ratios: liquidity, activity, solvency (or financing), and profitability. In
practice, these ratios may vary slightly depending on which accounts
the user decided to include or exclude.
Liquidity is the ability to satisfy the company’s short-term
obligations using assets that can be most readily converted into
cash. Liquidity ratios help measure the liquidity of a company. Here
are some common liquidity ratios:
Current ratio = Current assets/Current liabilities
Quick (acid test) ratio = (Current assets – Inventory)/Current liabilities
Working capital = Current assets – Current liabilities

Activity ratios are a computation of a firm’s operating efficiency.
Company activity is often measured by use of turnover ratios reflect
the number of times assets flow into and out of the company during
the period and serve as a gauge of the efficiency of putting assets to
work. Receivables, inventory, and total asset turnover are all
examples of activity ratios. Note that when you page 330
compare income statement (duration) accounts with
balance sheet (point in time) accounts, you need to average the
balance sheet accounts to match the period. Also for turnover ratios,
analysts may use 365 days or round down to 360 days depending on
preference.
Asset turnover ratio = Net sales/Average total assets
Receivable turnover ratio = Net sales/Average net accounts receivable
Average collection period ratio = 365/Receivables turnover
Inventory turnover ratio = Cost of goods sold/Average inventory
Average days in inventory ratio = 365/Inventory turnover
We use solvency (or sometimes called financing) ratios to help

assess a company’s ability to pay its debts and stay in business. In
other words, we assess the company’s financial risk—that is, the risk
resulting from a company’s choice of financing the business using
debt or equity. Debt-to-equity, long-term debt-to-equity, and times
interest earned ratios are also useful in assessing the level of
solvency.
Debt-to-equity ratio = Total liabilities/Shareholders’ equity
Times interest earned ratio = Income before interest and taxes/Interest expense
Profitability ratios are a common calculation when assessing a

company. They are used to provide information on the profitability of
a company and its prospects for the future.
Profit margin on sales ratio = Net income/Net sales
Return on assets ratio = Net income/Average total assets
Return on equity ratio = Net income/Average shareholders’ equity
Asset turnover ratio = Net sales/Average total assets
Equity multiplier ratio = Average total assets/Average total equity
Profitability ratios are commonly associated with the DuPont

ratio. The DuPont ratio was developed by the DuPont Corporation
to measure performance as a decomposition of the return on equity
ratio in this way.
Return on equity (ROE) = Profit margin x Asset turnover x Equity multiplier
It decomposes return on equity into three different types of ratios:

profitability (profit margin), activity (operating leverage or asset
turnover), and solvency (equity multiplier) ratios.
page 331
Diagnostic Financial Analytics

Is a debt-to-equity ratio value of 2.0 good or bad? Does an asset
turnover ratio value of 60 days indicate a problem with obsolete
inventory? Without a benchmark, these ratios give us nothing more
than a data point. We need to compare these to other descriptive
statistics to be able to make a judgment call. For example, if the
industry average debt-to-equity ratio is 1.0 or the average company
has $1 in debt to $1 in equity (a 50/50 split), the comparison would
tell us that the company is quite a bit over leveraged and it is $2 in
debt to every $1 in equity (a 66/33 split).
Benchmarks for financial statements can include direct
competitors, industry averages, or a company’s own past
performance. If a competitor has an asset turnover ratio of 40 days,
our 60 days means we’re less efficient at getting our product out the
door. But if last period we had an asset turnover of 65 days, our
current period’s 60 days reveals improvement in inventory
management since the last period. Using these diagnostic analytics
give a relative sense of place for firm performance. Refer to Exhibit
8-2 to see how Microsoft compares with Apple and Facebook in
converting revenue into profit.
EXHIBIT 8-2
Comparison of Ratios among Three Companies
Auditors will use ratio analysis to pinpoint potential audit issues
by considering how a company’s financial statements depart from
industry performance, a close competitor, or even the same
company’s prior-year performance. Competitors might use ratio
analysis to understand the vulnerabilities of a competitor. Bond
investors might use ratio analysis to see if a bond covenant is
violated (e.g., some bond contracts require a borrower to maintain a
current ratio above 1.0 to help ensure the loan can be paid off).
Predictive Financial Analytics

Predicting future performance of a company is the work of trading
analysts, researchers, the finance department, and managers
responsible for budgeting.
A horizontal analysis is an analysis that shows the change of a
value from one period to the next. This is sometimes called a trend
analysis. When you have two or more periods, you calculate the
proportional change in value from one to the next similar to a ratio
analysis. In Exhibit 8-3 below, we take Revenue in 2018 and divide it
by Revenue in 2017 to show a 114.28 percent change or a 14.28
percent increase from one year to the next for Microsoft.
EXHIBIT 8-3
Horizontal Analysis of a Common Size Financial Statement
page 332
Horizontal analysis can be used to calculate trends from one

period to the next or over time.
Change amount = Current year amount – Base year amount
Change percent = (Current year amount – Base year amount)/Base year amount
When you calculate the trend over a large period of time relative
to a single base year, you create an index. An index is a metric that
shows how much any given subsequent year has changed relative to
the base year. The formula is the same as above, but we lock the
base year value when creating our formula, shown in Exhibit 8-4.
EXHIBIT 8-4
Index Showing Change in Value Relative to Base Year
Using these trends and indices, we can better understand how a

company performs over time, calculate the average amount of
change and predict what the value is likely to be in the next period.
page 333
PROGRESS CHECK
1. Which ratios would a financial institution be most interested
in when determining whether to grant a loan to a business?
2. What would a horizontal trend tell you about a firm’s
performance?
VISUALIZING FINANCIAL DATA

LO 8-
2
Explain
how to
create
and read
visualiza
tions of
financial
stateme
nt data.
Visualizations help to highlight key figures present in the financial

data. Whether to describe the data or show the relative value in
diagnosing points of interest, color and graphs show many different
dimensions.
Showing Trends
Sparklines and trendlines are used to help, financial statement users
easily see the data visually and give meaning to the underlying
financial data. A sparkline is a small visual trendline or bar chart that
efficiently summarizes numbers or statistics in a single spreadsheet
cell. Because it generally can fit in a single cell within a spreadsheet,
it can easily add to the data without detracting from the tabular
results.
For what types of reports or spreadsheets should sparklines be
used? It usually depends on the type of reporting that is selected.
For example, if used in a digital dashboard that already has many
charts and dials, additional sparklines might clutter up the overall
appearance. However, if used to show trends where it replaces or
complements lots of numbers, it might be used as a very effective
visualization. The nice thing about sparklines is they are generally
small and just show simple trends rather than all the details
regarding the horizontal and vertical axes that you would expect on a
normal graph.
Exhibit 8-5 provides an example of the use of sparklines in a
horizontal trend analysis for Microsoft. It shows the relative value of
each line item and the overall trend.
EXHIBIT 8-5
Visualizing Financial Data with Heat Maps and Sparklines
Relative Size of Accounts

Another way to visualize financial data is to use heat maps
(conditional formatting in Excel) and charts. A heat map shows the
relative size of values by applying a color scale to the data. In Exhibit
8-5, the vertical composition of the accounts changes over the five-
year period. Color helps highlight dramatic shifts in each year, such
as the drop in income in 2015.
A balance sheet, on the other hand, has an inherent hierarchy of
accounts that is a good candidate for a sunburst diagram. As shown
in Exhibit 8-6, the center of the ring shows the main sections of the
balance sheet and their proportional size. As you move out, you see
the subgroups and individual accounts that make up the balance
sheet.
page 334
EXHIBIT 8-6
Sunburst Diagram Showing Composition of a Balance Sheet
For some additional examples of visualizations that show
financial data, including tree diagrams, geographic maps, chord
diagrams, and heat maps for word frequency in management
discussion and analysis, explore the following website:
rankandfiled.com.
PROGRESS CHECK
3. How might sparklines be used to enhance the DuPont
analysis? Would you show the sparklines for each
component of the DuPont ROE disaggregation, or would
you propose it be shown only for the total?
TEXT MINING AND SENTIMENT
ANALYSIS
LO 8-
3
Describe
the value
of text
mining
and
sentimen
t
analysis
of
financial
reporting
.
Some data analysis is used to determine the sentiment included in

text. For example, Uber might use text mining and sentiment
analysis to read all of the words used in social media associated with
its driving or the quality of its smartphone app and its services. The
company can analyze the words for sentiment to see how the social
media participants feel about its services and new innovations, as
well as perform similar analysis on its competitors (like Lyft or
traditional cab services).
Similar analysis might be done to learn more about financial
reports, SEC submissions, analyst reports, and other related
documents based on the words that are used. They might provide a
gauge of the overall tone of the financial reports. This tone might
help us understand management expectations of past or future
performance that might complement the numbers and figures in the
reports.
page 335
To provide an illustration of the use and predictive ability of text
mining and sentiment analysis, Loughran and McDonald2 use text
mining and sentiment analysis to predict the stock market reaction to
the issuance of a 10-K form by examining the proportion of negative
words used in a 10-K report. Exhibit 8-7 comes from their research
suggesting that the stock market reaction is related to the proportion
of negative words (or inversely, the proportion of positive words).
They call this method overlap. Thus, using this method to define the
tone of the article, they indeed find a direct association, or
relationship, between the proportion of negative words and the stock
market reaction to the disclosure of 10-K reports.
EXHIBIT 8-7
Stock Market Reaction (Excess Return) of Companies Sorted by Proportion of
Negative Words
The lines represent the words from a financial dictionary (Fin-Neg) and a standard English
dictionary (H4N-INF).
Source: Loughran, Tim, and Bill McDonald. (2011). “When Is a Liability Not a Liability?
Textual Analysis, Dictionaries, and 10-Ks.” Journal of Finance 66, no. 1, pp. 35–65.
They measure proportion first by developing a dictionary of 2,337
negative words in the financial context and then counting how many
of those words are used as compared to the total words used (called
Fin-Neg in Exhibit 8-7). One of their arguments is that a financial
dictionary is better than a dictionary created from standard English
usage. For that reason, they differentiate their financial dictionary
(Fin-Neg) from the negative words used in normal English usage (as
shown in Exhibit 8-7 as H4N-Inf). Whereas cost, expense, or liability
might be viewed as negative in normal English, they are not
considered to be negative words in the financial dictionary. The most
frequent negative words in the financial dictionary include words like
loss, claims, impairment, adverse, restructuring, and litigation.
PROGRESS CHECK
4. Which would you predict would have more positive
sentiment in a 10-K, the footnotes to the financial
statements or the MD&A (management discussion and
analysis) of the financial statements?
5. Why would you guess the results between the proportion of
negative words and the stock market reaction to the 10-K
issuance diverge the Fin-Neg and the H4N-Inf dictionary?
page 336
XBRL AND FINANCIAL DATA

QUALITY
LO 8-
4
Describe
how
XBRL
tags
financial
reporting
data.
XBRL is a global standard for tagging and reporting financial

information in a computer-readable format. XBRL stands for
eXtensible Business Reporting Language and is a type of XML
(extensible markup language) used for organizing and defining
financial elements. In the United States and other jurisdictions,
companies are required to tag each piece of financial data that
appears in their financial statements so that it is machine readable.
Once these instance documents are submitted and validated by the
regulatory body, they are immediately available for public
consumption by different types of financial statement users, including
financial analysts, investors, or lenders. These users can then
leverage data models to quickly analyze large amounts of data from
the entire population of listed companies with minimal effort.
As of June 2011, the Securities and Exchange Commission
requires all public company filers, including smaller reporting
companies and foreign private issuers, to file an XBRL instance
document, which contains the same information found the in the
traditional financial statements but in computer-readable format. In
addition to tagging financial values, such as account balances and
lease amounts, companies must tag every date, fact, figure,
percentage, and paragraph of text in management discussion and
analysis and footnotes.
The preparer of an XBRL instance document must begin by
identifying a correct taxonomy that defines and describes each key
standardized data element (like cash or accounts payable), shown in
Exhibit 8-8. The XBRL taxonomy also defines the relationships
between each element—for example, buildings and improvements
are a component of property, plant, and equipment, which is a
component of noncurrent assets, which is a component of assets,
which is in the statement of financial position (balance sheet), shown
in Exhibit 8-9.
EXHIBIT 8-8
Creating an XBRL Instance Document
EXHIBIT 8-9
Organization of Accounts within the XBRL Taxonomy
The current U.S. GAAP Financial Reporting Taxonomy can be
explored interactively at xbrlview.fasb.org. It defines more than
19,000 elements with descriptions and links to the page 337
FASB codification. For example, the XBRL tag for
cash is labeled “Cash” and is defined as follows:
Amount of currency on hand as well as demand deposits with banks or financial

institutions. Includes other kinds of accounts that have the general characteristics of
demand deposits. Excludes cash and cash equivalents within disposal group and
discontinued operation.3
The XBRL tag for cash and cash equivalents footnote disclosure
is labeled as “CashAndCashEquivalentsDisclosureTextBlock” and is
defined as follows:
The entire disclosure for cash and cash equivalent footnotes, which may include the
types of deposits and money market instruments, applicable carrying amounts,
restricted amounts and compensating balance arrangements. Cash and equivalents
include: (1) currency on hand (2) demand deposits with banks or financial institutions
(3) other kinds of accounts that have the general characteristics of demand deposits
(4) short-term, highly liquid investments that are both readily convertible to known
amounts of cash and so near their maturity that they present insignificant risk of
changes in value because of changes in interest rates. Generally, only investments
maturing within three months from the date of acquisition qualify.4
The use of tags allows data to be quickly transmitted and

received, and the tags serve as an input for financial analysts valuing
a company, an auditor finding areas where an error page 338
might occur, or regulators seeing if firms are in
compliance with various regulations and laws (like the SEC or IRS).
Preparers of the XBRL instance document compare the financial
statement figures with the tags in the taxonomy. When a tag does
not exist, the preparer can extend the taxonomy with their own
custom tags. The taxonomy and extension schema are combined
with the financial data to generate the XBRL instance document
which is then validated for errors and submitted to the regulatory
authority.
XBRL Data Quality

While XBRL enables data analytics models to quickly process and
present interesting patterns in the data, the user must be careful not
to trust all of the numbers at face value. Users may notice that data
values may be missing or incorrect. The XBRL-US Center for Data
Quality laments that:
Investors and analysts have been reluctant to use the data because of concerns
about its accuracy, consistency, and reliability. Inconsistent or incorrect data tagging,
including the use of custom tags in lieu of standard tags and input mistakes, causes
translation errors, which make automated analysis of the data unduly difficult.5
Part of the problem is that most companies outsource the

preparation of XBRL financial statements to other companies, such
as Fujitsu and R.R. Donnelley, and don’t validate the data
themselves. Another problem is that ambiguity in the taxonomy leads
companies to select incorrect tags or use extension tags where a
standard tag exists. Because these statements are not audited, there
is little incentive to improve data quality unless stricter validation
measures are put in place.
Improved data quality for analysts comes at a cost. As analysts
wait for improved data quality in XBRL data, they will turn to
expensive commercial datasets, such as Compustat, that have been
automatically and manually transformed to improve data quality.
Despite these issues, there is still some value in analyzing XBRL
data for illustrative purposes and some providers have additional
solutions to make XBRL data comparable. Sometimes outside data
vendors create standardized metrics to make the company
reported XBRL data more comparable. For example, Calcbench, a
data vendor that eases financial analysis for XBRL uses, makes
standardized metrics, noting:
IBM labels revenue as “Total revenue” and uses the tag “Revenues”, whereas Apple,
labels their revenue as “Net sales” and uses the tag “SalesRevenueNet”. This is a
relatively simple case, because both companies used tags from the FASB taxonomy.
Users are typically not interested in the subtle differences of how companies tag or
label information. In the previous example, most users would want Apple and IBM’s
revenue, regardless of how it was tagged. To that end, we create standardized
metrics.6
page 339
Different data vendors such as XBRLAnalyst and Calcbench

both provide a trace function that allows you to trace the
standardized metric back to the original source to see which XBRL
tags are referenced or used to make up the standardized metric.7
Exhibit 8-10 shows what a report using standardized metrics
looks like for Boeing’s balance sheet. Note the standardized tags
used for Boeing could be used for any of the SEC filers to gather
their balance sheet and other financial statements.
EXHIBIT 8-10
Balance Sheet from XBRL Data
Note the XBRL tag names in the far left column.
Source: https://www.calcbench.com/xbrl_to_excel
page 340
XBRL, XBRL-GL, and Real-Time Financial

Reporting
The desire for machine-readable data doesn’t stop at the financial
statements. Many financial reporting systems within enterprise
systems such as Oracle and SAP have a general ledger that is
consistent with XBRL, called XBRL-GL (XBRL-Global Ledger). That
means once the numbers are input into a financial system, they are
already tagged and able to be transmitted in real time to interested
users in a continuous reporting function.
Of course, there are a number of reasons this information is not
transmitted in real time. For example, the accounting information has
not yet been audited, and it may contain errors. Other information
such as goodwill or long-term debt will likely not change on a minute-
by-minute basis, so there would be no use for it on a real-time basis.
But as systems advance and continuous, real-time auditing becomes
more prevalent, and with our understanding of how and exactly what
type of real-time information might be used, there may be a chance
of providing real-time accounting information in the relative short
term by use of XBRL-GL.
Examples of Financial Statement Analytics

Using XBRL
We illustrate the DuPont Ratios in Exhibit 8-11 by considering a
calculation from some standard XBRL data.
EXHIBIT 8-11
DuPont Ratios Using XBRL Data
Source: https://www.calcbench.com/xbrl_to_excel.
You’ll note for the Quarter 2 analysis in 2009, for DuPont (Ticker
Symbol = DD), if you take its profit margin, 0.294, multiplied by asset
turnover of 20.1 percent multiplied by the financial leverage of 471.7
percent, you get a return on equity of 27.8 percent.
PROGRESS CHECK
6. How does XBRL facilitate Data Analytics by analysts?
7. How might standardized XBRL metrics be useful in
comparing the financial statements of General Motors,
Alphabet, and Alibaba?
8. Assuming XBRL-GL is able to disseminate real-time
financial reports, which real-time financial elements
(account names) might be most useful to decision makers?
And which information might not be useful?
9. Using Exhibit 8-11 as the source of data and using the raw
accounts, show the components of profit margin, operating
leverage and financial leverage and how they are combined
to equal ROE for Q2 2009 for DuPont (Ticker = DD).
page 341
Summary
Data Analytics extends to the financial accounting and
financial reporting space.
Financial statement analytics includes descriptive

analytics, such as financial ratios and vertical analysis;
diagnostic analytics, where we compare those to
benchmarks from prior periods or competitors; and
predictive analytics, including horizontal trend analysis.
Sparklines and trendlines are efficient and effective tools
to visualize firm performance, and sunburst diagrams
and heat maps help highlight values of interest.
Sentiment analysis could be used with financial
statements, other financial reports, and other financially
related information to gauge positive and negative
meaning from otherwise text-heavy notes.
The XBRL taxonomy provides tags for 19,000 financial
elements and allows for the use of company-defined
tags when the normal XBRL tags are not suitable.
By tagging financial elements in a computer readable
manner, XBRL facilitates the accurate and timely
transmission of financial reporting to all interested
stakeholders.
XBRL and Data Analytics allow timely analysis of the
financial statements and the computation of financial
ratios. We illustrated its usage by showing the DuPont
ratio framework.
Key Words
common size financial statements (328) A type of
financial statement that contains only basic accounts that
are common across companies.
DuPont ratio analysis (330) Developed by the DuPont

Corporation to decompose performance (particularly
return on equity [ROE]) into its component parts.
financial statement analysis (328) Used by investors,

analysts, auditors, and other interested stakeholders to
review and evaluate a company’s financial statements
and financial performance.
horizontal analysis (331) An analysis that shows the

change of a value from one period to the next.
index (332) A metric that shows how much any given

subsequent year has changed relative to the base year.
ratio analysis (328) A tool used to evaluate relationships

among different financial statement items to help
understand a company’s financial and operating
performance.
sparkline (333) A small visual trendline or bar chart that

efficiently summarizes numbers or statistics in a single
spreadsheet cell.
standardized metrics (338) Metrics used by data

vendors to allow easier comparison of company reported
XBRL data.
vertical analysis (328) An analysis that shows the

proportional value of accounts to a primary account, such
as Revenue
XBRL (336) XBRL stands for extensible Business
Reporting Language and is a type of XML (extensible
markup language) used for organizing and defining
financial elements.
XBRL-GL (340) Stands for XBRL-General Ledger; relates

to the ability of enterprise system to tag financial elements
within the firm’s financial reporting system.
XBRL taxonomy (336) Defines and describes each key

data element (like cash or accounts payable). The
taxonomy also defines the relationships between each
element (like inventory is a component of current assets
and current assets is a component of total assets).
page 342
ANSWERS TO PROGRESS
CHECKS
1. Liquidity ratios (e.g., current ratio or quick ratio) would
tell the bank whether the business could make
payments. Solvency ratios (e.g., debt-to-equity ratio)
would indicate how leveraged the company was and
the likelihood of paying us back. It may also determine
the interest rate we charge.
2. The horizontal analysis shows the trend over time. We
could see if revenues are going up and costs are going
down as the result of good management or the opposite
in the case of inefficiencies or decline.
3. Answers may vary on how to visualize the data. It might
depend on the type of reporting that is selected. For
example, is it solely a digital dashboard, or is it a report
with many facts and figures where more sparklines
might clutter up the overall appearance? The nice thing
about sparklines is they are generally small and just
show simple trends rather than details about the
horizontal and vertical axes.
4. The MD&A section of the 10-K has management
reporting on what happened in the most recent period
and what they expect will happen in the coming year.
They are usually upbeat and generally optimistic about
the future. The footnotes are generally background
looking and would be much more factual-based,
careful, and conservative. We would expect the MD&A
section to be much more optimistic than the footnotes.
5. Accounting has its own lingo. Words that might seem
negative for the English language are not necessarily
negative for financial reports. For this reason, the
results diverge based on whether the standard English
usage dictionary (H4N-inf) or the financial dictionary
(Fin-Neg) is used. The relationship between the excess
stock market return and the financial dictionary is what
we would expect.
6. By each company providing tags for each piece of its
financial data as computer readable, XBRL allows
immediate access to each type of financial statement
user, be they financial analysts, investors, lenders, for
their own specific use.
7. When journal entries and transactions are made in an
XBRL-GL system, there is the possibility of real-time
financial reporting. In the author’s opinion, income
statement information (including sales, cost of goods
sold, and SG&A expenditures) would be useful to
financial users on a real-time basis. Any information
that does not change frequently would not be as useful.
Examples include real-time financial elements,
including goodwill; long-term debt; and property, plant,
and equipment.
8. Standardized metrics are useful for comparing
companies because they allow for similar accounts to
have the same title regardless of the account names
used by the various companies. They allow for ease of
comparison across multiple companies.
9. Profit margin = (Revenues – Cost of
revenue)/Revenues = ($7.088B – $5.007B)/ $7.088B =
29.4%
Operating leverage = Sales/Assets = ($7.088B /
$35.258B) = 20.1%
Financial leverage = Assets/Equity = $35.258B /
$7.474B = 471.7%
ROE = Profit margin × Operating leverage (or Asset
turnover) × Financial leverage = 0.294 × 0.201 × 4.717
= 0.278
page 343

1. The DuPont analysis of return on equity (ROE) includes
all of the following component ratios except:
a. asset turnover.
b. inventory turnover.
c. financial leverage.
d. profit margin.
2. XBRL stands for:
a. Extensible Business Reporting Language.
b. Extensive Business Reporting Language.
c. XML Business Reporting Language.
d. Excel Business Reporting Language.
3. Which term defines and describes each XBRL financial
element?
a. Data dictionary
b. Descriptive statistics
c. XBRL-GL
d. Taxonomy
4. Which stage of the IMPACT model (introduced in
Chapter 1) would the use of sparklines fit?
a. Track outcomes
b. Communicate insights
c. Address and refine results
d. Perform test plan
5. What is the name of the output from data vendors to
help compare companies using different XBRL tags for
revenue?
a. XBRL taxonomy
b. Data assimilation
c. Consonant tagging
d. Standardized metrics
6. What is the term used to describe the process of
assigning XBRL tags internally within a financial
reporting/enterprise system?
a. XBRL tagging
b. XBRL taxonomy
c. XBRL-GL
d. XBRL dictionary
7. What computerized technique would be used to
perform sentiment analysis on an annual accounting
report?
a. Text mining
b. Sentiment mining
c. Textual analysis
d. Decision trees
8. What type of ratios measure a firm’s page 344
operating efficiency?
a. DuPont ratios
b. Liquidity ratios
c. Activity ratios
d. Solvency ratios
9. What type of ratios measure a firm’s ability to pay its
debts and stay in business?
a. DuPont ratios
b. Liquidity ratios
c. Activity ratios
d. Solvency ratios
10. What is considered an essential component of planning
an audit and carrying out substantive testing that
involves ratio analysis?
a. Environmental analysis
b. Competitive analysis
c. Management integrity analysis
d. Analytical procedures
1. Which would you predict would have more positive
sentiment in a 10-K, the financial statements or the
MD&A (management discussion and analysis) of the
financial statements? More positive sentiment in the
footnotes or MD&A? Why?
2. Would you recommend the Securities and Exchange
Commission require the use of sparklines on the face of
the financial statements? Why or why not?
3. Why do audit firms perform analytical procedures to
identify risk? Which type of ratios (liquidity, solvency,
activity, and profitability ratios) would you use to
evaluate the company’s ability to continue as a going
concern?
4. Go to https://xbrl.us/data-rule/dqc_0015-lepr/ and find
the XBRL element name for Interest expense and
Sales, General, and Administrative expense.
5. Go to https://xbrl.us/data-rule/dqc_0015-lepr/ and find
the XBRL element name for Other NonOperating
Income and indicate whether XBRL says that should
normally be a debit or credit entry.
6. Go to finance.yahoo.com and type in the ticker symbol
for Apple (AAPL) and click on the statistics tab. Which
of those variables would be useful in assessing
profitability?
7. Can you think of any other settings, besides financial
reports, where tagged data might be useful for fast,
accurate analysis generally completed by computers?
How could it be used in a hospital setting? Or at your
university?
8. Can you think of how sentiment analysis might be used
in a marketing setting? How could it be used in a
hospital setting? Or at your university? When would it
be especially good to measure the sentiment?
Problems
1. Can you think of situations where sentiment analysis
might be helpful to analyze press releases or earnings
announcements? What additional information might it
provide that is not directly in the overall announcement?
Would it be useful to have sentiment analysis
automated to just get a basic sentiment measure
versus the base level of sentiment expected in a press
announcement or earnings announcement?
2. We noted in the text that negative words in page 345
the financial dictionary include words like
loss, claims, impairment, adverse, restructuring, and
litigation. What are other negative words might you add
to that list? What are your thoughts on positive words
that would be included in the financial dictionary,
particularly those that might be different than standard
English dictionary usage?
3. You’re asked to figure out how the stock market
responded to Amazon’s announcement on June 16,
2017, that it would purchase Whole Foods—arguably a
transformational change for Amazon, Walmart, and the
whole retail industry.
Required:
a. Go to finance.yahoo.com, type in the ticker symbol for
Amazon (AMZN), click on historical data, and input
the dates around June 16, 2017. Specifically, see how
much the stock price changed on June 16.
b. Do the same analysis for Walmart (WMT) over the
same dates, which was arguably most directly
affected, and see what happened to its stock price.
4. The preceding question asked you to figure out how the
stock market responded to Amazon’s announcement
that it would purchase Whole Foods. The question now
is if the stock market for Amazon had higher trade
volume on that day than the average of the month
before.
Required:
a. Go to finance.yahoo.com, type in the ticker symbol for
Amazon (AMZN), click on historical data, and input
the dates from May 15, 2017, to June 16, 2017.
Download the data, calculate the average volume for
the month prior to June 16, and compare it to the
trading volume on June 16. Any effect on trading
volume of the Whole Foods announcement by
Amazon?
b. Do the same analysis for Walmart (WMT) over the
same dates and see what happened to its trading
volume. Any effect on trading volume of the Whole
Foods announcement by Amazon?
5. Go to Loughran and McDonald’s sentiment word lists at
https://sraf.nd.edu/textual-analysis/resources/ and
download the Master Dictionary. These lists are what
they’ve used to assess sentiment in financial
statements and related financial reports. Give five
words that are considered to be “negative” and five
words that are considered to be “constraining.” How
would you use this in your analysis of sentiment of an
accounting report?
6. Go to Loughran and McDonald’s sentiment word lists at
https://sraf.nd.edu/textual-analysis/resources/ and
download the Master Dictionary. These lists are what
they’ve used to assess sentiment in financial
statements and related financial reports. Give five
words that are considered to be “litigious” and five
words that are considered to be “positive.”
page 346
Lab 8-1 Create a Horizontal and
Vertical Analysis Using XBRL Data
Company summary
This lab will pull in XBRL data from Fortune 100
companies listed with the SEC. You have the option to
analyze a pair of companies of your choice based on your
own interest level. This lab will have you compare other
companies as well.
Data
The data used in this analysis are XBRL-tagged data from
Fortune 100 companies. The data are pulled from
FinDynamics, which in turn pulls the data from the SEC.
Technique
You will use a combination of spreadsheet formulas and
live XBRL data to generate a spreadsheet that is
adaptable and dynamic. In other words, you will create a
template that can be used to answer several financial
statement analysis questions.
Software needed
Google Sheets (sheets.google.com)
iXBRLAnalyst script
(https://findynamics.com/gsheets/ixbrlanalyst.gs)

Part 1: Identify questions related to the income
statement.
Part 2: Master the data.
Part 3: Perform an analysis of the financial data.
Part 4: Address and refine your results.

Financial statement analysis frequently involves
identifying relationships between specific pieces of data.
We may want to see how financial data have changed
over time or how the composition has changed.
Q1. Select a Fortune 100 company, such as Apple

(AAPL) or Nike (NKE), and identify three
questions you might want to know about that
company’s financial performance over the past
three years. For example, “What is the trend of
operating costs?”
Q2. Form a hypothesis for each of your questions.
For example, “I expect Nike’s operating costs
have gone up.”

To create a dynamic spreadsheet, you must first connect
your sheet to a data source on the Internet. In this case,
you will use Google Sheets because it is hosted online
and then add the iXBRLAnalyst script to connect it to
FinDynamics so you can use formulas to query financial
statement elements.
1. Log into Google Sheets (sheets.google.com), and
create a new, blank sheet called XBRL Common.
2. Click Tools > Script Editor from the menu.
3. In a new window (i.e., tab), go to
findynamics.com/gsheets/ixbrlanalyst.gs.
4. Copy and paste the entire script from the FinDynamics
page into the Script Editor window, replacing any
existing text. page 347
5. Click Save and name the project XBRL.
Click OK.
6. Close the Script Editor window and return to your
Google Sheet.
7. Reload/refresh the page. If you see a new
iXBRLAnalyst menu appear, you are now connected to
the XBRL data.
8. Test your connection by typing in the following formula
anywhere on your sheet:
=XBRLFact(“AAPL”,“AssetsCurrent”,“2017”). If your
connection is good, it should return the value
128645000000 for Apple Inc.’s 2017 balance in current
assets.
9. Delete the formula and continue to the next step.
Note: Once you’ve added the iXBRLAnalyst script to a
Google Sheet, you can simply open that sheet, then go to
File > Make a copy . . . , and the script will automatically
be copied to the new sheet.
The basic formulas available with the iXBRLAnalyst
script are:
=FinValue(company, tag, year, period, member, scale)

=XBRLFact(company, tag, year, period, member,
scale, true)
=SharePriceStats(company, date, duration, request)
where:
company = ticker symbol (e.g., “AAPL” for Apple Inc.)
tag = XBRL tag or normalized tag (e.g.,
“NetIncomeLoss” or “[Net Income]”)
year = reporting year (e.g., “2017”)
period = fiscal period (e.g., “Q1” for 1st Quarter or “Y”
for year)
scale = rounding (e.g., “k,” “thousands,” or “1000” for
thousands) [Note: There is an error with rounding, so it is
suggested to simply divide the formula by the scale
instead, e.g. =XBRLFact(c,t,y,p)/scale.]
Because companies frequently use different tags to
represents similar concepts (such as the tags ProfitLoss
or NetIncomeLoss to identify Net Income), it is important
to make sure you’re using the correct values.
FinDynamics attempts to coordinate the diversity of tags
by using normalized tags that use formulas and
relationships instead of direct tags. Normalized tags must
be contained within brackets []. Some examples are given
in Lab Table 8-1A.
If you’re looking for specific XBRL tags, you can
explore the current XBRL taxonomy at xbrlview.fasb.org.
LAB TABLE 8-1A

Normalized Accounts Created by FinDynamics for XBRLAnalyst
Statement of
Balance Sheet Income Statement
Cash Flows
[Cash, Cash [Revenue] [Cash From
Equivalents and [Cost of Revenue] Operations
Short-Term [Gross Profit] (CFO)]
Investments]
[Selling, General & [Changes in
[Short-Term Administrative Working
Investments] Expense] Capital]
Statement of
Cash Flows
[Accounts [Research & [Changes in
Receivable, Development Accounts
Current] Expense] Receivables]
[Inventory] [Depreciation [Changes in
[Other Current (&Amortization), IS] Liabilities]
Assets] [Non-Interest [Changes in
[Current Assets] Expense] Inventories]
[Net of Property, [Other Operating [Adjustments of
Plant & Expenses] Non-Cash
Equipment] [Operating Items, CF]
[Long-Term Expenses] [Provision For
Investments] [Operating Income] Doubtful
[Intangible [Other Operating Accounts]
Assets, Net] Income] [Depreciation
[Goodwill] [Non-Operating (&Amortization),
Income (Expense)] CF]
[Other
Noncurrent [Interest Expense] [Stock-Based
Assets] Compensation]
[Costs and
[Noncurrent Expenses] [Pension and
Assets] Other
[Earnings Before
Retirement
[Assets] Taxes]
Benefits]
[Accounts [Income Taxes]
[Interest Paid]
Payable and [Income from
Accrued [Other CFO]
Continuing
Liabilities, Operations] [Cash from
Current] Investing (CFI)]
[Income from
[Short-Term Discontinued [Capital
Borrowing] Operations, Net of Expenditures]
[Long-Term Taxes] [Payments to
Debt, Current] [Extraordinary Items, Acquire
[Other Current Gain (Loss)] Investments]
Liabilities] [Net Income]
Statement of
Cash Flows
[Current [Net Income [Proceeds from
Liabilities] Attributable to Investments]
[Other Parent] [Other CFI]
Noncurrent [Net Income [Cash From
Liabilities] Attributable to Financing
[Noncurrent Noncontrolling (CFF)]
Liabilities] Interest] [Payment of
[Liabilities] [Preferred Stock Dividends]
[Preferred Dividends and Other [Proceeds from
Stock] Adjustments] Sale of Equity]
[Common [Comprehensive [Repurchase of
Stock] Income (Loss)] Equity]
[Additional Paid- [Other [Net Borrowing]
in Capital] Comprehensive
[Other CFF]
Income (Loss)]
[Retained [Effect of
Earnings [Comprehensive
Exchange Rate
(Accumulated Income (Loss)
Changes]
Deficit)] Attributable to
Parent] [Total Cash,
[Equity Change]
Attributable to [Comprehensive
Income (Loss) [Net Cash,
Parent] Continuing
Attributable to
[Equity Operations]
Noncontrolling
Attributable to [Net CFO,
Interest]
Noncontrolling Continuing
Interest] Operations]
[Stockholders’ [Net CFI,
Equity] Continuing
[Liabilities & Operations]
Equity] [Net CFF,
Continuing
Operations]
[Net Cash, DO]
Statement of
Cash Flows
[Net CFO, DO]
[Net CFI, DO]
[Net CFF, DO]
page 348

Data
We will begin by creating a common size income
statement for one company over a three-year period.
10. In your Google Sheet, begin by entering the values for
the tags, as shown:
LAB EXHIBIT 8-1A
A B
1 Company AAPL
2 Year 2016
3 Period Y
4 Scale 1000000
11. Then set up your financial statement using the following

normalized tags and periods. Note: Because we
already identified the most current year in A2, we’ll use
a formula to find the three most recent years.
LAB EXHIBIT 8-1B

12. Now enter the =XBRLFact() formula to pull page 349
in the correct values, using relative or
absolute references (e.g., $A7, $B$1, etc.) as
necessary. For example, the formula in B7 should be
=XBRLFact($B$1,$A7,B$6,$B$3)/$B$4.
13. If you’ve used relative references correctly, you can
either drag the formula down and across columns B, C,
and D, or copy and paste the cell (not the formula itself)
into the rest of the table.
14. Use the formatting tools to clean up your spreadsheet,
then take a screenshot (label it 8-1A).
Next, you can begin editing your dynamic data and
expanding your analysis, identifying trends and ratios.
15. In your Google Sheet, use a sparkline to show the
change in income statement accounts:
a. In cell E7, type: =SPARKLINE(B7:D7). Next, copy the
sparkline down the column.
b. Note: The line is trending toward the left.
16. Now perform a vertical analysis in the columns to the
right showing each value as a percentage of revenue:
a. Copy cells B6:D6 into F6:H6.
b. In F7, type =B7/B$7.
c. Drag the formula to fill in F7:H19.
d. Format the numbers as a percentage.
e. Add a sparkline in Column I.

Now that you have a common-size income statement,
replace the company ticker in cell B1 with your selected
company’s ticker and press Enter. The data on the
spreadsheet will update.
Q3. Look at the trends and composition of the

income statement, then answer your three
questions from Q1.
Q4. How did the actual results compare with your

hypothesis?
Q5. Replace the company ticker with a competitor

of your company (e.g., MSFT vs. AAPL). How
do their trends compare with your initial
company?
Q6. How could you expand this spreadsheet to
include multiple competitors’ data on the same
sheet for quick analysis?
End of Lab
Lab 8-2 Create Dynamic Common
Size Financial Statements
XBRLAnalyst allows us to easily create common size
financial statements. Using the skills learned in Lab 8-1,
now extend the analysis to identify some companies
based on their financial performance. The page 350
Fortune 100 companies listed in Lab Exhibit
8-2A operate in a variety of industries. Their FY2016
revenue and assets appear below:
LAB EXHIBIT 8-2A

Background Information on Selected Fortune 100 Companies
Revenue Assets
Company (millions) (millions)
FY2016 FY2016
BANK OF AMERICA (BAC), $80,104 $2,187,702
through its subsidiaries,
provides various banking and
financial products and services
for individual consumers, small-
and middle-market businesses,
institutional investors,
corporations, and governments
in the United States and
internationally.
WALMART (WMT) operates $482,130 $199,581
retail stores in various formats
worldwide. The company
operates in three segments:
Walmart U.S., Walmart
International, and Sam’s Club.
Revenue Assets
FY2016 FY2016
CISCO (CSCO) designs, $49,247 $121,652
manufactures, and sells Internet
protocol (IP)–based networking
and other products related to
the communications and
information technology
industries worldwide.
COCA-COLA (KO) is a $41,863 $87,270
beverage company engaging in
the manufacture, marketing,
and sale of nonalcoholic
beverages worldwide.
BOEING (BA) engages in the $94,571 $89,997
design, development,
manufacture, sale, and support
of commercial jetliners, military
aircraft, satellites, missile
defense, human space flight,
and launch systems and
services worldwide.
EBAY (EBAY) provides online $8,979 $23,847
platforms, tools, and services to
help individuals and merchants
in online and mobile commerce
and payments in the United
States and internationally.
AMAZON (AMZN) operates as $135,987 $83,402
an online retailer in North
America and internationally.
Revenue Assets
FY2016 FY2016
MERCK (MRK) provides $39,807 $95,377
various health solutions through
its prescription medicines,
vaccines, biologic therapies,
animal health, and consumer
care products worldwide.
WALT DISNEY COMPANY $55,632 $92,033
(DIS) is an entertainment
company that operates
television and movie studios as
well as theme parks.
MONDELEZ (MDLZ) produces $25,923 $61,538
consumer food products, such
as Oreo cookies.
In Lab Exhibit 8-2B, you’ll find the common size ratios for
each Lab Exhibit 8-2A company’s income statement (as a
percentage of revenue) and balance sheet (as a
percentage of assets).
LAB EXHIBIT 8-2B

Mystery Ratios
page 351
Author note: The current version of the XBRLAnalyst plug-in does not load
data for the Walt Disney Company (DIS). You will see #ERROR! in place of
the expected values. You can complete this assignment using the remaining
companies.
1. Use a Google Sheet with the iXBRLAnalyst script as
well as the normalized accounts in Lab Exhibit 8.2B (or
search for XBRL tags in the FASB taxonomy if
normalized accounts aren’t available) to recreate the
ratios above.
2. Take a screenshot (label it 8-2A) of your completed
worksheet.
Q1. Using the skills learned from your prior

financial accounting classes, your ability to
extract information from XBRL, and your
knowledge of common-size financial
statements, match the company names in Lab
Exhibit 8-2A with their corresponding ratios in
each column of Lab Exhibit 8-2B.
page 352
Column A = ______________ which

company?
Column B = ______________
Column C = ______________
Column D = ______________
Column E = ______________
Column F = ______________
Column G = ______________
Column H = ______________
Column I = ______________
Column J = ______________
End of Lab
Lab 8-3 Analyze Financial
Statement Ratios
Financial analysts, investors, lenders, auditors, and many
others perform ratio analysis to help review and evaluate
a company’s financial statements and financial
performance. This analysis allows the stakeholder to gain
an understanding of the financial health of the company
and gives insights to allow more insightful and, hopefully,
more effective decision making.
In this lab, you will access XBRL data to complete
data analysis and generate financial ratios to compare the
financial performance of several companies. Financial
ratios can more easily be calculated using spreadsheets
and XBRL. You will (1) select an industry to analyze, (2)
create a copy of a spreadsheet template, (3) input ticker
symbols from three U.S. public companies, and (4)
calculate financial ratios and make observations about the
state of the companies using these financial ratios.
Data
Financial Elements from XBRL from SEC Filings
Software needed
Google Account
Google Sheets
Browser connected to Internet
Specifically, you will:

Part 2: Master the data and prepare for analysis.
Part 3: Input ticker symbols into the template.
Part 4: Analyze the financial ratios.

Interested stakeholders of the firm need access to real-
time, accurate financial data. Since 2011, stakeholders
have used XBRL data to meet this need.
page 353
Q1. How does XBRL fulfill the need for real-time,

accurate financial data?
Q2. Why is it useful to compare multiple
companies at once?
Part 2: Master the Data and Prepare

for Analysis
To master the data and prepare for analysis, we need to
pick which industry and which companies to analyze.
1. Below is a list of 15 Fortune 100 companies in five
different industries. Each of these companies has
attributes and strategies that are similar to and different
from its competitors. Choose one industry to analyze.
Retail: Walmart (WMT), Target (TGT), Costco (Cost)
Technology: Microsoft (MSFT), Apple (AAPL),
Facebook (FB)
Pharmaceutical: Johnson & Johnson (JNJ), Merck
(MRK), Bristol-Myers Squibb (BMY)
Finance: Citigroup (C), Wells-Fargo (WFC), JPMorgan
Chase (JPM)
Energy: ExxonMobil (XOM), Chevron (CVX),
ConocoPhillips (COP)
Create a copy of a spreadsheet template in the
following way:
2. Open a web browser and go to drive.google.com.
3. If you haven’t done so already, sign in to your Google
account.
4. Go to http://tinyurl.com/xbrlratios. You will see a
spreadsheet similar to Lab Exhibit 8-3A.
LAB EXHIBIT 8-3A

XBRL Financial Ratios in Google Sheets
Microsoft Excel 2016
5. Click File > Make a copy. . . as shown in Lab Exhibit 8-

3A.
6. Rename your spreadsheet if desired and click OK to
save a copy to your Drive. A new tab will open with your
copy of the spreadsheet. You may now edit the values
and formulas.
page 354
Part 3: Input Ticker Symbols

Refer to Lab Exhibit 8-3B for your industry’s ticker
symbols.
LAB EXHIBIT 8-3B

Input Ticker Symbols
7. Referring to Lab Exhibit 8-3B for your industry’s ticker

symbols, in the Main Company Ticker field, input the
ticker of the company you would like to focus your
analysis on and press Enter. In a moment, the value on
the spreadsheet will change to Loading. . . and then
show your company’s financial figures.
8. In the Most Recent Year field, enter the most recent
reporting year. It may be the current year or the
previous year.
9. In the Period field, enter either FY for a fiscal year or
Q1 for first quarter, etc.
10. In the Round to field, choose the rounding amount.
1,000 will round to thousands of dollars; 1,000,000 will
round to millions of dollars.
11. In the Comparable 1 Ticker field, input the ticker of a
second company you would like to compare with your
first company.
12. In the Comparable 2 Ticker field, input the ticker of a
third company you would like to compare with your first
company.
13. Take a screenshot (label it 8-3A) of your figure with the
financial statements of your chosen companies.
Part 4: Analyze the Financial Ratios

First, review the Facts sheet (or tab) to determine whether
there are any values missing for the companies you are
analyzing. Describe what impact (if any) the missing data
have on the ratios. Once you have determined whether
any data are missing, you have a chance to find some
interesting trends and comparisons in the data. You will
click through the sheets at the bottom to review the ratios.
To aid in this analysis, the template also includes
sparklines that provide a mini-graph to help you quickly
visualize any significant values or trends.
Q3. Review the 14 financial ratios shown across
the tabs and make some conclusions or
judgments about the values, trends, or
comparisons with the other companies. For
example, if one company has a significantly
higher debt-to-equity ratio than the other two,
what might be driving this?
Q4. Has the company you are analyzing seen any
major changes in its ratios in the past three
years? Which of the three companies is most
liquid in the most current year?
Q5. How has your company managed short-term

liabilities over the last three years?
Q6. Analyze liquidity, profitability, financing
(leverage), and activity for your company.
Where is it strong?
Q7. Consider the DuPont framework to interpret
the results and make sure you make a
judgment about your company’s financial
position based upon the data.
End of Lab
page 355
Lab 8-4 Use PivotTables to Analyze

Data from an XBRL Database
Company summary
As the chapter mentioned, there are 19,000 tags in the
XBRL taxonomy, which doesn’t even include the custom
tags that organizations have created for themselves. The
normalized tags XBRLAnalyst provides can be helpful, but
sometimes you will need to find a more specific tag. One
way that you can do this is by using SQL to query an
XBRL database for all tags that are similar to the
normalized tag you are working with.
Data
We have provided a subset of the XBRL database in two
database files, an Access database and a SQLite
database. This lab will specifically walk through
connecting to the Access database, XBRL.accdb, and
how to work with that data in Excel. If you prefer to work
with the data in SQLite, you can answer the questions
directly in the database with SQL queries.
We have used the Arelle open-source XBRL platform
to build our subset, which in turn pulls the data from the
SEC.
Technique
You will use both the database tool of your choice
(Microsoft Access or SQLite) and Microsoft Excel. You
will use the database tool to view the data, then you will
work with the data in Microsoft Excel by connecting to a
database and using PivotTables.
Software needed
Microsoft Access (or SQLite)
Microsoft Excel
Part 1: Identify questions related to XBRL tags and

taxonomy.
Part 2: Analyze tags and then do more in-depth
querying.

One of the aspects that querying the XBRL database can
be most helpful for is quickly viewing a list of tags that are
similar or for quickly viewing a list of companies that have
something in common over the years or even during a
specific filing period.
Q1. Identify three questions that would be
interesting regarding finding commonalities in
XBRL tags or in filtering the data to view a
subset based on similar criteria.
Q2. If you didn’t know how to use SQL to query an
XBRL database, how would you go about
trying to answer the three questions you
identified in Q1?
The questions we will answer are the following:
1. To see another way of working with the data, using a
filter to only see companies that meet a certain type of
criteria, what large accelerated filers filed?
2. To help us identify how many different iterations there
are within one type of financial statement data element,
we will create a query to show us all of the XBRL tags
that contain “cash” in their description.

The XBRL database in full is a very large database. Our
subset is a bit easier to manage. Lab Exhibit 8-4A is a
database schema of the tables, attributes, and
relationships in the Access database that you will work
with in this lab. The database schema in full can be found
at this URL: http://arelle.org/wordpress/wp- page 356
content/uploads/2014/07/sql_diagram.png if
you are interested.
LAB EXHIBIT 8-4A

XBRL Database Schema—Subset
Lab Exhibit 8-4B describes the most common table
functions.
LAB EXHIBIT 8-4B

Table Explanations
Table
Description
Name
Entity Information on the entity that submitted the
filing
Filing Information about individual filings
Report Link from the information about the filing to the
information in the filing
Period Date information
Unit To what degree any observation was recorded
Data – Identify Large Accelerated
Filers
To answer the first question (to identify the number of
large, accelerated filers), we will first assess the data in
the database, then we will connect the database to Excel
to further analyze the data.
3. Begin by identifying which tables contain the data you
will need to answer the problem. Look at the Table
Explanations figure (Lab Exhibit 8-4B).
Q3. Which tables seem to have relevant data to (1)
describe a filer to determine if it is categorized
as large, accelerated filers and (2) determine
the filing date?
page 357
4. In the database, double-click into the tables to get a

glimpse of the data contained in each. Open the file
Lab_8-4_XBRLsubset.accdb (or the SQLite database
Lab_8-4_XBRLsubset_SQLite).
Q4. Which attributes are relevant?
5. You can write a query in Access to answer the question
at this point, if you prefer, but in this lab we will connect
the database to Excel to create a PivotTable to find the
answer. Open a new, blank Excel file.
6. From the Data tab on the ribbon, click Get External
Data and then click From Access.
7. Browse to the file location to select it.

8. Select Enable selection of multiple tables, and then
select the check boxes next to Entity and Filing, then
click OK. Click OK twice more.
9. It should default to a PivotTable Report, but if it doesn’t,

select that option and click OK.
10. It will take a little while for the data to load, page 358
but once it loads, place Name in the Rows,
and create a Slicer for Filer_Category to filter for Large
Accelerated Filers.
11. Take a screenshot that shows the results of your
PivotTable (label it 8-4A).
For Question 2, identify how many different iterations
there are within one type of financial statement data
element, the Aspect table will be most useful. This table
contains the basic building blocks of XBRL taxonomy, and
specifically the Name column contains the different
names of the XBRL elements that have been used in the
sample.
12. Double-click this table in the Access Database
XBRLsubset.accdb to get a glimpse of the different
elements stored.
13. Return to Excel to import the Aspect table. From the
Data tab on the ribbon, click From Access.

14. Browse to the file location to select it.
15. Select the Aspect table, then click OK.
16. It should default to a PivotTable Report, page 359

but if it doesn’t, select that option and click
OK.
17. It will take a little while for the data to load, but once it
loads, place Name in the Rows. This will create a
distinct list of all the element names included in the
Aspect table.
18. To filter for the names that contain Cash, click the filter
button next to Row Labels on the PivotTable, and
select Label Values, then select Contains.
19. Type Cash in the Label Filter box, then click OK.
20. Take a screenshot of your PivotTable (label it 8-4B).

Based on the massive amount of tags that contain the
word cash in them, we may decide to be more specific
with the query.
21. This time, refine the query or your PivotTable filter to
show only the tags that begin with the word Cash.
22. Take a screenshot of your refined results (label it 8-4C).
Q5. How would you further drill down into the first
question about the large filers?
Q6. Do you think the number of outputs you got for
the different types of tags with the word Cash
is reasonable? What recommendation would
you have regarding the numerous elements in
the taxonomy?
End of Lab
1Source: PCAOB, AS 2305, https://pcaobus.org/Standards/Auditing/Pages/AS2305.aspx.

2Tim Loughran and Bill McDonald, “When Is a Liability Not a Liability? Textual Analysis,
Dictionaries, and 10-Ks,” Journal of Finance 66, no. 1 (2011), pp. 35–65.
3Source: https://xbrl.us/xbrl-taxonomy/2017-us-gaap/
4Source: https://xbrl.us/xbrl-taxonomy/2017-us-gaap/
5Source: https://xbrl.us/data-quality/
6Source: Accessed August 2017. https://knowledge.calcbench.com/hc/en-
us/articles/230017408-What-is-a-standardized-metric.
7https://knowledge.calcbench.com/hc/en-us/articles/230017408-What-is-a-standardized-
metric.
page 360
Chapter 9
Tax Analytics

We highlight the use of data analytics for the tax function. First, we
consider how tax departments could better control the data they
receive from the financial reporting system. Second, we investigate
how data analytics is used to help with tax compliance issues.
Finally, we consider how data analysis might be used to assist in tax
planning including what-if analysis for new legislation, the possibility
of a merger with another company, a shift in product mix or a plan to
set up operations in a new low-tax jurisdiction (and/or transfer
pricing).
A Look Back
In Chapter 8, we focus on how to access and analyze financial
statement data. We highlight the use of XBRL to quickly and
efficiently gain computer access to financial statement data. Next,
we discuss how ratios are used to analyze financial performance.
We also discuss the use of Sparklines to help users visualize trends
in the data. Finally, we discuss the use of text mining to analyze the
sentiment in financial reporting data.
A Look Forward
In Chapter 10, we bring all of the accounting Data Analytics concepts
together with a set of exercises that walk all the way through the
IMPACT model. The chapter serves as a great way to bring together
all of the elements learned in the course.
page 361
Jim McIsaac/Getty images
Knowing the tax liability for a move to a new jurisdiction is important for corporations
and individuals alike. For example, a tax accountant might have advised LeBron
James not to sign with the Los Angeles Lakers in summer 2018 because it is
expected it will cost him $21 million more in extra state income taxes since California
has higher taxes than Ohio. Tax data analytics for this type of “what-if scenario
analysis” is important for tax planning purposes. It has wide applications when
contemplating new legislation, a merger possibility, a shift in product mix, or a plan to
set up operations in a new low-tax jurisdiction. Amazon recently used this when
considering the tax incentives for property taxes, payroll taxes, and sales taxes for its
second headquarters location (or its HQ2) which landed near Washington, DC.
Source: https://www.forbes.com/sites/seanpackard/2018/07/02/lebrons-move-could-
cost-him-21-million-in-extra-state-taxes/#6517d3156280, accessed August 2, 2018;
https://www.cnbc.com/2018/11/13/amazon-tax-incentives-in-new-york-city-virginia-
and-nashville.html,accessedAugust2,2018.
OBJECTIVES
LO 9-1 Describe how tax data has traditionally been

managed and how it should be managed for data
analysis purposes.
LO 9-2 Describe the use of Data Analytics in tax planning.
LO 9-3 Understand the use of tax analytics in visualizations.
LO 9-4 Understand the use of tax data for tax planning and
perform what-if scenario analysis.
page 362
INTRODUCTION TO TAX ANALYTICS

LO 9-
1
Describe
how tax
data has
tradition
ally been
manage
d and
how it
should
be
manage
d for
data
analysis
purpose
s.
With more and more data available, just like other areas in
accounting, there is an increased focus on tax analytics. New
regulations are requiring greater detail, and tax regulators are getting
more adept at the use of analytics. In addition to the regulator side,
tax filers now have more data to support their tax calculations.
Here are a few ways increased access to data and tax analytics
are used by regulators and by companies:
IRS and tax analytics. The IRS has a huge trove of data about
each taxpayer. There are three main sources of information,
including the following:
Not only do they have data of the reportable financial transactions
that occur during the year (including W-2s, Form 1099s, Schedule
K-1s), but also the IRS has a repository of tax returns from prior
years that they have stored in a data warehouse.
The IRS mines and monitors personal data from social media (such
as Facebook, Twitter, Instagram, etc.)1 about taxpayers. For
example, posts about a new car, new house, or fancy vacation
could help the IRS capture the taxpayer dodging or misreporting
income. Divorce lawyers certainly use the same tactics to learn the
lifestyle and related income of a divorcing spouse!
The IRS has personal financial data about each taxpayer, including
Social Security numbers, bank accounts, and property holdings.
While most of this is gathered from prior returns and transactions
(see item 1), the IRS can also access your credit report during an
audit or criminal investigation to determine if spending/credit looks
proportional to income and if they are trying to collect an
assessment.
Each of these sources of information can help the IRS to
establish a profile (using the profiling test approach discussed in
Chapter 3). The IRS has an algorithm called Discriminant Function
that pulls historical data for average amount and type of deductions
related to income level and predicts the likelihood of underreported
income. When the amount self-reported by the taxpayer is
significantly less than the amount estimated, additional investigation
and a potential tax audit might be warranted if the potential tax
revenue is greater than the expected cost of the investigation.
Documenting book-tax differences. One way for tax regulators to
assess if companies are paying sufficient tax is to look at the
differences between the amount of income reported for financial
reporting purposes (like form 10-Q or 10-K submitted to the SEC)
and the amount reported to the IRS (or other tax authorities) for
income tax purposes. Increasingly, tax software and analytics (such
as Hyperion or Corptax) is used to help with the reconciliation to
find both permanent and temporary differences between the two
methods of computing income and also to provide needed support
for IRS schedule M-3 (Form 1120).
R&D Tax Credit. The R&D Tax Credit is a tax credit under Internal
Revenue Code section 41 for companies that incur research and
development (R&D) costs. To receive this credit, firms must
document an appropriate level of detail before receiving R&D tax
credit. For example, companies have to link an employee’s time
directly to a research activity or to a specific project to qualify for the
tax credit. Let’s suppose that a firm spent money on qualifying R&D
expenditures but simply did not keep the sufficient detail needed as
supporting evidence to receive the credit. Analytics could be used to
consider to find the needed detail (timesheets, calendars, project
timelines, document meetings between various employees, time
needed for management review, etc.) to qualify for the R&D tax
credit.
page 363
In terms of the IMPACT model, types of appropriate questions for

the tax function in terms of data analytics might be as follows:
1. What can tax analytics do to reduce the overall current and future
tax liability for an individual or for a company?
2. How might tax analytics reduce the cost of compliance and tax
planning by companies?
3. If certain tax legislation passes, what level of exposure (additional
tax) might the company face?
PROGRESS CHECK
1. What are examples of tools the IRS has to monitor tax
evasion?
2. How can tax analytics support and potentially increase the
amount of R&D tax credit taken by a company?
MASTERING THE DATA THROUGH

TAX DATA MANAGEMENT
LO 9-
2
Describe
the use
of Data
Analytics
in tax
planning.
The tax function typically uses data from the financial reporting
system (or enterprise system). However, the financial reporting
system is primarily designed and used for financial accounting
purposes, where transactions that have an economic impact are
recorded as an input for the financial statements and other financial
reporting purposes. In addition, these financial reporting systems
along with other data have also been used for management
accounting purposes to allow management to calculate the cost of a
product or to optimize a product mix that would maximize profits for
the firm. There is generally not a completely separate information
system solely collecting tax data needed for tax compliance and tax
planning.
With little integration between the financial reporting system and
the needs of the tax function, tax departments would manually
collect and extract data from its financial reporting system and
generalized data warehouse. After gathering data from these
generalized data warehouses, tax departments would use Excel
spreadsheets to capture and store the detail needed to support tax
calculations. Such lack of integration hampered efforts of tax
accountants to have the needed information to comply with tax law,
to minimize current taxes and to allow for tax planning for future
transactions.
With recent advances in technology, there are increasing
opportunities for tax departments to have greater control of their
data, which allows them to work more effectively and efficiently.
Specifically, instead of use of a generalized data warehouse,
enterprise systems increasingly use specific data marts for their tax
function. Data marts are defined as being a subset of the data
warehouse oriented toward a specific need. Such a tax data mart is
used to extract past and real-time data from the financial reporting
system that is most applicable to the tax function. Tax departments
are able to specify what data might affect their tax calculations for
their tax data mart and have a continuous feed of that data. Such a
tax data mart allows tax departments to more completely own the
data than from a more generalized data warehouse or generalized
data mart. They can add to that tax data mart, other relevant
information that might come from other sources.
They are also able to keep it as a centralized repository so that
different users of the tax function can have access to the data.
Exhibit 9-1 provides a good illustration of how data is accumulated
and subsequently dedicated for the tax function. Consistent with the
IMPACT model, tax data warehouses and tax data marts help tax
departments to “master the data” to address tax questions and
issues inside the company.
page 364
EXHIBIT 9-1
Tax Data in a Data Warehouse
Accounting firms can also keep track of its clients using another
type of data mart; for example, a tax data mart kept at an accounting
firm that might have marketing implications. Let’s suppose an
accounting firm has a tax data mart that keeps track of clients and
their unrealized capital gains. The 2018 Tax Cuts and Jobs Act Tax
Reform offers a major change to investors, allowing them to invest
in opportunity zones (in low-income communities) to defer or
completely eliminate taxes on realized capital gains if reinvested in
opportunity zones. While only a fraction of the estimated total
unrealized capital gains market of $6.1 trillion actually qualifies for
opportunity zones,2 there seems to be an almost endless set of
investors that could reap tax savings via an opportunity zone. If a tax
data mart allows accounting firms to know which investors have
unrealized capital gains, they can effectively market tax assistance,
education about opportunity zones, or market investments in
opportunity funds to them directly.
PROGRESS CHECK
3. Why do tax departments need to extract data for tax
calculation from a financial reporting system?
4. How is a tax data mart specifically able to target the needs
of the tax department?
TAX DATA ANALYTICS

VISUALIZATIONS
LO 9-
3
Understa
nd the
use of
tax
analytics
in
visualiza
tions.
Tax Data Analytics Visualizations and Tax

Compliance
Increasingly, tax regulators are using Data Analytics to evaluate tax
compliance by those with potential tax liability. Tax regulators use
Data Analytics to see if companies are close to actually paying what
would be expected based on tax rates and expected income or
sales. To date, companies have not engaged in the same level of
Data Analytics. If for no other reason, companies might engage in
Data Analytics to avoid a tax audit. In some sense, this allows
companies to “see what the regulator is seeing.” The European
Union is way ahead of the United States on use of Data Analytics in
a tax audit, both by the regulator and the company hoping to not be
audited.
page 365
Evaluating Sales Tax Liability

Evaluating sales tax liability can quickly be complicated by customer
sales returns where sales taxes are returned to customers. That
complexity is compounded by the differing tax rates in each city,
county and state jurisdiction. In preparation for an audit, sales tax
regulators could ask for gross sales or net sales (gross sales less
returns) by store and compute the total taxes owed to see if it is
close to the amount of sales taxes actually paid by the company.
Companies can run the same type of analysis to see where they
stand to avoid an audit or at least be prepared in the eventuality that
a tax audit does occur.
With the recent Supreme Court Wayfair decision and the tidal
wave of states passing legislation to copy the decision, collection of
sales tax on every online purchase (based on where the customer
lives) is a serious compliance issue. Companies need to get data
collection processes in place to collect, summarize, and process this
information so that they can have functional compliance with the new
laws in states where they sell online.
A dashboard (similar to that introduced in Chapter 4) is a type of
visualization that might be helpful for compliance with state sales tax.
The comprehensive labs at the end of this chapter provide an
example of how Data Analytics might be used with respect to state
sales tax data. Companies pay sales tax regularly based on the
amount of sales they collect.
Evaluating Income Tax Liability

Tax data analytics allows tax departments to view multiple years,
periods, jurisdictions (state or federal or international, etc.) and
differing scenarios of data typically through use of a dashboard .
Dashboards allow tax departments to evaluate those jurisdictions
where current state income tax liabilities have departed most from
the liability of prior years. This allows tax departments to evaluate
further why current year jurisdictional taxable income and tax liability
have changed from the past and address if any issues or
irregularities occur.
TAX DATA ANALYTICS VISUALIZATIONS ALLOWS A

WAY TO MONITOR AND TRACK KPIs
As noted in this text, a key output of rich data from tax analytics
is the ability to create visualizations. As noted in Chapter 7,
tracking KPIs using visualizations is a good way to easily see
how well the company is performing. Such KPIs might be used
to monitor different aspects of the tax function.
In the article, “Defining Success: What KPIs Are Driving the
Tax Function Today,”3 PwC points out four general categories of
tax-focused KPIs, including tax cost, tax risk, tax efficiency and
effectiveness, and tax sustainability. Below we list some KPIs
that might be used to measure performance in each of these
areas:
Tax cost: The actual amount of tax paid. Example KPIs

include:
Effective tax rate (ETR).
Cash taxes paid.
Effect of loss carry-forwards.
Expiration of tax credits.
page 366
Tax adjustments in response to new tax
legislation.
Deferred taxes.
Tax risk(financial and reputational): With increased regulator

and stakeholder scrutiny, firms bear the financial and reputation
risk of the misreporting or tax provision adjustments. Example
KPIs include:
Frequency and magnitude of tax audit adjustments.
Frequency of concerns pertaining to the organization’s tax
position.
Levels of late filing or error penalties and fines.
Number of resubmitted tax returns due to errors.
Tax efficiency and effectiveness: This includes the efficiency
and effectiveness of technology, processes, and people in
carrying out the tax function. Example KPIs include:
Levels of technology/tax training.
Amount of time spent on compliance versus strategic activities.
Level of job satisfaction of the tax personnel.
Employee turnover of the tax personnel.
Improved operational efficiency.
Tax sustainability: Refers to the ability to sustain similar tax
performance over time. Example KPIs include:
Number of company tax audits closed and significance of
assessment over time.
The effective tax rate (ETR) over time.
Additionally, tax managers should track permanent differences
between book and tax revenue and expenses to ensure
compliance and dispute overpayments of taxes. These include:
Penalties and fines (excluded from taxable income).
Meals and entertainment (100 percent books, 50 percent tax).
Interest on municipal bonds (non-taxed income).
Life insurance proceeds (non-taxed income).
Dividends received deduction (taxed based on percentage of
ownership).
Excess depreciation.
These tax-focused KPIs appear on dashboards or cockpits,
consistent with the “C” (communicate insights) and the “T”
(tracking outcomes) of the IMPACT model. Cockpits are similar
to dashboards but are much narrower in scope and focus than
a dashboard. This focus allows the tax function to highlight
potential high impact or single areas of concern like
reconciliation. We also note that the tax sustainability KPIs, in
particular, measure performance over time and are consistent
with the “T” (tracking outcomes) of the IMPACT model.
PROGRESS CHECK
5. Why is ETR (effective tax rate) a good example of a tax
cost KPI? Why is ETR over time considered to be a good
tax sustainability KPI?
6. Why would a company want to track the levels of late filing
or error penalties as a tax risk KPI?
page 367
TAX DATA ANALYTICS FOR TAX

PLANNING
LO 9-
4
Understa
nd the
use of
tax data
for tax
planning
and
perform
what-if
scenario
analysis.
Tax planning is the analysis of potential tax liability and formulation

of a plan to reduce the amount of taxes paid. It involves forecasting
corporate activity and calculating the anticipated tax liabilities or
benefits from operations in various jurisdictions. Tax analytics helps
organizations operate in a way that helps them be as tax efficient as
possible by identifying opportunities to minimize the amount of
current and future taxes paid as well as to recover tax overpayment.
Tax accountants can utilize the abundance of detailed transaction
and metadata (e.g., descriptions of data, such as categories) to filter
and analyze the data, identify opportunities for tax savings, and plan.
Tax savings and recovery are especially important because they
represent value-adding functions of tax accountants since every tax
dollar saved goes directly to the bottom line (e.g., net income after
tax).
Changes in tax legislation, changes in ownership, expansion into
new territories, and transfer pricing for intercompany sales affect
future tax liability. Beyond calculating tax rates across multiple
jurisdictions, tax planning involves identifying transactions and
investments that are subject to deductions, credits, and other
exclusions from income. Tax planning may involve the following
questions:
What will be the impact of a new tax rate on our tax liability?
Are we minimizing our tax burden by tracking all eligible deductible
expenses and transactions that qualify for tax credits?
What would be the impact of relocating our headquarters to a
different city, state or country?
What is the tax exposure for owners in the case of a potential
merger or significant change in ownership?
Do our transfer pricing contracts on certain products put us at
higher risk of a tax audit because they have abnormal margins?
What monthly trends can we identify to help us avoid surprises?
Can we reduce the number of assumptions in our tax plan?
How are we addressing tax complexities resulting from online sales
due to new sales tax legislation?
How would tax law changes affect our pension or profit-sharing
plans and top employee compensation packages (including stock
options)?
How would the use of independent contractors affect our payroll tax
liabilities?
The answers to these questions come from analysis of current
transaction data and a collection of parameters that represent
potential assumption changes. A combination of descriptive and
predictive analytics with visualizations provide guidance for decision
makers in each of these cases.
Descriptive tax analytics provide insight into the current
processes, policies, and calculations related to determining tax
liability. These analytics involve summarizing transactions by
jurisdiction or category to more accurately calculate tax liability.
Diagnostic tax analytics might help identify items of interest, such as
high tax areas or excluded transactions. For example, creating a
trend analysis for sales and use tax paid in different locations would
help identify seasonal patterns or abnormal transaction volume that
warrant further investigation.
Predictive tax analytics use historical data and new information to
identify future tax liabilities. On the basic level, this includes
regression and what-if analyses and requires a specific target, such
as the value of a tax credit or deferred tax asset. The addition of
ancillary data, including growth rates, trends, and other identified
patterns, aids to the usefulness of these analyses. Additionally, tax
analytics rely on tax calculation logic and tax determination, such as
proportional deductions, to determine the potential tax liability.
page 368
What-If Scenarios
What-if scenario analysis tests the impact of various input data on
an expected output. In tax, this means the manipulation of inputs—
such as multiple tax rates, a series of transactions, and varying profit
margins—to estimate the future outputs, including estimated book
income, cash taxes paid, and effective tax rates. These analyses
attempt to optimize the inputs to reach a desired goal, such as
minimizing the effective tax rate or generating a portfolio of possible
outputs given the inputs. In these cases, we need to estimate the
possible inputs and outputs as well as determine the expected
probabilities of those items.
For example, assume the Pennsylvania General Assembly is
debating a reduction in the statutory corporate income tax rate from
10% to either 8% or 7% with a positive (+5%), neutral, or negative
(−5%) change in corporate income. A company with expected
earnings before tax of $1,000,000 might see potential tax savings
shown in Table 9-1.
TABLE 9-1
Estimated Change in Tax Burden under Different Income Tax Proposals
Based on average earnings before tax of $1,000,000. Negative values represent tax
savings.
Change in Taxable Income / Change

10% 8% 7%
in Tax Rate
Positive change (+5%) 5,000 (16,000) (26,500)
Neutral change (+0%) 0 (20,000) (30,000)
Negative change (−5%) (5,000) (24,000) (33,500)
By itself, this analysis may indicate the path to minimizing tax

would be the lower tax rate with negative growth. An estimate of the
joint probabilities of each of the nine scenarios determines the
expected value of each, or the most likely impact of a change (as
shown in Table 9-2) and the dollar impact of the expected change in
value (in Table 9-3). For example, there is a 0.05 probability (as
shown in Table 9-2) that there will be +5% change in taxable income
but no change in tax rate. This would result in a $250 increase in
taxes (as shown in Table 9-3). In this case, the total expected value
of the proposed decrease in taxes is $15,575, which is the sum of
the individual expected values as shown in Table 9-3.
TABLE 9-2
Joint Probabilities of Changes in Tax Rate and Change in Income
TABLE 9-3
Expected Value of Each of the Scenarios
Change in Taxable Income / Change in

10% 8% 7%
Tax Rate
Positive change (+5%) 250 (1,600) (2,650)
Neutral change (+0%) 0 (4,000) (3,000)
Negative change (−5%) (500) (2,400) (1,675)
The usefulness of the what-if analysis is that decision makers can

see the possible impact of changes in tax rates across multiple
scenarios. This model relies heavily on assumptions that drive each
scenario, such as the initial earnings before tax, the expected
change in earnings, and the possible tax rates. Data Analytics help
confirm or refine the details guiding the scenarios so the decision
maker doesn’t have to rely on as many assumptions. The more
analyzed data that are available to inform the assumptions of the
model, the more accurate the estimates and page 369
expected values can be. Here, data analysis of
before-tax income and other external factors can help determine
more accurate probability estimates. Likewise, an analysis of the
legislative proceedings may help determine the likelihood of a
change.
What-If Scenarios for Potential Legislation,
Deductions, and Credits
Changes in the tax code complicate tax estimates and payments.
Potential changes in legislation are generally complex, involving
identification of qualifying transactions, calculating partial transaction
amounts, analyzing groups of transactions, and determining the
impact of the change from current policy. Changes involve updating
rules and decision aides, as well as capturing previously ignored
metadata (such as categories).
Just like scenario analysis involving changes to corporate tax
rates, we examine another scenario analysis with the use of R&D tax
credits. For example, the United States allows companies to take a
research credit of up to 20 percent of qualified research expenditures
(QREs) used to develop new products exceeding a calculated base
amount, limited by a ceiling. The use of tax analytics to determine
adjustments to the research credits requires a correct determination
of expenses related to qualifying research. If any part of the research
credit were to change—such as the percentage, base amount, or
ceiling—companies would need to anticipate the change and enact
policies and reporting to calculate the new values.
Data Analytics helps refine the model by more accurately
calculating current levels of activity and estimating trends for most
likely changes in the future. To determine the level of research
activity within a firm, the system designers would need to
appropriately code transactions and individuals that qualify for the
research credit. Some of these inputs and variables require compiled
data that includes:
Qualified research activities.
Wages, bonuses, and stock options for employees engaged in,
supporting, or supervising qualified research.
Supplies used to conduct qualified research.
Contract research expense paid for qualified research by a third
party.
Average gross receipts over a four-year period.
Limits on research credit.
Carry-forward credit balance.
If the metadata and tagging for qualified research activities is
inaccurate or missing, additional data ETL would be required.
Scenarios involving changes to the research credit would most
likely include the following variables:
Fixed-base percentage.
Ceiling for fixed-base percentage.
Floor of current QREs.
Credit percentage.
Current and future levels of qualified research activity.
One current change to research and development expenses that
provides an interesting opportunity for analysis is the recent change
in U.S. tax code. While companies currently expense research and
experimental (R&E) expenditures in the year they are incurred, as of
December 31, 2021, the IRS will require companies page 370
to capitalize R&E expenditures and amortize them
over 5 years. This will result in an increase in taxable income and
corresponding tax liability. What-if scenario analysis of current
research activity and expected activity through 2022 can provide
insight into the amount of tax that is likely to be collected after the
change goes into effect.
PROGRESS CHECK
7. What are some data a tax manager would need in order to
perform a what-if analysis of the potential effects of a stock
buyback?
8. How does having more metadata help a tax accountant
minimize taxes?
Summary
Recent advances in Data Analytics extend to the tax
functions, allowing them to work more effectively,
efficiently, and with greater control over the data.
New regulations are requiring greater detail, and tax

regulators are getting more adept in the use of analytics.
In addition to the regulator side, tax filers now have
more data to support their tax calculations.
While the tax department has traditionally just used data
from the financial reporting system, there are increasing
opportunities to control and expand upon available tax
data to help address the most important tax questions.
Tax visualizations (dashboards, cockpits) can be helpful
in monitoring how the tax function is doing in meeting its
KPIs (key performance indicators).
Tax data analytics can be especially powerful in doing
tax planning and formulating what-if scenarios.
Key Words
2018 Tax Cuts and Jobs Act Tax Reform (364) Tax
legislation offering a major change to the existing tax
code.
data mart (363) A subset of the data warehouse focused
on a specific function or department to assist and support
its needed data requirements.
data warehouse (363) A data warehouse is a repository

of data accumulated from internal and external data
sources, including financial data, to help management
decision making.
tax data mart (363) A subset of a company-owned data

warehouse focused on the specific needs of the tax
department.
tax planning (367) Predictive analysis of potential tax

liability and the formulation of a plan to reduce the amount
of taxes paid.
what-if scenario analysis (368) Evaluation of the impact

of different tax scenarios/alternatives on various outcome
measures including the amount of taxable income or tax
paid.
page 371
ANSWERS TO PROGRESS
CHECKS
1. The IRS has all reportable financial (taxable)
transactions at their disposal (including W-2s, Form
1099s, Schedule K-1s, etc.); access to social media
feeds for taxpayers; and access to credit reports, bank
accounts, etc., to help monitor the potential for tax
evasion.
2. Analytics could be used to find the needed detail
(timesheets, calendars, project timelines, document
meetings between various employees, time needed for
management review, etc.) to qualify for the R&D Tax
Credit.
3. Tax data marts are a repository of data from the
financial reporting and other systems to get the data to
support tax department needs.
4. Tax departments are able to specify what data might
affect their tax calculations for their tax data mart and
have a continuous feed of that data. This data mart is
essentially one where the tax department can, in some
sense, “own” the data because no other group has
rights to modify it.
5. The ETR (effective tax rate) is generally used as a
measure of the tax cost used by the tax department to
understand how well they are keeping the tax cost at a
minimum. The lower the effective tax rate, the more
effective the tax department is at finding ways to
structure transactions to minimize taxes and find
applicable tax deductions and tax credits (like the R&D
tax credit or other tax loopholes). Monitoring the level of
the ETR over time helps us know if the tax department
is persistent and consistent in reducing the taxes paid,
or if this rate is highly variable. Generally, most tax
professionals would consider the more stable the ETR
over time, the better. Tracking ETR over time as part of
the tax sustainability KPIs allows management and the
tax department figure out if the ETR is persistent or if
the rate bounces around each year in an unsustainable
way.
6. The greater the number of levels of late filings or error
penalties, the more vulnerable the company is to
penalties, tax audits, and missed tax saving
opportunities.
7. Data may include the possible price of the stock, the
potential capital gains incurred by the stockholders, and
number of shares.
8. The more metadata, the better the tax accountant can
accurately calculate the amounts of taxable and
nontaxable items. For example, they can more clearly
identify expenses that qualify for the research and
development credit or track meal and entertainment
expenses that may trigger tax presence in other
locations.

1. Which stage of the IMPACT model (introduced in
Chapter 1) would the use of tax cockpits fit?
a. Track outcomes
b. Master the data
c. Address and refine results
d. Perform test plan
page 372
2. Tax departments interested in maintaining
their own data are likely to have their own
a. tax reporting system.
b. tax data mart.
c. tax dashboard.
d. tax analytics.
3. According to the textbook, an example of a tax
efficiency and effective KPI would be:
a. number of audits closed.
b. ETR (effective tax rate) over time.
c. number of resubmitted tax returns due to errors.
d. amount of time spent on compliance versus strategic
activities.
4. According to the textbook, an example of a tax
sustainability KPI would be:
a. frequency of concerns pertaining to the organization’s
tax position.
b. level of job satisfaction of the tax personnel.
c. levels of technology/tax training.
d. number of audits closed and significance of
assessment over time.
5. According to the textbook, an example of a tax cost KPI
would be:
a. employee turnover of the tax personnel.
b. levels of technology/tax training.
c. ETR (effective tax rate).
d. levels of late filing or error penalties.
6. The task of tax accountants and tax departments to
minimize the amount of taxes paid in the future is
called:
a. tax planning.
b. tax compliance.
c. tax minimization.
d. tax sustainability.
7. According to the textbook, an example of a tax risk KPI
would be:
a. employee turnover of the tax personnel.
b. levels of technology/tax training.
c. ETR (effective tax rate).
d. levels of late filing or error penalties.
8. allows tax departments to view multiple years,
periods, jurisdictions (state or federal or international,
etc.), and differing scenarios of data, typically through
use of a dashboard.
a. Tax data analytics
b. Tax data warehouses
c. Tax compliance data
d. Tax planning
9. is defined as predictive analysis of potential
tax liability and the formulation of a plan to reduce the
amount of taxes paid.
a. Tax data analytics
b. Tax data warehouses
c. Tax compliance data
d. Tax planning
page 373
10. is defined as the evaluation of the
impact of different tax scenarios/alternatives on various
outcome measures including the amount of taxable
income or tax paid.
a. Tax planning
b. What-if scenario analysis
c. Tax compliance
d. Data warehouse
1. Explain how the IRS might use social media data to
profile taxpayers who might be underpaying taxes.
What additional information would the IRS need to
consider in addition to social media data to build a full
taxpayer profile?
2. Why would a company be interested in documenting
the book-tax differences to identify potential items of
interest to the IRS?
3. Explain why the needs of the tax accountant are
different than the needs of the financial accountants.
Why does this lead to a tax data warehouse or tax data
mart?
4. Why would tracking a client’s unrealized capital gains
be important to businesses trying to capitalize on the
tax opportunities inherent in opportunity zones (a new
investment opportunity available as a result of the Tax
Cuts and Jobs Act of 2017)? How would accounting
firms access this data regarding their clients?
5. Why would employee turnover of the tax personnel be a
good KPI to track a company’s overall tax efficiency and
effectiveness? What does low employee turnover (as
compared to high turnover) allow a tax department to
do?
Problems
1. How do visualizations of tax compliance assist a
company in its efforts to reduce tax risk and minimize
the costs of tax preparation and compliance? In your
opinion, what would be needed to consistently make
visualizations a key part of the tax department
evaluation of tax risk and tax cost minimization?
2. How does tax planning differ from tax compliance? Why
might the company leadership be more excited about
the value-creating efforts of tax planning versus that of
tax compliance?
3. How do Data Analytics facilitate what-if scenario
analysis? How does the presence of a tax data mart
help with the needed data to support such analysis?
4. In your opinion, which of the four general categories of
tax KPIs mentioned in the text would be most important
to the CEO? Support your opinion.
5. Explain why tax sustainability would be of interest to the
tax department. What does it allow them to do if they
are able to gain tax sustainability versus what they are
not able to do without tax sustainability?
6. Descriptive analytics help calculate tax liability more
accurately. Give some examples of tax-related
descriptive analytics.
7. Predictive analytics help identify future tax liabilities.
What data would a tax accountant need in order to
perform a predictive analysis?
8. Explain how probability helps refine a what-if analysis.
9. Assume that a company has the option of staying in a
tax jurisdiction with an effective tax rate of 20 percent or
moving to a different location where the effective tax
rates are 11 percent and 4 percent. What other drivers
besides the tax rate may affect the decision to stay or
move?
page 374
10. If a company knows that the IRS will
change a tax calculation in the future, such as the
capitalization of research and experimental expense in
2021, what actions might management take today to
reduce their tax liability when the new policy goes into
effect?
11. Match the tax analytics definitions to their terms: data
mart, data warehouse, tax planning, tax data mart,
what-if scenario analysis.
Tax
Tax Analytics Definition Analytics
Term
A subset of the data warehouse focused on
a specific function or department to assist
and support its needed data requirements
A repository of data accumulated from
internal and external sources, including
financial data, to help management decision
making
Predictive analysis of potential tax liability
and the formulation of a plan to reduce the
amount of taxes paid
A subset of a company-owned data
warehouse focused on the specific needs of
the tax department
Evaluation of the impact of different tax
scenarios/alternatives on various outcome
measures including the amount of taxable
income or tax paid
page 375
Lab 9-1 State Sales Taxes and

Create a Data Visualization
Lab Summary
Since taxes vary by state, this lab teaches how to gather
the state sales tax data, and how to analyze and visualize
it.
Data
Ch 9 State_SalesTax.xlsx
Technique
Some experience with spreadsheets and simple
visualization in Tableau is useful for this lab.
Software needed
Excel
Screen capture tool (Windows: Snipping Tool; Mac:
Cmd+Shift+4)
Tableau
Part 1: Look up state sales tax on the Internet.

Part 2: Create a spreadsheet with the sales tax data in
Excel.
Part 3: Visualize and analyze sales tax data in
Tableau.

Q1. What advantage does a state have in charging
a relatively high state sales tax? In contrast,
what advantage does a state have in charging
relatively low (or no) sales tax? If one state
has a high sales tax and the state next door
has a low sales tax, would that affect your
answer to the first two questions?
Part 2: Master the Data: Prepare Data
for Analysis Using Excel
1. Open the Excel spreadsheet, Ch 9
State_SalesTax.xlsx.
2. You are going to create a frequency distribution and
histogram, and to do so you need to set up some bins.
Input the following set of numbers somewhere in the
Excel data file (Ch 9 State_SalesTax.xlsx) to use as the
bins for your frequency distribution and histogram:
0
0.019
0.038
0.054
0.073
Part 3: Perform an Analysis by

Creating a Frequency Distribution and
Histogram in Excel
3. Create a frequency distribution and histogram by
accessing the Data Analysis ToolPak from the Excel
ribbon.
page 376
4. Select Histogram from the Data Analysis window and

click OK.
5. Select the tax rate column for the Input Range.
6. Select the bins that you entered for the Bin Range.
page 377
7. Place a checkmark in the box next to Chart Output.
8. Click OK.
9. Take a screenshot of your frequency distribution and

the histogram. Label it “Lab 9-1 Histogram.”
Analyze the Results

Q2. Based on the histogram of the distribution of
tax rates, are you surprised by its shape where
there are observations on both extremes but
none in the middle? Why would some states
have zero sales tax, but most states have
higher sales taxes?
Part 3, continued: Perform an

Analysis Using Tableau
10. Open Tableau and connect to the Ch 9 Lab 9-1
State_SalesTax.xlsx file.
11. Ensure that the fields imported into Tableau in the
correct format—geographic string for State and number
for State Tax Rate.
page 378
12. Click into Sheet 1 and double-click on State from the

Dimensions section. Automatically, Latitude and
Longitude pills will appear in the Rows and Columns
shelves, and Tableau will default to a symbol map since
the State field is formatted as a geographic string.
13. Double-click on TaxRate from Measures area.
14. In the Show Me section of Tableau, change the map
from a symbol map to a filled map. If the shading in
each state does not automatically adjust to reflect the
different levels of tax rates, drag the pill SUM(State
Sales Tax) to Color on the Marks shelf.
15. Right-click on the label for Sheet 1 to rename the sheet

as Filled Map.
page 379
16. Take a screenshot of the filled map and label it “Lab 9-1
Filled Map.”
17. Right-click the label for Filled Map (previously labeled
Sheet 1) to duplicate the sheet.
18. Using the different options available in the Show Me
tab, create another visual that you find useful. Re-label
the sheet appropriately. If your Show Me tab has
disappeared, you can click Ctrl +1 on your keyboard for
it to reappear.
19. Take a screenshot of your new visual and label it “Lab
9-1 Visual.”
Q2. Based on this visualization, which states

paying 0 percent sales tax are next to states
with really high sales tax rates?
End of Lab
Lab 9-2 Comprehensive Case 1:

Dillard’s Store Data: Calculate Sales
Tax for Dillard’s States
Company summary
Wikipedia site on DDS. You’ll quickly note that William T.
Data
The data for this lab and other Dillard’s labs is available
at http://walton.uark.edu/enterprise/. Your instructor will
either give you specific instructions on how to access the
data, or there will be information available in Connect.
The 2016 Dillard’s data covers all transactions over the
period 1/1/2014 to 10/17/2016.
Additionally, the data found in the Ch 9
State_SalesTax.xlsx is used.
Software needed
the VMWare at the University of Arkansas)
Microsoft Excel

Calculate total sales across all Dillard’s stores year over
year in order to compare sales tax paid across periods.

What is an estimate of the amount of sales taxes we owe
from Dillard’s purchases? What is an estimate of the
amount of sales taxes we owe from Dillard’s net
purchases?

1. Before we can analyze the data and look up the sales
tax owed in each state, we need to extract the sales
data From SQL Server and load it into Excel. To do so,
you will need to connect to SQL Server data through
Excel using Excel’s Get & Transform tool. If you need a
refresher on how to use that tool, refer to the textbook’s
Appendix F.
page 380
2. Once you connect to the SQL Server
dataset in Excel and expand the options to input a
query, input the following SQL query to extract the data
needed for our analysis.
SELECT State, SUM(Tran_Amt) AS Amount
FROM Store
INNER JOIN Transact
ON Store.Store = Transact.Store
WHERE Tran_Type = ‘P’ AND YEAR(Tran_Date) =
2015 AND State <> ‘U’
GROUP BY State
3. Thirty rows (including the header) should load into your
spreadsheet.
4. Add a column to your SQL Server data table titled State
Sales Tax Rate.
5. Save your file as Ch_9_Comprehensive_Lab_1.xlsx.
6. Open the Ch 9 State_SalesTax.xlsx file and copy the
Sales Tax table to columns E and F of the file you just
created, Ch_9_Comprehensive_Lab_1.xlsx. You will
use this table to calculate the sales tax Dillard’s owed
in each state. You can find this file in the “DAA 2e Data
Files” folder on the remote desktop.
7. Use a VLookup to populate the new column that
matches the state sales tax with the sales amount for
each state.
The Lookup_Value is the state field.
The table_array is the entire Sales Tax table that you
copied into this Excel file. Refer to this either as Table1
(if made into a table) or refer directly to the cells in
columns E and F.
The col_index_number is 2.
The range_lookup value is FALSE.
8. Add a column to your SQL Server data table titled State

Tax Owed to calculate the sales tax owed in each state
in 2016 using the formula shown in the image below.
page 381
Part 2, Continued: Refining the Query

The query you entered in the previous steps included a
filter that only showed transactions that were marked with
a “P” for their tran_type, indicating a customer purchase.
However, if there were returns during the period, these
should also be included in the calculation so that the
dollar amounts associated with the returns are subtracted
from the total.
Q1. Why is it important to subtract the return
transaction amount when calculating sales tax
owed?
Re-run the query with the following text (which omits

the filter for only purchases), and then repeat the steps to
add the two additional columns (State Sales Tax Rate and
Tax Owed) to look up the total amount of sales tax owed
for each state, but now deducting the returns.
SELECT State, SUM(Tran_Amt) AS Amount
FROM Store
INNER JOIN Transact
ON Store.Store = Transact.Store
WHERE YEAR(Tran_Date) = 2015 AND State <> ‘U’
GROUP BY State
Q2. How vast is the difference in the amount of

total state sales tax Dillard’s would end up
paying if you didn’t correct the error by refining
the query (and reducing it by the amount of
returned items)?
Q3. Is it safe to assume that all of the returns made

in 2015 are aligned with the fiscal period for
which Dillard’s owes sales tax? For example,
would all returns in 2015 be from 2015 sales?
What should Dillard’s due to account for any
potential misalignment?
End of Lab

Tax for Dillard’s States Part 2—
Compare Year over Year
Company Summary
Data
data or there will be information available on Connect.
The 2016 Dillard’s data covers all transactions over the
period 1/1/2014 to 10/17/2016.
page 382
Software needed
Microsoft Excel

Calculate total sales across all Dillard’s stores year over
year and month over month in order to compare sales
tax paid across periods.
Compare 2014, 2015, and 2016 sales tax data across
states in parallel periods.

1. Before we can analyze the data, we need to extract the
data From SQL Server and load it into Excel. To do so,
click New Query from the Data tab, and follow the path
to select from Database and from SQL Server
Database.
page 383
2. To address this question, you will need to connect to

SQL Server data through Excel using Excel’s Get &
Transform tool. If you need a refresher on how to use
that tool, refer to the textbook’s Appendix F.
Select year(Tran_Date) as year,
datename(month,tran_date) as month, state,
sum(Tran_Amt) as SalesAmount
From TRANSACT
INNER JOIN STORE
WHERE State <> ‘U’
Group By year(Tran_Date),
datename(month,tran_date), STATE
Order By year(Tran_Date),
datename(month,tran_date), STATE
3. Once the query runs (it may take a few minutes), click
Load the data into Excel; 986 rows should load.
4. Now, we need to add two columns to this table, a State
Tax Rate column and a State Tax Owed column.
5. Use a VLookup and your state sales tax table to
populate the tax_rate column (replicate the steps 4-7
from Lab 9-2 Comprehensive Lab 1).
6. Multiply the State Tax Rate by the SalesAmount to
populate the State Tax Owed column.

7. From the insert tab on the ribbon, click PivotTable to
insert a PivotTable.
8. Ensure that the data from Query 1 is selected, and click

OK.
To view how state sales tax owed changed in each

state from 2014 to 2015, add the State field to the
PivotTable Rows, the Year field to the PivotTable columns,
and State Tax Owed to the PivotTable Values (ensure that
Tax Owed is summarized by SUM).
page 384
9. To limit the years viewed, click the drop-down next to

Column Labels on the PivotTable, and unselect the
checkmark next to 2016. Click OK to apply the filter.
page 385
Part 4: Address, Refine, and

Communicate the Results
The filtered data provides the information you need to see
the states that Dillard’s owed more (or less) state sales
tax in from 2014 to 2015, but it would be easier to
understand by cleaning up the data included and with a
visualization.
10. The Grand Totals don’t add important information for
our analysis in this PivotTable, so they need to be
removed. From the PivotTable Design tab on the
ribbon, click Grand Totals, and turn them off for rows
and columns:
11. To visualize the data, we’ll add sparklines next to each

state’s numbers. Place your cursor in cell D5 next to the
end of the data for Alabama (AL). From the Insert tab in
the ribbon, click Line from the Sparklines section to add
a Sparkline.
12. In the Create Sparklines box, select Alabama’s set of

data for the Data Range and click OK (the Location
Range should default to D5 because of where your
cursor was when you began inserting the sparkline).
13. Copy the sparkline all the way down your data so that
you can see how the State Tax Owed changed year
over year for each state.
page 386
Q1. Are total state sales taxes increasing or

decreasing each year from 2014 to 2015 in the
state of Missouri? The state of Ohio?
End of Lab

Tax for Dillard’s States Part 3—
Calculate City Tax and Compare Tax
Owed Year over Year
Company Summary
Data
data, or there will be information available on Connect.
Optionally, a data file for Arkansas’ cities sales tax is
included: Ch 9 Lab - Comprehensive Lab 3 Total Sales
Tax in AR Cities.xlsx
Software needed
Microsoft Excel

Calculate total sales across all Dillard’s stores in a
particular state (the example given in this lab is for
Arkansas) each January, year over year, in order to
compare sales tax paid in each city across periods.

Compare 2014, 2015, and 2016 sales tax data across
cities.

1. To address our question, you will need to connect to
Once you connect to the SQL Server dataset in
Excel and expand the options to input a query, input
the following SQL query to extract the data needed for
our analysis.
SELECT YEAR(Tran_Date) AS Year, State, City,
SUM(Tran_Amt) AS Amount
FROM Transact
INNER JOIN Store
page 387
ON Transact.Store = Store.Store
WHERE DATENAME(MONTH,Tran_Date) = ’January’
AND State = ’AR’
GROUP BY YEAR(Tran_Date), State, Store.City
ORDER BY YEAR(Tran_Date), State, Store.City
Notice that this query text has a filter for state =
‘AR’. If you would prefer to look up city tax in a
different state, simply change the criteria to any
state’s abbreviation you choose.
2. Using your web browser, search for the city tax in the
state that you are filtering for. Typically, the tables for
the popular cities in each state are fairly easy to find,
and that will likely suffice. Alternatively, a clean version
of City Sales tax data for the state of Arkansas is
provided that you can use (File name: Ch 9 Lab -
Comprehensive Lab 3 Total Sales Tax in AR
Cities.xlsx).
3. Using your ETL skills that you learned in Chapter 2,
clean the city tax data in Excel so that you can use a
VLookup to calculate the city tax owed in each of the
state’s cities that have Dillard’s stores.
4. Following the same steps that you learned in Lab 9-2,
steps 6-8, calculate the total tax owed in each of the
state’s cities that have Dillard’s stores.
Part 3: Perform the Analysis and

Communicate Results
5. Using the skills you learned in Lab 9-3, steps 8-12,
create a PivotTable to compare the total sales tax owed
in each city across the three years of data. Then, create
sparklines to visualize the change over the three years.
Q1. Are sales taxes increasing or decreasing each

year from 2014 to 2016 in the city of Little
Rock? Fayetteville?
End of Lab

Dillard’s Store Data: Does a State’s
Tax Rate Affect Dillard’s Decision to
Open Stores There?
Company Summary
Data
data, or there will be information available on Connect.
Software needed
Microsoft Excel
page 388

Run a regression to test to see if a state’s sales tax rate
influenced their decision to place stores in that state.

Does a state’s sales tax rate have a significant impact on
Dillard’s decision to open stores there?

1. To address our question, you will need to connect to
Once you connect to the SQL Server dataset in
Excel and expand the options to input a query, input
the following SQL query to extract the data needed for
our analysis.
SELECT State, COUNT(Store) as NumStores
FROM Store
WHERE State <> ’U’
GROUP BY State
After loading the data, you should see 30 rows
(including a header).
2. Following the same steps that you learned in Lab 9-2,
steps 6-8, add a column to include the sales tax rate for
each state.

3. To run a regression to see if a state’s sales tax rate
helps to explain how many stores Dillard’s would have
placed in a given state, first click into the Data Analysis
ToolPak in the Data Tab.
4. In the Data Analysis window, Select Regression and

click OK.
5. Select the NumStores column of data for your Y Range
and the State sales tax rate column of data for your X
Range.
page 389
Remember that when you select the
labels in your Y and X ranges, place a check mark next
to Labels, as shown in the visual.
6. Click OK to run the regression.
Part 4: Address and Refine the

Results: Interpret R-Square and P-
Value
Q1. The coefficient for state sales tax is positive
but not significantly different from zero. What
does that suggest? Is the coefficient different
than what you would expect? Why or why not?
Q2. The adjusted R-squared tells us how much of

the variance is explained by this model. The
adjusted R-squared is less than zero,
suggesting it doesn’t do a very good job. What
do you think would do a better job in predicting
the number of stores in a state than sales tax
rate?
End of Lab
1https://washington.cbslocal.com/2014/04/16/report-irs-data-mining-facebook-twitter-
instagram-and-other-social-media-sites/,accessed August 2018.
2https://www.forbes.com/sites/jenniferpryce/2018/08/14/theres-a-6-trillion-opportunity-in-
opportunity-zones-heres-what-we-need-to-do-to-make-good-on-it/#527391d46ffc,
accessed August 15, 2018.
3“Defining Success: What KPIs Are Driving the Tax Function Today” PwC, September 2017,
https://www.pwc.com/gx/en/tax/publications/assets/pwc_tax_function_of_the_future_tax_f
unction_KPI_sept17.pdf, accessed August 14, 2018.
page 390
Chapter 10
Project Chapter (Basic)

This chapter will take you through a series of problems to help you
analyze and communicate answers to accounting questions that are
asked every day. This will provide a review of the Data Analytics
concepts we’ve discussed in the previous chapters and put them into
perspective. For each analysis, we will look at the data from a
managerial, auditing, and financial accounting perspective.
A Look Back
Chapter 9 discussed the application of Data Analytics to tax
questions and looked at how data can drive strategy and help
managers understand the underlying tax behavior.
A Look Forward
Chapter 11 will revisit the Dillard’s sales and returns data to provide
an advanced overview of different analytical tools and techniques to
provide additional understanding of the data.
page 391
Source: Microsoft PowerBI
Tools like Tableau and Power BI are popular because they enable quick analysis of
simple descriptive and diagnostic analytics. By creating visual answers to data
problems, accountants can tell stories that help inform management decisions, aid
auditors, and provide insight into financial data.
Both Tableau and Power BI enable more simplified analysis by incorporating
natural language processing into their cloud-based offerings. Instead of dragging
dimensions and measures to build the analyses, you can simply ask a question in a
natural sentence, and the tool will map your question to your existing data model.
OBJECTIVES
LO 10-1 Analyze data in the order-to-cash process.

LO 10-2 Analyze data in the procure-to-pay process.
page 392
EVALUATING BUSINESS PROCESSES

As a manager, auditor, or financial accountant, your role is to
understand how different business processes operate and ensure
controls exist over those processes. Companies use Data Analytics
to summarize data for reports, evaluate performance, and identify
risk in these cycles.
In this basic project chapter, we will work through a series of
questions that help you understand how data from different aspects
of each business process can answer a variety of questions
depending on the user’s perspective.
There are two main question sets that explore the following:
1. Question Set 1 looks at the order-to-cash process or sales/revenue
cycle within a company; you will summarize flows of sales order
transactions, accounts receivables, and customer activity.
2. Question Set 2 moves into the procure-to-pay process or
purchasing cycle to evaluate purchasing activity and potential
savings when paying or interacting with vendors.
Each set walks you through the analysis using a prepared
Tableau workbook. The data has already been cleaned and
converted to mimic the audit data standard format. As you complete
the steps, you will encounter several questions that ask you about
your approach to the analysis, how to interpret the results, and how
you would expand the analysis using a comprehensive data set.
QUESTION SET 1: ORDER-TO-CASH
LO
10-1
Analyze
data in
the
order-to-
cash
process.
The order-to-cash (O2C) process or sales cycle involves three main

processes:
1. Sales order processing.
2. Order fulfillment and shipping.
3. Billing and cash collections.
Managers are concerned with making the process as efficient as
possible to ensure increased sales volume, sufficient profitability, and
fast cash collection.
Auditors should test sales transaction and master data to ensure
that only authorized users are processing orders, that sales prices
match master data and aren’t altered, and that customers aren’t
exceeding approval limits in addition to evaluating prompt payment
by customers through the aging of accounts receivable.
Financial accountants are interested in determining the amount of
sales revenue on the income statement and accounts receivable
balance on the balance sheet as well as the calculation of bad debts
expense.
To answer the following questions, use the Tableau Workbook
O2C found on Connect. The O2C data, shown in Exhibit 10-1, has
joined together the following tables into one file.
EXHIBIT 10-1
Order-to-Cash Data
page 393
When you open the O2C Tableau Workbook, you will see the
data pre-populated and ready for analysis, similar to Exhibit 10-2.
You are now ready to complete the questions in this set.
EXHIBIT 10-2
O2C Data Ready for Analysis in Tableau

Question 1.1: How Efficiently Are We
Collecting Our Cash?
Before you begin, put yourself in a manager’s shoes, and answer the
following questions about the order-to-cash process. You might want
to discuss these with a classmate, review Chapter 7, or search the
Internet for some suggestions.
Q1-1 What risks are present if you take too long to collect our
accounts receivable?
Q1-2. What are some analyses you could perform that would
provide insight into how efficiently your company is
collecting cash from customers? Are there any KPIs that
would be appropriate here?
Q1-3. In your opinion, what would be an appropriate
benchmark for the average number of days sales
outstanding (i.e., Accounts receivable/Sales × 365)?
Would management want this number higher or lower?
page 394
Now let’s go to Tableau to find some answers about sales:

1. Open the Tableau workbook called O2C.twbx.
2. Rename Sheet 1 to Total Sales.
3. On the Sales sheet, drag Sales Order Amount measure to the Text
button in Marks pane.
4. Drag the Sales Order Date dimension to the Filters shelf. Choose
Years, then click Next, check 2020, and click OK.
5. Drag the Sales Order Date dimension to the Rows pane.

Q1-4. What are the total sales for 2020?
6. Click the + next to YEAR(Sales Order Date) in the Rows pane to
show the quarters.
Q1-5. Which quarter had the highest sales?
7. Create a new worksheet called Sales by Customer.
8. Drag the Sales Order Amount measure to the Columns shelf.
9. Drag the Customer Account Name dimension to the Rows shelf
and click Add All Members if prompted.
0. In the bar chart, hover over the Sales Order Amount title and click
the sort button to show sales by customer.
Q1-6. Which customer had the most sales? How much did we sell
to that customer?
page 395
Now let’s find some answers related to accounts receivable:

1. Create a new worksheet called AR by Customer.
2. We need to create some calculations and parameters to answer
questions about receivables:
a. Create a new parameter to set your reporting date. Click the down
arrow next to Dimensions and choose Create Parameter. . . Name
the parameter Report Date, set the data type to Date and the
current value to 12/31/2020 (choose this date from the calendar
pop up), and click OK.
b. Now add a calculated field to determine the age of accounts. Click

the down arrow next to Dimensions and choose Create
Calculated Field. . . Name the field Age, enter the equation
[Report Date]-[Invoice Due Date], and then click OK.
c. Create a new calculated field to determine the account balance.

Name the field Balance and enter the equation [Invoice Amount]-
IFNULL([Receipt Amount],0)-IFNULL([Adjustment Amount],0). The
IFNULL will place a 0 value in for invoices without matching
receipts. Click OK.
3. Now we’re ready for the analysis. Drag the Customer page 396
Account Name (if prompted, choose Add All
Members), Invoice ID, and Invoice Due Date dimensions to the
Rows shelf. To show the actual date, right-click the YEAR(Invoice
Due Date) and choose Measure > Maximum from the menu.
4. Drag the Measure Names dimension to the Columns shelf and the
Measure Values measure to the Text button in the Marks card.
5. In the Measure Values pane, remove all of the measures except
SUM(Adjustment Amount), SUM(Age), SUM(Balance),
SUM(Invoice Amount), and SUM(Receipt Amount). Drag and
drop the items to reorder them as Invoice Amount, Receipt Amount,
Adjustment Amount, Balance, and Age.
6. Drag the Balance measure to the Filters shelf. Click Next, then set
the minimum value to 1 and click OK. Now you will see a list of all
outstanding invoices.
Q1-7. What is the value of the the first outstanding invoice? What
is its age?
Q1-8. Some of the accounts have a negative age. What does that
mean?
Finally, let’s determine how old the accounts receivable are on
average.
7. Create a new worksheet called DSO to calculate the days sales
outstanding KPI.
8. Create a new Calculated Field called Days Sales Outstanding.
The equation should be SUM([Balance])/SUM([Sales Order
Amount])*365. Click OK.
9. Drag the Days Sales Outstanding measure to the Text button in
the Marks card.
Q1-9. What is the current days sales outstanding KPI value? What
does it mean?
Question 1.2: Is the Delivery Process

Following the Expected Procedure?
As an auditor, you’re interested in determining whether the delivery
process follows the expected sequence. Specifically, does the
delivery follow the sales order, and has each delivery been matched
with an invoice? Before you continue with your analysis, answer the
following questions about the delivery process.
Q1-10. Under what circumstances might a delivery take place
before a sales order? Should this happen?
Q1-11. What types of controls would prevent the system from
skipping a process or step?
0. Return to your O2C Tableau Workbook and create a new worksheet
called Exceptions.
1. Create a new calculated field called Order to Ship Days with the
equation [Shipment Date]-[Sales Order Date] and click OK.
2. Drag Order to Ship Days to the Label button in the Marks pane
and Sales Order ID to the Rows shelf (if prompted, click Add All
Members).
3. Click the Sort button next to the Order to Ship Days two times to
sort by ascending order.
Q1-12. Which orders were created after shipment? How do you
know?
4. Now create a new worksheet called No Invoice to determine
whether any orders have shipped but have not been invoiced yet.
5. Drag Sales Order ID, Shipment Date, Shipment ID, and Invoice
ID to the Rows shelf (choose Add All Members if prompted).
6. Right-click Shipment Date and choose Measure > Maximum from
the menu.
7. Drag Invoice ID to the Filters shelf. Choose Select from list, click
None, then check Null, and click OK. page 397
Q1-13. Of the 12 orders listed, which one is the most problematic?

Why?
Q1-14. Why aren’t the other orders suspicious?
Q1-15. While you still have your auditor hat on, what are some
additional analyses you could perform to understand
whether the process or processes are being followed or
controls are functioning properly?
page 398
Question 1.3: What Is the Total Revenue and

Balance in Accounts Receivable?
Finally, imagine you are preparing financial statements and need to
calculate the revenue and accounts receivable balances for the
income statement and balance sheet, respectively.
Q1-16. How could analytics provide additional insight into
financial information beyond calculating balances?
8. Return to your O2C Tableau Workbook and create a new worksheet
called Aging Report. This sheet relies on the Balance calculated
field you created previously.
9. Create buckets for the day categories. Right-click the Age measure
and choose Create > Bins from the menu.

0. Leave the new field name, set the size of bins to 30, page 399
and click OK.
1. Drag the Balance measure to the Columns pane and the Age (bin)
dimension to the Rows shelf.
2. Drag the Balance measure to the Label button in the Marks pane to
show the value of each bucket alongside the bar chart.
Q1-17. What is the value of open accounts that are between 60
and 90 days old? Hint: Look for the 60 bin.
Q1-18. Should we be writing any of these accounts off, in your
opinion?
3. Finally, let’s evaluate write-offs. Create a new worksheet called Bad
Debts.
4. Create a new Calculated Field called Write-off Percent with the
equation SUM([Adjustment Amount])/SUM([Sales Order Amount]).
This will show the percent of sales order dollars that were written off
during the year.
5. Drag Sales Order Date dimension to the Columns shelf. Right-click
the Sales Order Date and choose Month from the menu.
6. Drag Sales Order Amount and Adjustment Amount measures to
the Rows shelf.
7. In the Marks card, under All, change the chart type from Automatic
to Bar.
8. Right click the Adjustment Amount label in the graph and choose
Dual Axis to combine the charts.
9. Right click the Adjustment Amount label again and choose

Synchronize Axis to place the numbers on the same scale.
0. Finally, drag the Write-off Percent measure to the Label button in
the Marks card to add the percentage value to the write-offs.
1. Right click the AGG(Write-off Percent) in the Marks card and click
Format from the menu. In the formatting pane, change the default
numbers from Automatic to Percentage.
Q1-19. If management expects about 5 percent of the accounts
receivables to be written off, which month(s) exceeded that
estimate?
Q1-20. Looking at the trend of write-offs for the end of the year, do
you expect there to be no write-offs in November and
December? Why or why not? How would you account for
these?
page 400
Question 1.4: What Else Can You Determine

about the O2C Process?
We’ve discussed a few different ways to analyze O2C data to
understand the processes and controls. Now it’s your chance to find
answers to your own questions.
2. Identify five questions that you think management or auditors would
want to know about the O2C process. If you need help, search for
some common questions asked by accountants on the Internet.
3. Using the the data you have already loaded into Tableau, generate
at least three analyses and visualizations that will help you find the
answers to your five questions.
Q1-21. Write your first question and provide an answer based on
your analysis.
Q1-22. Write your second question and provide an answer based
on your analysis.
Q1-23. Write your third question and provide an answer based on
your analysis.
Q1-24. Write your fourth question and provide an answer based on
your analysis.
Q1-25. Write your fifth question and provide an answer based on
your analysis.
QUESTION SET 2: PROCURE-TO-PAY

LO
10-2
Analyze
data in
the
procure-
to-pay
process.
The procure-to-pay process (P2P) or purchasing cycle for a retailer
involves four main processes:
1. Create and submit a purchase order.
2. Receive inventory.
3. Receive an invoice.
4. Pay the invoice.
The procure-to-pay process has some additional challenges in
that there are numerous opportunities to divert company funds.
Therefore you should focus on the risk of unauthorized and fictitious
payments (e.g., to shell vendors) and ensure that the process is
appropriately controlled.
Managers would want to ensure that inventory matches what
they (or their company) ordered, that invoices aren’t paid more than
once, and that the payments are sent to approved parties.
Auditors are interested in testing the internal controls that
govern who can create orders, receive items, and approve
payments. In addition to segregation of duties, they may be
interested in matching each of the documents (i.e., purchase order,
receiving report, and vendor invoice) in a three-way match.
To answer the following questions, use the Tableau workbook
and Excel data files found on Connect. The P2P data, shown in
Exhibit 10-3, has joined together the following tables into one file.
When you open the P2P Tableau Workbook, you will see the data
pre-populated and ready for analysis, similar to Exhibit 10-4. You are
now ready to complete the questions in this set.
Question 2.1: How Long Are We Taking to

Pay Our Invoices?
Before you begin, put yourself in a manager’s shoes, and answer the
following questions about the procure-to-pay process. You might
want to discuss these with a classmate, review Chapter 7, or search
the Internet for some suggestions.
page 401
EXHIBIT 10-3
Procure-to-Pay Data
EXHIBIT 10-4
P2P Data Ready for Analysis in Tableau

page 402
Q2-1. What risks are present if you take too long to pay our
accounts payable?
Q2-2. What are some analyses you could perform that would
provide insight into how efficiently your company is
processing payments to suppliers? Are there any KPIs
that would be appropriate here?
Q2-3. In your opinion, what would be an appropriate
benchmark for the average number of discount dollars
lost as a percentage of available discount dollars? How
about erroneous payments as a percentage of total
payments? Would management want these numbers to
be higher or lower?
Now let’s go to Tableau to find some answers about purchases

and payments:
1. Open the Tableau workbook called P2P.twbx. If prompted for a data
file, choose the P2P Data.xlsx file that you downloaded from
Connect. Alternatively, choose the Purchase_Orders+ (P2P Data)
option and click OK.
2. Rename Sheet 1 to Total Purchases.
3. On the Total Purchases sheet, drag the Purchase Order Amount
Local measure to the Text button on the Marks card.
4. Drag the Purchase Order Date dimension to the Filters shelf.
Choose Years, then click Next, check 2020, and click OK.
5. Drag the Purchase Order Date dimension to the Rows shelf.

Q2-4. What are the total purchases for 2020?
6. Click the + next to YEAR(Purchase Order Date) in the Rows shelf
to show the quarters.
Q2-5. Which quarter had the highest total purchases?
7. Create a new worksheet called Purchases by Supplier.
8. Drag the Purchase Order Amount Local measure to the Columns
shelf.
9. Drag the Supplier Account Name dimension to the page 403
Rows shelf and click Add All Members if prompted.
0. In the bar chart, hover over the Purchase Order Amount Local
title and click the sort button to show purchases by supplier.
Q2-6. Which supplier did we purchase the most from? How much
did we purchase from that supplier?
Now let’s find some answers related to accounts payable:
1. Create a new worksheet called AP by Supplier.
2. We need to create some calculations and parameters to answer
these questions:
a. Create a new parameter to set your reporting date. Click the down
arrow next to Dimensions and choose Create Parameter. . . Name
the parameter Report Date, set the data type to Date and the
current value to 12/31/2020, and click OK.
b. Now add a calculated field to determine the age of accounts. Click

the down arrow next to Dimensions and choose Create
Calculated Field. . . Name the field Age, enter the equation
[Report Date]-[Invoice Due Date], and then click OK.
c. Create a new calculated field to determine the account balance.

Name the field Balance and enter the equation [Invoice Amount]-
IFNULL([Payment Amount],0). The IFNULL will place a 0 value in
for invoices without matching payments. Click OK.
3. Now we’re ready for the analysis. Drag the Supplier page 404
Account Name, Invoice ID, and Invoice Due Date
dimensions to the Rows shelf. If prompted, add all members. To
show the actual date, right-click the YEAR(Invoice Due Date) and
choose Measure > Maximum from the menu.
4. Drag the Measure Names dimension to the Columns shelf and the
Measure Values measure to the Text button in the Marks card.
5. In the Measure Values pane, remove all of the measures except
SUM(Age), SUM(Balance), SUM(Invoice Amount), and
SUM(Payment Amount). Reorder them to show the Invoice
Amount, Payment Amount, Balance, and Age.
6. Drag the Balance measure to the Filters shelf. Click Next, then set
the minimum value to 1 and click OK. Drag the Payment Amount
measure to the Filters shelf. Click Next, then set the minimum and
maximum values to 0 and check Include Null Values and click OK.
Now you will see a list of all outstanding invoices.
Q2-7. What is the invoice amount of the first outstanding invoice
for Danbam? What is its age?
Q2-8. Some of the accounts have a negative age. What does that
mean?
Finally, let’s determine how old the accounts payable are on
average.
7. Create a new worksheet called DPO to calculate the days payable
outstanding KPI.
8. Create a new Calculated Field called Days Payable Outstanding.
The equation should be SUM([Balance])/SUM([Purchase Order
Amount Local])*365. Click OK.
9. Drag the Days Payable Outstanding measure to the Text button in
the Marks card.
Q2-9. What is the current days payable outstanding KPI value?
What does it mean?
Question 2.2: Are There Any Erroneous

Payments?
Auditors would be interested in evaluating the origin of invoices and
payables to make sure that they are paid correctly and aren’t out of
normal behavior. Before you continue your analysis, answer the
following questions about the payment process.
Q2-10. What statistical tools can we use to diagnose behavior
that is outside of normal behavior?
Q2-11. How might an outlier be used to focus the auditors on
high-risk transactions?
Let’s look at the data and use an average Z-score value to
determine which suppliers are receiving an abnormally high amount
of purchases.
0. Return to your P2P Tableau workbook and create a new worksheet
called Outliers.
1. Create a new calculated field called Average Purchases with the
equation WINDOW_AVG(SUM([Purchase Order Amount Local])).
2. Create another calculated field called Std Dev Purchases with the
equation WINDOW_STDEVP(SUM([Purchase Order Amount
Local])).
3. Finally create a calculated field called Z-Score Purchases with the
equation (SUM([Purchase Order Amount Local])-[Average
Purchases])/[Std Dev Purchases].
4. Drag Z-Score Purchases to the Columns shelf and Supplier
Account Name to the Rows shelf.
5. Right-click Z-Score Purchases in the Columns shelf and choose
Compute Using > Supplier Account Name.
6. Hover over the Z-Score Purchases label in the x-axis of the graph
and click Sort Descending.
Q2-12. Which supplier(s) had an abnormally high dollar amount of
purchases? What do you consider “abnormally high”? Are
these suspicious?
7. Now create a new worksheet called No Order to determine if any
invoices have been received that don’t match existing orders.
8. Drag Invoice ID and Purchase Order ID (from page 405
Invoices Received) to the Rows shelf. Click Add All
Members if prompted.
9. Drag the Purchase Order ID (from Invoices Received) to the
Filters shelf. Choose Select from the list, click None, then click
Null, and click OK.
Q2-13. Of the invoices received, which are the most problematic?
Why?
Q2-14. Why aren’t the other purchases suspicious?
Q2-15. While you still have your auditor hat on, what are some
additional analyses you could perform to understand
whether the purchase process is being followed or controls
are functioning properly?
Question 2.3: Are We Missing Out on

Discounts by Paying Late?
Finally, imagine you are being offered discounts on your payments. If
you pay within the discount period, you may receive up to 5 percent
off of the invoice amount.
Q2-16. When might a large company prefer to forfeit discounts
on its invoices?
For this step to work properly, you’ll need to adjust the data
model to show only invoices that have been paid.
0. Return to your P2P Tableau workbook and save a copy of the
workbook as P2P Paid.twbx. If Tableau asks, choose Save as
Extract.
1. Click the Data Source tab. If Tableau prompts you for a data file,
browse to the P2P Data.xlsx file you downloaded from Connect.
2. Click the join between Invoices_Received and
AP_Cash_Application and change the join type from a Left Join to
an Inner Join.
3. Now create a new worksheet called Forfeited Discounts.

4. Create a new calculated field called Discounts Available with the
calculation SUM([Invoice Amount]*[Terms Discount Percentage
(Invoices!Received)]).
5. Create a new calculated field called Discounts Taken with the
calculation SUM([Invoice Amount]-IIF(ISNULL([Payment Amount]),
[Invoice Amount], [Payment Amount])).
6. Create a new calculate field called Discounts page 406
Forfeited with the calculation [Discounts Available]-
[Discounts Taken].
7. Create another calculated field called Forfeited Ratio with the
calculation [Discounts Forfeited]/[Discounts Available].
8. Drag Discounts Available and Discounts Taken to the Columns
shelf and Discounts Forfeited to the Detail button in the Marks
card. They will become aggregate functions (AGG) to show the total
amounts by supplier.
9. Drag the Supplier Account Name to the Rows shelf.
0. Drag Forfeited Ratio to the Detail button in the Marks card.
1. In the Marks card, under All, change the chart type from Automatic
to Bar.
2. Right-click the Discounts Taken label in the graph and choose
Dual Axis to combine the charts.
3. Right-click the Discounts Taken label again and choose
Synchronize Axis to place the numbers on the same scale.
4. Sort by Discount Available.
Q2-17. How much discount (in dollars) have we forfeited from our
top three suppliers?
Q2-18. Should we adjust our policy to pay them more quickly, in
your opinion?
Q2-19. What alternative options could we use to reduce our
accounts payable, in your opinion?
Question 2.4: What Else Can You Determine

about the P2P Process?
We’ve discussed a few different ways to analyze P2P data to
understand the processes and controls. Now it’s your chance to find
answers to your own questions.
5. Identify five questions that you think management or auditors would
want to know about the P2P process. If you need help, search for
some common questions asked by accountants on the Internet.
6. Using the the data you have already loaded into Tableau, generate
some analyses and visualizations that will help you find the answers
to your questions.
Q2-20. Write your first question and provide an answer based on
your analysis.
Q2-21. Write your second question and provide an answer based
on your analysis.
Q2-22. Write your third question and provide an answer based on
your analysis.
Q2-23. Write your fourth question and provide an answer based on
your analysis.
Q2-24. Write your fifth question and provide an answer based on
your analysis.
page 407
page 408
Chapter 11
Project Chapter (Advanced):
Analyzing Dillard’s Data to Predict
Sales Returns

Similar to the previous chapter, this chapter will take you through a
series of problems to help you analyze and communicate answers to
typical accounting questions related to predicting sales returns. To
answer these questions, we will return to the Dillard’s data set found
on the University of Arkansas Remote Desktop to explore the data
as it relates to understanding and predicting sales returns. After
exploring the data, we will work through a series of questions to
describe the state of sales returns, diagnose why the returns are the
way they are, and predict future sales returns based on a selection
of explanatory variables. This will provide a review of the Data
Analytics concepts we’ve discussed in the previous chapters and put
them into perspective.
A Look Back
Chapter 10 had a project chapter that emphasized basic data
analytic skills related to the order to cash and purchase to pay
processes.
page 409
Chris Salata/ZUMA Press/Newscom.
Retail’s Ticking Time Bomb: Sales (Particularly Online Sales) Returns

Americans returned $260 billion in merchandise to retailers in 2015. While the
average rate of return at retailers is just 8 percent, it increases on average to 10
percent during the holiday sales returns. However, it increases dramatically for online
sales—to 30 percent or higher with clothing returns from online sales hitting 40
percent. With a much higher return rate, as online retailers such as Amazon
continue to increase their market share of total retail sales, it’s only going to get
worse.
What’s more? Not only is product being returned in greater numbers, but the
value of the unwanted and damaged returns is greatly diminished:
Unwanted and damaged goods either get tossed out or sent through a lengthy
chain of liquidators and wholesalers, paying pennies on the dollar to the
retailer before eventually selling them to bargain-hunting consumers.*
Because accountants are required to estimate sales returns (and the diminished
value of returned items), and offset sales in the same period that the original sales
are made, accountants need to establish a reasonable and hopefully reliable method
to estimate such returns. This chapter establishes various descriptive, diagnostic,
and predictive analytics that may be used to help evaluate the estimate of sales
returns.
Source: https://www.cnbc.com/2016/12/16/a-260-billion-ticking-time-bomb-the-costly-
business-of-retail-returns.html, accessed April 2019.
https://www.forbes.com/sites/stevendennis/2018/02/14/the-ticking-time-bomb-of-e-
commerce-returns/#46d599754c7f, accessed April 2019.
*Source: CNBC LLC
OBJECTIVES
LO 11-1 Analyze returned sales to find explanatory

(independent) variables using descriptive and
exploratory analytics.
LO 11-2 Illustrate hypothesis testing using diagnostic
analytics to compare and contrast sales returns
around the holiday season, as well as comparing
sales returns in Arkansas (where Dillard’s is based)
and the rest of the country.
LO 11-3 Predict returned sales in future periods using
predictive analytics.
page 410
ESTIMATING SALES RETURNS

The recent revenue recognition standards1 increased the emphasis
on valid, reasonable estimates of sales returns matched to the same
time period the original sale was made. Companies must assess
whether their models and methods of estimating returns are
appropriate.
In this chapter, we will work through a project to describe
(Question Set 1), diagnose (Question Set 2), and predict sales
returns for Dillard’s (Question Set 3). Using these various analyses,
we develop a potential model useful in predicting Dillard’s sales
returns and test it for reasonableness.
There are three main question sets:
1. Question Set 1 focuses on exploring the sum of returns and the
percentage of returned sales by state, product, and month using
Tableau data visualizations.
2. Question Set 2 continues the analysis with hypothesis tests to see if
the percent of sales returned is significantly higher in Arkansas
(where Dillard’s is headquartered) versus other states, as well as
to see if the percent of sales returned is significantly higher during
the holiday season (December and January) than any other time of
the year.
3. Question Set 3 focuses on exploring how historical data can help
predict the future percentage of returned sales through PivotTables,
PivotCharts, and regression testing in Excel.
Each question set has a set of instructions to guide you through
mastering the data, performing the analysis, and communicating
your results.
QUESTION SET 1: DESCRIPTIVE AND

EXPLORATORY ANALYSIS
LO
11-1
Analyze
returned
sales to
find
explanat
ory
(indepen
dent)
variables
using
descripti
ve and
explorat
ory
analytics
.
In this question set, we use analytics to find explanatory variables.
Question 1.1: Which Attributes Could Help

Predict Percentage of Returned Sales?
To answer this question, we will analyze the Dillard’s database ER-
Diagram in Microsoft SQL Server and run some short queries to
explore the data.
1. Look at the ER-Diagram for Dillard’s Data in Appendix J
WCOB_DILLARDS ER Diagram.
Q1-1. Which of the attributes listed seem like they might be
interesting as explanatory variables to help Dillard’s explain
and plan for percentage of returned sales?
Dive into the Tables: Create Queries to See What Sort of

Data Each Attribute Represents
To get a better idea of what data is represented by each of the
attributes, we can run preview queries to see the top 10 rows of each
table. If you previously completed Lab 2-7, you explored the data in
a similar way. Getting a preview of just the top 10 rows in a table is
helpful because the queries can run quickly (a benefit page 411
when the tables hold a massive amount of data—
running a query to return all of the records would take quite a long
time!).
2. Ensure that you are connected to the University of Arkansas
Remote Desktop. Open Microsoft SQL Server Management Studio
to access the WCOB_Dillards data and input the Server Name in
the Connect to Server window that your instructor provided you.
(Leave the default for authentication to Windows Authentication,
and click Connect.)
Source: SQL
3. To write a new query, select New Query from the menu at the top of
the SQL Server application. Ensure that the database selected is
WCOB_DILLARDS.
Source: SQL
4. To view the top 10 rows in the TRANSACT table, type the following
query into the query window:
SELECT TOP 10 *
FROM TRANSACT
5. To see the result of the query, click Execute. F5 also works to run
queries as a PC shortcut.
Source: SQL
6. Explore the STORE, SKU, and DEPARTMENT tables with queries

to preview the first 500 rows, to get a better feel for the data.
Q1-2. What benefit can you gain from selecting only the top few
rows of your data, particularly from a large dataset?
page 412
Q1-3. After looking at the top 10 rows in the DEPARTMENT table,
what do you think the attributes DEPTCENT_DESC,
DEPTDEC_DESC, and DEPT_DESC represent? How are
the three attributes different from one another?
Q1-4. After executing these queries, which attributes do you think
would be useful to help understand percentage of returned
sales?
Question 1.2: How Can We Explore the

Product Hierarchy Through Data
Visualization?
The three attributes that you discovered in the DEPARTMENT table,
DEPTCENT_DESC, DEPTDEC_DESC, and DEPT_DESC,
represent “century,” “decade,” and “department,” and it is the way
Dillard’s organizes its product hierarchy. Its product hierarchy
contains Department Century (example: Children’s), Department
Decade (Example: Basics, girls, infants, seasonal, etc.), and
Department (example: girls coats, girls swim, boys coats, etc.). The
individual products are only identified by SKU and Item ID (not
description).
We can use Tableau to explore which centuries, decades, and
departments have the highest average percentage of returned sales
in our sample period.
Explore and Visualize Data in Tableau
Analyze Returns by Product and State

7. While still in the UArk system, open Tableau.
8. Because the raw data is stored in SQL Server, click Microsoft SQL
Server, then enter the Server and Database information associated
with the Dillard’s data.
9. We can work with data from several tables at once. Drag the
following tables into the Data source window to work with them in
Tableau:
TRANSACT
SKU
STORE
DEPARTMENT
0. Click Sheet 1 to begin working with your data.
1. Double-click on the Measure Sale Price and the Dimension Tran

Type to get a glimpse of the grand totals of returns and purchases
in the dataset.
Tableau responded to your double-click with three defaults: it placed
the measure (Sale Price) in the Rows and the dimension (Tran Type)
in the Columns, it created a bar chart, and it defaulted the aggregate
measure to SUM. All of these defaults are sufficient for the level of
analysis we are doing at this stage.
Q1-5. We can drill down into this grand total to make a more
meaningful bar chart. Before we do so, make a
prediction about which types of products get returned the
most frequently to Dillard’s.
page 413
2. To focus on only returns, we can add Tran Type to the Filters shelf.
Remove the checkmark in the box next to P so that the only data
we see are Returns data, and click OK.

3. Once this filter is created, you can apply it to every worksheet in this
Tableau workbook. Right-click the Tran Type filter pill and select
Apply to Worksheets. From the options provided, select All Using
Related Data Sources.
4. Double-click the Deptcent Desc dimension. This is the page 414

top of the product hierarchy; it contains the broadest
categories. Deptcent Desc should be automatically added to the
Columns after Tran Type.
5. We can make this easier to read by swapping the rows and
columns and by sorting the data. These buttons are located near
each other near the top of the Tableau window. Click each of them.
Q1-6. Which two Centuries have the highest sum of returned

sales?
6. Right-click Sheet 1 to rename this sheet. Name it Returns by
Century.
7. Right-click the tab (now named Returns by Century) page 415

again to duplicate it.
This will allow you to modify the same bar chart by adding more
detail while still preserving the original chart.
8. Rename the new sheet (defaulted to being named Returns by
Century (2)) Sandbox so that it is recognizable as a sheet for
exploring.
9. Double-click the Deptdec Desc dimension to further drill down into
the product details. Deptdec Desc will be added to the Rows shelf.
0. Continue working with the data in a variety of ways to explore it.
Removing DeptCent Desc provides you a way to look at the returns
by Decade without the hierarchy. Adding Dept Desc provides even
more detail. You can also click the Show Me tab to change the way
your data is visualized.
Q1-7. Write down three insights that you have gained from
exploring the data in the Sandbox sheet.
1. Create a new sheet by clicking the icon to the right of the current
sheet you are working on:
2. Double-click the dimension State to create a map. page 416

Tableau will default to a symbol map with a symbol in
each state that Dillard’s has had returns in during our sample
period. If it does not default to a filled map, State may not have
imported as a geographic datatype. Right-click on the Abc icon next
to the State attribute to select Geographic Role, and then assign it
as a State/Province datatype.
3. Double-click the measure Sale Price to provide more detail to the
map.
4. This map will be easier to read as a filled map instead of a symbol
map, so select Filled Map from the Show Me tab.
5. Rename this sheet Returns by State.

To explore the data in greater detail, we can create a Dashboard
with both of our sheets (Returns by Century and Returns by State)
and have them interact with one another.
6. Click the New Dashboard icon.
page 417
7. Drag and drop the two sheets, Returns by Century and Returns by
State, into the dashboard and arrange them so that they are one on
top of the other.
8. Click anywhere in Returns by Century to bring up interactive buttons
to the right. Click the Filter button (it looks like a funnel) to use this
sheet as a filter.
9. Do the same for Returns by State.

Using the sheets as filters in the dashboard allows you to click
one century to see how the levels of percentage of sales returned
changes in each state, and also allows you to click one state to see
which centuries have higher percentage of sales returned in each
state.
Explore the data in the dashboard and answer the next two
questions:
Q1-8. Which Century has the highest sum of returns in Utah?
Q1-9. Which state has a higher sum of returns in the
Cosmetics century than any other?
Analyze Returns by Month

It would also be interesting to see which months have the highest
amount of returns.
Q1-10. Make a guess about which months you think likely see
the highest amount of returns.
0. Create a new worksheet and re-name it Returns by Month.
1. Double-click the Sale Price measure.
2. Double-click the Tran Date dimension. Tableau defaults to showing
the change in returns over year in a line chart. Click the expand
button on the YEAR(Tran Date) pill in the Columns shelf to see
quarters, then click the expand button on the new QUARTER pill to
see Months.
3. To make the distinction in months clearer, change the line charts to

a bar chart on the Show Me tab.
page 418
4. Remove the Quarter pill from the Rows shelf—this will make it
easier to focus on each month over the three years of data
provided.
Q1-11. Which month saw the highest amount of returns in 2016?
2015? 2014? Why do you think these months see such high
amounts of returns?
After looking through total sums of returns across products,
states, and months, it’s time to turn our focus to a more precise
measure of returns: the percentage of returned sales.
Q1-12. Why can we draw more precise insights from analyzing
percentage of returned sales than just analyzing sums of
returns?
Q1-13. Do you think the same two months that saw high sums
of returns in 2014, 2015, and 2016 will also be the two
months with the highest percentage of returned sales?
Why or why not?
In order to explore percentage of returned sales, we first need to
transform the data using Excel’s Power Query tools, then we’ll pull
the transformed data into Excel to see how the percentage of
returned sales differs from the sums of returns across states and
across months.
page 419
Transform the Data to Prepare It for Analysis of

Percentage of Returned Sales
Here is a summary of the Master the Data work that you need to do
in Excel (remember that you need to connect to the WCOB_Dillards
database in Excel on the University of Arkansas Remote Desktop):
5. Navigate to the Data tab on the Excel ribbon and click New Query
> From Database > from SQL Server Database, then retrieve data
using the following query (remember to click Advanced Options to
input the query):
select year(tran_date) as year, month(tran_date) as month, state,
transact.store,
tran_type, sum(sale_price) as amount
from transact
inner join store
on transact.store = store.store
where year(tran_date) = 2014 or year(tran_date) = 2015
group by year(tran_date), month(tran_date), state, transact.store,
tran_type
order by year(tran_date), month(tran_date), state
This query returns only the specific attributes we will use to
analyze the data. For the next several steps of analysis, we are
going to focus solely on differences across months, states, and
years. We are also limiting the dataset to two years so that we can
work with this data in Excel (if we bring in all of the years’ worth of
data, there are too many records for Excel to hold).
6. If the Power Query Editor does not show up immediately, click Edit
in the data preview window to open the Power Query Editor.
The first transformation we need to do is to create two separate
columns: one column for returns and one column for purchases. This
will allow us to look at the separate transaction types more easily
(without having to add a filter), and it will also allow us to create a
Percentage of Sales Returned column.
7. In the Power Query Editor, select the Tran_Type Column, then
select the Transform tab on the ribbon. Select Pivot Column and
use amount for the Value Column.
After this transformation is done, each Purchase transaction
resulted in a “null” value in the new Returns column, and each
Returns transaction resulted in a “null” value in the Purchases
column. We need to replace each null value with 0.
8. Replace Values (Transform Tab) in both the new P and R columns;
replace Null with 0.
9. From the Add Column tab, click Custom Column to create a % of
Sales Returned column using the formula: (if [p]=0 then 0 else
[R]/[P]).
0. From the Home tab on the ribbon, click Close and Load to load the
data into Excel. You should have 7,139 rows (excluding the header)
loaded.
1. Save your Excel file as Chapter11Dataset.xlsx. Ensure that you are
saving the file on the University of Arkansas Remote Desktop so
that you can access it again with Tableau from the remote desktop.
page 420
Load and Work with Data in Tableau

2. After mastering the data in Excel, open Tableau.
3. Since your data is stored in Excel, click Microsoft Excel, then
browse to the file location to Open it.
The data will appear in the Data Source window. Notice that each
of the attributes has a datatype that Tableau automatically selected.
4. Year and Month are not easy to filter when they’re classified as a
Number, so click the # sign on each variable and change them to
String. This will have Tableau treat those variables as text, making
them easier to filter.
Analyze Percentage of Sales Returned by Month

Now that data is loaded, we can create a map to see how average
percentage of returned sales compares across states.
5. Click Sheet 1 to begin working with your data.
6. Rename this sheet Sales Returned by Month.
7. Double-click on the Measure % of Sales Returned.

Tableau defaulted the aggregate measure to SUM. It would be
more meaningful to view the % of Sales Returned aggregated as an
Average.
8. Right-click the SUM(% of Sales Returned) pill in the Rows shelf.
9. Select Measure (Sum) from the window that pops up.
0. Select Average.
page 421
1. Double-click Year and Month dimensions to add them to the

Columns shelf.
Q1-14. Which month has the highest percentage of returned sales
in 2014? 2015?
Q1-15. Which month has the lowest percentage of returned sales
in both years?
Q1-16. Why do you think this difference occurs from analyzing
sums of returns?
Analyze Percentage of Sales Returned by State

2. Create a new worksheet.
3. Double-click the dimension State to create a map. Tableau will
default to a symbol map with a symbol in each state that Dillard’s
has had sales or returns during our sample period.
4. Double-click the measure % of Sales Returned to provide more
detail to the map.
5. This map will be easier to read as a filled map instead of a symbol
map, so select Filled Map from the Show Me tab.
6. The aggregate for % of Sales Returned defaulted to page 422

Sum again, so change it to an Average.
7. Re-name this sheet Returns by State.
To explore the data in greater detail, we can create a Dashboard
with both of our sheets (Sales Returned by Month and Returns by
State) and have them interact with one another.
8. Click the New Dashboard icon.

9. Create a dashboard that contains both sheets. Set both sheets as
filters for the dashboard so that you can explore the data.
Q1-17. Which state had the highest amount of returns in 2014?
2015? (Remember to activate your visuals as filters for the
dashboard. To filter for 2014 or 2015, select the labels of
2014 or 2015).
Q1-18. What was the percentage of sales returned in California in
January 2014?
QUESTION SET 2: DIAGNOSTIC

ANALYTICS—HYPOTHESIS TESTING
LO
11-2
Illustrate
hypothes
is testing
using
diagnosti
c
analytics
to
compare
and
contrast
sales
returns
around
the
holiday
season,
as well
as
compari
ng sales
returns
in
Arkansa
s (where
Dillard’s
is based)
and the
rest of
the
country.
Question 2.1: Is the Percentage of Sales

Returned Significantly Higher in January
After the Holiday Season?
In the previous activity for this dataset, you found that January has
the highest percentage of sales returned of all of the months. This is
likely due to the holiday season. While it’s helpful to see which
month has the highest amount of returns, it would be more useful to
find out if the difference is statistically significant. In order to do so,
we can run a hypothesis test to determine if a significantly higher
percentage of sales are returned during the month of January than
the rest of the months.
To do so, we need to return to the Power Query Editor. Ensure
you are in the University of Arkansas Remote Desktop. If you have
closed the file you worked with in Question Set 1, re-open it.
If you have closed the file, you will need to access the Power
Query window. To do so, return to the spreadsheet that contains the
data that was extracted via the query originally and select a cell of
the table to activate the Query tab in the Excel ribbon. From the
Query tab, select Edit to open the Power Query editor. If prompted
to do so, click Edit Permissions and Run the Native Database
Query in the window that pops up. Then repeat a similar process by
clicking Edit Credentials. The query data should show up in the
Power Query Editor now.
In the Chapter11Dataset.xlsx Power Query Editor, do the
following:
1. From the Add Column tab in the ribbon, select Conditional
Column.
2. In the Conditional Column window that pops up, fill in page 423
the condition (you can leave the column name as
‘Custom,’ we will pivot this column shortly, so the name is
unimportant). Indicate that anytime the month variable equals 1,
then the column should display “Holiday,” otherwise, it should
display “Non-Holiday”. Once you’ve input all of the information, click
OK.
At this point we need to be able to pivot the new Holiday/Non-

Holiday dummy variable column based on % of Sales Returned.
This will allow you to have one column devoted to the % of Sales
Returned during the Holiday (January) month, and another column
for the % of Sales Returned during the non-holiday months. To do
so, we need to first duplicate the % of sales returned column, then
we will pivot the Holiday/Non-Holiday dummy variable column
based on the % of Sales Returned duplicate column.
3. From the Power Query Editor ribbon, click Add Column if you are
not already there.
4. We will add a duplicate column of the % of Sales Returned
column. Select that column, then click Duplicate column.
5. Once the duplicate column is created, we can pivot the page 424
Holiday/Non-Holiday column. First you will need to
click the Transform tab on the ribbon of the Power Query Editor
window.
6. Select the Holiday/Non-Holiday custom column, then click Pivot
Column. Use the column you just copied, % of Sales Returned –
Copy as the values column. Click OK.
We will use these two new Holiday and non-Holiday columns to

run a t-test to see if the percentage of sales returned during the
December and January holiday season is significantly higher than
the rest of the year.
7. Load the transformed data into Excel by navigating to the Home tab
on the ribbon and selecting Close & Load.
It may take a couple minutes for the data to load.

8. Once the data has loaded, navigate to the Data Analysis Toolpak in
the Data tab in the ribbon. If the Data Analysis Toolpak hasn’t been
added in, see Appendix B for directions on how to add page 425
it. Click Data Analysis to open the Analysis Tools
window.
9. Scroll to find the t-Test: Two Sample Assuming Unequal

Variances tool and click OK.
0. In the t-Test window, you will need to input your variable ranges. For
Variable 1 Range, select all of the values that correspond with the
Holiday column.
1. Follow the same pattern for Variable 2 by selecting all of the data
that correspond with the Non-Holiday column.
page 426

2. If you selected labels (Holiday and Non-Holiday) in addition to just
the data, place a check mark in the box next to Labels.
3. Click OK to run the hypothesis test. The output for the hypothesis
test will appear on a new sheet in your Excel workbook.
Q2-1. Based on the p-values (or the t-statistic and critical values),
are the returns as a percentage of sales in January greater,
less than, or the same as the returns as a percentage of
sales for the rest of the year?
Q2-2. What can we conclude about returns?
Question 2.2: Is the Percentage of Sales

Returned Significantly Different in Arkansas
Than the Rest of the Country?
Following similar steps that you completed in Question 2.1 above,
determine if the percentage of sales returned is significantly different
in Arkansas, where Dillard’s is based, than the rest of the country.
page 427
First, you will need to create dummy variable columns for “AR”
and “Non-AR”—follow the same steps that you did in steps 1-7 to
create the Holiday and non-Holiday variables (Open the Power
Query window by selecting Edit from the Query tab in the ribbon.
From the Add Column tab, create a conditional column to separate
AR from Non-AR states. Next, duplicate the % of Sales Returned
column again. From the Transform tab, Pivot your new custom
column for AR/Non-AR - use the copy of % of Sales Returned as the
values).
Your next step will be to run a new t-test, similar to the steps you
completed in steps 8-12.
Q2-3. Using the p-values (or the t-statistic and critical values),
are the returns as a percentage of sales in Arkansas and
January greater, less than, or the same as the returns as
a percentage of sales for the rest of the country?
Q2-4. What can we conclude about returns?
QUESTION SET 3: PREDICTIVE

ANALYTICS
LO
11-3
Predict
returned
sales in
future
periods
using
predictiv
e
analytics
.
Question 3.1: By Looking at Line Charts for
2014 and 2015, Does the Average Percentage
of Sales Returned in 2014 Seem to Be
Predictive of Returns in 2015?
After assessing how different variables impact returns, we have a
better idea of how to help Dillard’s prepare for returns—both across
states and for the holiday season. We can also get an idea of how
much a previous year’s percentage of returned sales can help
predict the next year’s. To answer this question, we will create a
PivotTable, a PivotChart, and use slicers to see if we can improve
our predictions by looking state-by-state.
Create a PivotTable and a PivotChart to Compare 2014

and 2015 Returns
1. Open the data you transformed and built into spreadsheet
Chapter11Dataset.xlsx.
2. Create a PivotTable from the dataset.
To view a comparison of how the average percentage of sales
returned changed each year, we want to view the years as columns,
and see a row for each month’s average percent of sales returned
for either year.
3. Drag Month to the Rows.
4. Drag Year to columns.
5. Drag % of Sales Returned to Values.
6. Change the aggregate for % of Sales from Sum to Average using
the Value Field Settings.
7. To view this data as a line chart, we need to insert a PivotChart. To
do so, click the Analyze tab on the ribbon, underneath PivotTable
tools (if you do not see this option, select one of the cells in your
PivotTable to open the interactive ribbon).
8. Click PivotChart to see the different chart options.
9. Change the default to Line and then click OK. page 428
Q3-1. Based on the line chart, does 2014 seem to be useful in

predicting 2015’s percentage of sales returned?
0. Now we can add a Slicer to interactively filter the PivotTable and
PivotChart to see how the values changed by state each year. Click
into the data of your PivotTable to make the options for adding
slicers available.
1. In the Insert tab on the ribbon, select Slicer.
2. Select State and then click OK. page 429

3. Filter first for AR (Arkansas). Try again to filter for UT (Utah).
Q3-2. Based on these line charts, does it seem more beneficial to
predict percentage of sales returned on a state-by-state
basis or in aggregate form?
Question 3.2: Using Regression, What Can

We Predict for Returns as a Percentage of
Sales Based on Historical Transactions?
4. Because the line graphs seemed to suggest that the percentage of
sales returned in 2014 will help predict the percentage of sales
returned in 2015, we can run a regression to build a model that will
help stores predict the percentage of sales that will be returned
each month. First, we will need to transform the data one more
time. This transformation requires removing several columns, so
you may choose to create a duplicate file of your page 430
current chapter 11 spreadsheet if you need to save
your progress from Question Set 2.
5. Return to the worksheet that contains the raw data. From this
worksheet, you can access the Query tab in the ribbon and click
Edit in order to open the Power Query tool.
6. We ultimately need to pivot the year column so that we have two
separate columns for comparing 2014 and 2015, but before we can
do that, we need to remove several columns. Select columns P, R,
Holiday, Non-Holiday, Non-AR, and AR, and then select Remove
Columns from the ribbon. You can remove each column individually,
or you can select multiple columns at a time by pressing “Ctrl” on
your keyboard as you select the columns.
7. Now you’re ready to pivot the year column. Select the year column,
and then click Pivot Column from the Transform tab.
8. Select % of Sales Returned as your Values, then click page 431

OK.
9. The last step to prepare this data for analysis is to remove zeros
and null values from your new 2014 and 2015 columns. Select the
drop-down next to the column for 2014 and remove the checkmarks
next to (null) and 0.
0. Do the same for the 2015 column.

1. Now your data is ready for Analysis. Click Close & Load from the
Home tab to return to Excel.
2. To run a Regression Analysis, you need to have the Data Analysis
Toolpak added in to Excel (see Statistics Appendix B for information
on how to do so).
3. Select Data Analysis from the Data tab in the ribbon.
4. Scroll to find the Regression tool in the Analysis Tools page 432
window and click OK.
5. Because you will be predicting 2015 values based on 2014’s

values, 2014’s values are our (X) variables and 2015’s values are
our (Y) variables. Enter the following information in the Regression
window (these input ranges should come from the original data
worksheet, not the PivotTable):
Input Y Range: This is the column containing the 2015 data.
Input X Range: This is the column containing the 2014 data.
Labels: Place a checkmark in this box if you included the labels for
2014 and 2015 in your selections for the Y and X ranges.
Click OK.
Q3-3. Looking at your regression output, was the relationship
between 2014 and 2015 percentage of sales returned
significant? How can you tell?
Q3-4. Brainstorm at least four other data items (e.g., economy,
type of customer, etc.) that would be helpful in predicting
future the next year’s percentage of sales returned.
Q3-5. Upon project completion of Question Sets 1, 2, and 3, what
have you learned from completing this analysis?
1Accounting Standards Codification (ASC) 606, Revenue from Contracts with Customers,
as amended, and created by Accounting Standards Update (ASU) 2014-09, Revenue
from Contracts with Customers.
page 433
page 434
Appendix A
Basic Statistics Tutorial
POPULATION VS. SAMPLE

Restaurant stores and retail stores are often faced with the decision
of whether to stay open on Sunday. Like Chick-Fil-A, retail owners
and restaurant owners sometimes like to close on Sunday to allow
their employees to spend time with families or simply to take a break
for the day.
What percentage of the restaurants and retail owners close on
Sunday? We’d love to ask a survey question on SurveyMonkey or
Qualtrics and get every retail/restaurant owner to respond. If we
could get every response from every restaurant/retail owner, we’d
call this the results of the population, defined as the total set of
observations. Because it is virtually impossible to get every owner to
respond to our survey, we often just get a sample (defined a subset
of the data collected from the population) to reply and hope that the
results we find from the sample are representative of what we would
find had we been able to get the total population to respond.
As a side note, in the past, auditors were only able to view a
sample of the accounting transactions that were made. But with new
high-powered analytics tools as well as having accountants trained
to use such tools, auditors are increasingly able to consider data
from the full population instead of a small sample!
PARAMETERS VS. STATISTICS: WHAT

IS THE DIFFERENCE?
Whereas a parameter comes from a population, a statistic comes
from a sample. For example, the population average (or mean)
would be the parameter we would call the greek letter mu (μ). For
example, the population average of stores close on Sunday might be
24 percent. However, since we’re only able to survey a sample, the
result of surveying the sample would be sample statistic average x-
bar, or . If we don’t know the true population average, μ, we will use
the sample average, or, to make inferences about the true population
average.
DESCRIBING THE SAMPLE BY ITS

CENTRAL TENDENCY, THE MIDDLE,
OR MOST TYPICAL VALUE
To learn more about a sample, we often use measures of the central,
the middle, or most typical value to describe the sample. The mean,
median, and mode are three common measures used to assess
central tendency.
page 435
The sample arithmetic mean is the sum of all the data points
divided by the number of observations. The median is the midpoint
of the data and is especially useful when there are skewed numbers
one way or another. The mode is the observation that occurs most
frequently.
DESCRIBING THE SPREAD (OR

VARIABILITY) OF THE DATA
The next step after describing the central tendency of the data is to
assess its spread, or variability. This might include considering the
maximum and minimum values and the difference between those
two values, which we define as the range.
The most common measures of spread or variability is standard
deviation or variance, where each ith observation in the sample is xi,
and the total number of observations is N. The standard deviation,
the greek letter sigma, σ, is computed as follows:
And relatedly, the variance, σ2 is computed as follows:
The greater the sample standard deviation or variance, the

greater the variability.
PROBABILITY DISTRIBUTIONS
There are three primary probability distributions used in statistics and
data analytics, including normal distribution, the uniform distribution,
and the poisson distribution.
Normal Distribution
A normal distribution is arguably the most important probability
distribution because it fits so many naturally occurring phenomenon
in and out of accounting—from the distribution of return on assets to
the IQ of the human population.
The normal distribution is a bell-shaped probability distribution
that is symmetric about its mean, with the data points closer to the
mean more frequent than those data points further from its mean. As
shown in Exhibit A-1, data within one standard deviation (+/− one
standard deviation) includes 68 percent of the data points. Within
two standard deviations, 95 percent of the data points; three
standard deviations, 99.7 percent of the data points.
EXHIBIT A-1
Normal Distribution and the Frequency of Observations around Its Mean (Using 1, 2,
or 3 Standard Deviations)
A z-score is computed to tell us how many standard deviations

(σ), a data point (or observation), xi, is from its population mean, μ,
using the formula z = (xi − μ)/σ. A z-score of 1 page 436
suggests that the observation is one standard
deviation above its mean. A z-score of –2 suggests that the
observation is two standard deviations below its mean.
Many of the statistical tests employed in data analysis are based
on the normal distribution and how many standard deviations a
sample observation is from its mean.
Uniform Distribution and Poisson

Distribution
The uniform distribution is a probability distribution where all
outcomes are equally likely. Like in a fair coin toss, the distribution of
heads and tails are equally likely. A deck of cards has an equal
distribution of hearts, clubs, diamonds, or spades. Likewise, a deck
of cards has an equal distribution of “queen”s and “3”s. A poisson
distribution is a distribution with a low mean and highly skewed to the
right.
HYPOTHESIS TESTING
As we learn in Data Analytics, data by itself is not really that
interesting. It is using data to answer, or at least address, questions
posed by management that makes it interesting.
Management might pose a question in terms of a hypothesis, like
their belief that sales at their stores are higher on Saturdays than on
Sundays. Perhaps they want to know this answer to decide if they
will need more staff to support sales (e.g., cashiers, shelf stockers,
parking lot attendants, etc.) on Saturday as compared to Sunday. In
other words, management holds an assumption that sales are higher
on Saturday than on Sundays.
Usually hypotheses are paired in two’s: the null hypothesis and
the alternate hypothesis.
The first is the base case, often called the null hypothesis, and
assumes the hypothesized relationship does not exist. In this case,
the null hypothesis would be stated as follows:
Null hypothesis: H0: Sales on Saturday are less than or equal to
sales on Sunday.
The alternate hypothesis would be the case that management
believes to be true.
Alternate hypothesis: HA: Sales on Saturday are greater than
sales on Sunday.
For the null hypothesis to hold, we would assume that Saturday
sales are the same as (or less than) Sunday sales. Evidence for the
alternate hypothesis occurs when null hypothesis does not hold and
is rejected at some level of statistical significance. In other words,
before we can reject or fail to reject the null hypothesis, we need to
do a statistical test of the data with sales on page 437
Saturday and Sundays and then interpret the results
of that statistical test.
STATISTICAL TESTING
There are two types of results from a statistical test of hypotheses
that may occur or may be interpreted in different ways: the p-value
and/or the critical values.
The p-Value
We describe a finding as statistically significant by interpreting the p-
value.
A statistical test of hypothesis may return a p-value. The p-value
is the result of a test that either rejects or fails to reject the null
hypothesis. The p-value is compared to a threshold value, called the
significance level (or alpha). A common value used for alpha is 5
percent or 0.05 (as is 1 percent or 0.01).
The p-value is compared to the alpha threshold. A result is
statistically significant when the p-value is less than alpha. This
signifies a change was detected: that the default hypothesis can be
rejected.
If p-value > alpha: Fail to reject the null hypothesis (i.e., not
significant result).
If p-value <= alpha: Reject the null hypothesis (i.e., significant
result).
For example, if we were performing a test of whether Saturday
sales were greater than Sunday sales and the test statistic was a p-
value of .09, we would state something like, “The test found that the
Saturday sales are not different than Sunday sales, failing to reject
the null hypothesis at a 5% level of significance.”
This statistical result should then be reported to management,
reporting the results of the statistical test.
The Confidence Interval

The significance level can be computed by subtracting alpha from 1
to give a confidence level of the hypothesis given the statistical test
of the data.
For example, if the confidence level is 95 percent, then alpha (α)
is 5 percent. In Exhibit A-2, the 95 percent of the figure represents
the confidence interval—we are 95 percent confident that the true
population parameter of Saturday and Sunday sales falls
somewhere in that area.
EXHIBIT A-2
Statistical Testing Using Alpha, p-Values, and Confidence Intervals
Therefore, statements such as the following can also be made:

With a p-value of 0.09, the test found that Saturday and Sunday
sales are not different than Sunday sales, failing to reject the null
hypothesis at a 95% confidence level.
This statistical result should then be reported to management,
reporting the results of the statistical test.
page 438
INTERPRETING THE STATISTICAL

OUTPUT FROM A SAMPLE T-TEST OF
A DIFFERENCE OF MEANS OF TWO
GROUPS
A sample t-test is a statistical test used to compare the means of two
sets of data observations.
For example, it might be comparing means of two independent
groups. For example, a t-test might be used to compare the mean
return on asset (ROA) for companies in the retail industry to the
mean ROA for companies in the entertainment industry to see if one
is statistically higher than the other.
Or it could be a paired t-test of the same group of companies but
at different times. Such a t-test might compare the mean return on
assets (ROA) for companies in the retail industry in 2020 to the
mean ROA same companies in the retail industry in 2021.
Let’s suppose that a company is trying to understand if its rate of
sales returns is higher around the end-of-year holidays than at other
times (non-holidays) during the year. To assess whether the sales
returns are different, a t-test is performed in Excel to see if there is a
difference of daily mean sales returns (as a percentage of total sales
that day) between the holidays and non-holiday periods. After
performing the t-test, Excel returns the following statistical output:
The t-test output found that the mean holiday sales returns over
1,167 days is 0.13 (or 13 percent) of sales, and the mean non-
holiday sales returns are 0.119 (or 11.9 percent) of sales. The
question is if those two numbers are statistically different from each
other. The t Stat of 7.86 and the p-value (shown as “P(T<=t) one
tail") is 3.59E-15 (i.e., well below .01 percent), suggesting the two
sample means are significantly different from each other.
The t-test output notes the difference in crucial p-values for a
one-tailed t-test and a two-tailed t-test. A one-tailed t-test is used if
we hypothesize that holiday returns are significantly greater (or
significantly smaller) than non-holiday returns. A two-tailed t-test is
used if we don’t hypothesize holiday or non-holiday returns are
greater or smaller than the other, only that we expect the two sample
means will be different from each other.
page 439
INTERPRETING THE STATISTICAL

OUTPUT FROM A REGRESSION
Regressions are used to help measure the relationship between one
output variables and various inputs. We can think about this like an
algebraic equation where y is the dependent variable and x is the
independent variables, where y = f(x). As an example, we
hypothesize a model where y (or College Completion Rate) =
f(factors potentially predicting college completion rate) including the
independent variable SAT score (SAT_AVG). In other words, we
hypothesize that college completion rates depend on SAT scores.
Through regression analysis, we can assess if the college
completion rate is statistically related to the SAT score.
As you recall from Lab 2-5 and Lab 3-2, we are considering the
relationship between SAT scores and the college completion rate for
first-time, full-time students at four-year institutions.
Here is the regression output from Lab 3-2:
There are many things to note about the regression results. The
first is that the overall regression model did better than chance at
predicting the college completion rate as shown by the “F”-score. We
note that by seeing the p-score representing “Significance F” result is
very small, almost zero, suggesting there virtually zero probability
that the completion rate can be explained by no independent
variables than a model that has independent variables. This is
exactly the situation we want suggesting we should be able to
identify a factor that explains completion rates.
There is another statistic used to measure how the overall
regression model did at predicting the dependent variable of
completion rates. The adjusted R-squared is a value between 0 and
1. An adjusted R-squared value of 0 represents no ability of the
model to explain the dependent variable and an adjusted R-squared
value of 1 represents perfect ability of the model to explain the
dependent variable. In this case, the adjusted R-squared value is
0.642, which represents a reasonably high ability to explain the
changes in the college completion rate.
The statistics also report that the SAT score (SAT_AVG) helps
predict the completion rate. This is shown by the “t Stat” that is
greater than 2 (or less than –2) for SAT_AVG (with t Stat of 47.74)
and a p-value less than an alpha of 0.05 (as shown with the p-value
of 1.564E-285). As expected, given the positive coefficient on
SAT_AVG, the greater the SAT score, the greater the college
completion rate.
page 440
Appendix B
Accessing the Excel Data Analysis
Toolpak
Excel offers a toolpak that helps perform much of the data analysis,
called the Excel Data Analysis Toolpak.
To run a correlation, form a histogram, run a regression, or
perform other similar analysis using the Excel Data Analysis Toolpak,
we need to make sure our Analysis Toolpak is loaded up, by looking
at the ribbon of Data > Analysis and seeing if the Data Analysis
Add-In has been installed.
If it has not yet been added, go to File> Options > Add-Ins,

select the Analysis Toolpak, and select OK:
page 441
In the Add-ins window that appears, place a check mark next to

Analysis ToolPak and then click OK. This will add the Data Analysis
ToolPak to the Data tab so you can perform additional data analysis.
Step 2:
To perform the additional data analysis, please select Data >
Analysis > Data Analysis. A dialog box will open.
In this text, we will highlight the use of the following analysis tools:
Correlation: To understand the if and the extent to which variables
are related to each other.
Descriptive statistics: To understand the basic statistics, including the
mean, standard deviation, minimums, and maximums of a data set.
Histogram: To understand the frequency of the data using a display
of rectangles with area proportional to the underlying frequency of
the data.
Regression: To understand the relation between specific dependent
variable values and independent variable inputs.
T-tests: To understand the probability of a difference in means
between either two independent samples or a paired sample through
time.
page 442
Appendix C
Excel (Formatting, Sorting,
Filtering, and PivotTables)
BASIC FORMATTING AN INCOME

STATEMENT USING EXCEL
FUNCTION SUM()
Suppose we want to put the following data into the appropriate
income-statement format:
Revenues 50000
Expenses
Cost of Goods Sold 20000
Research and Development Expenses 10000
Selling, General, and Administrative Expenses 10000
Interest Expense 3000
Required:
1. Add a comma as a 1000 separator for each number.
2. Insert the words “Total Expenses” below the list of expenses.
3. Calculate subtotal for Total Expenses using the SUM() command.
4. Insert a single bottom border under Interest Expense and under the
Total Expenses subtotal.
5. Insert the words “Net Income,” and calculate Net Income (Revenues
– Total Expenses).
6. Format the top and bottom numbers of the column with a $ currency
sign.
7. Insert a Bottom Double Border to underline the final Net Income
total.
page 443
Solution:
1. Open Appendix C Data.xlsx and access the sheet named “Income
Statement Formatting.”
2. Add a comma as a 1000 separator for each number.
Highlight the column with all of the numbers. Right click on
Format Cells. . . to open this dialog box:
Click on Number and set Decimal places to zero. Click on Use

1000 Separator (,) and click OK.
3. Insert the words “Total Expenses” below the list of expenses.
Type “Total Expenses” at the bottom of the list of expenses.
4. Calculate subtotal for Total Expenses using the SUM() command.
Use the SUM() command to sum all of the expenses, as follows.
page 444
Here is the result:
Revenues 50,000
Expenses
Cost of Goods Sold 20,000
Research and Development Expenses 10,000
Selling, General, and Administrative Expenses 10,000
Interest Expense 3,000
Total Expenses 43,000
5. Insert a single bottom border under Interest Expense and under the
Total Expenses subtotal.
Use the icon indicated to add the bottom border.

6. Insert the words “Net Income” and calculate Net Income (Revenues
– Total Expenses).
Type “Net Income” at the bottom of the spreadsheet. Calculate
Net Income by inserting the correct formula in the cell (here, =B2-
B9):
Microsoft Excel
page 445
7. Format the top and bottom numbers of the column with
a $ currency sign.
Right click on each number and Format Cells, select currency
and no decimal points and click OK.
Microsoft Excel
8. Insert a Bottom Double Border to underline the final Net Income

total.
Place your cursor on the cell containing Net Income (7,000).
Then select Bottom Double Border from the Font > Borders menu.
This is the final product:
Revenues $50,000
Expenses
Cost of Goods Sold 20,000
Research and Development Expenses 10,000
Selling, General, and Administrative Expenses 10,000
Interest Expense 3,000
Total Expenses 43,000
Net Income $7,000

Microsoft Excel
(Level 1) Basic Data Manipulation (Filters, Sorts, PivotTables)

9. Open Appendix C Data.xlsx and access the sheet named “Basic
Data Manipulation.”
0. Look at the data.
page 446
(Level 2) Sorting the Data

1. Let’s sort the data. To do so, go to Data > Sort & Filter > Sort.
Microsoft Excel
2. Let’s sort by sales price from largest to smallest. Input Sales into
the Sort by, select Largest to Smallest in the dialog box, and select
OK.
Microsoft Excel
page 447
The highest sales price appears to be Apricots at a cost of $140.
Microsoft Excel
Looking down at the bottom of this list, we see that the lowest
sales price appears to be bananas for $2.52.
(Level 3) Filtering the Data
Next, let’s filter the data to only look at only the banana transactions.
3. Let’s sort the data. To do so, go to Data > Sort & Filter > Filter.
4. An upside down triangle (or a chevron) will appear. Click the
chevron in cell F1, click Select All to unselect all, and then select
only the word “Banana”.
5. The resulting data should appear as follows:
Microsoft Excel
6. Alternatively, we could filter based on date to get all transactions on

3/2/2021. We first need to clear the filter in cell F1, by clicking on
the Filter symbol and selecting Select All.
page 448
7. Click the chevron in cell C1, click Select All to unselect
all, and then select only the word “2021” then “March” then “2”.
Microsoft Excel
(Level 4) Pivot Tables

8. Let’s compute the accumulated gross margin for bananas, apricots,
and apples by taking the sales.
9. First, unclick the filter at Data > Sort & Filter > Filter by clicking on
and unselecting Filter.
0. Next, let’s compute the gross margin for each line item in the
invoice. In cell J1, input the words “Gross Margin”. Underline it with
a bottom border. In cell J2, input “=H2-I2” and hit <Enter> in this
way.
Microsoft Excel
1. Copy the result from cell J2 to J3:J194.

2. Now it is time to use the pivot table. Recall that a pivot table
summarizes selected columns in a spreadsheet, but doesn’t change
the spreadsheet itself. Recall we are trying to summarize the
accumulated gross margin for bananas, apricots and apples.
3. Select Insert > Tables > PivotTable.
Analytics Tool: Excel PivotTables
PivotTables allow you to quickly summarize large amounts of
data. In Excel, click Insert > PivotTable, choose your data
source, then click the checkmark next to or drag your fields to
the appropriate boxes in the PivotTable Fields pane to identify
filters, columns, rows, or values. You can easily move attributes
from one pane to another to quickly “pivot” your data. Here is a
brief description of each section:
page 449
Rows: Show the main item of interest. You usually want

master data here, such as customers, products, or accounts.
Columns: Slice the data into categories or buckets. Most

commonly, columns are used for time (e.g., years, quarters,
months, dates).
Values: This area represents the meat of your data. Any

measure that you would like to count, sum, average, or
otherwise aggregate should be placed here. The aggregated
values will combine all records that match a given row and
column.
Filters: Placing a field in the Filters area will allow you to filter
the data based on that field, but it will not show that field in the
data. For example, if you wanted to filter based on a date, but
didn’t care to view a particular date, you could use this area of
the field list. With more recent versions of Excel, there are
improved methods for filtering, but this legacy feature is still
functional.
4. Make sure all data is selected as follows in Table/Range and select
OK.
Microsoft Excel
5. The empty pivot table will open up in a new worksheet, ready for the
pivot table analysis.
Columns:
Rows: [Description]
ΣValues: [Gross Margin]
Inputs into PivotTable

Drag [Description from FIELD NAME into the Rows and [Gross
Margin] from FIELD NAME into ΣValues fields in the pivot table. The
ΣValues will default to “Sum of Gross Margin”.
page 450
The resulting pivot table will look like this:
6. The analysis suggests that the gross margin for apples is $140.39;
for apricots, $78.02 and for bananas, $77.08.
THE VLOOKUP FUNCTION

One of Excel’s most useful tools for looking up data from two
separate tables and providing matching information based on related
fields (often based on a primary key/foreign key relationship) is the
VLookup function.
To demonstrate the VLookup function, we will work with
sales transactions and tax rate data. Sometimes when you access
or request sales transaction data, the tax rate may not be included.
In order to calculate the amount of tax owed for each transaction, we
need to match the state tax rate with the state the customer is from.
This would be an arduous task to do manually, particularly in
datasets that are large. We can use Excel’s VLookup function to
match the state the customers are from with the state tax rate.
Open Excel File Appendix C Data.xlsx and access the sheet
named “VLookup.”
The dataset contains information about sales transactions to
different customers. This is similar to the dataset that you use in Lab
2-2 of chapter 2. There are 132 unique transactions, but 150 rows—
this is because some of the transactions had multiple products on
them.
1. The location of each store where the sales were made.
2. A second table with sales tax information for each state.
You may need to scroll to the right to see the sales tax table.
page 451
Data Dictionary:
Sales_Transactions table:
Sales_Order_ID: Unique identifier for each individual Sales

Order
Sales_Order_Date: Date each sales order was placed
Sales_Order_Quantity_Sold: Quantity of each product sold

on the transaction
product_description: Description of the product sold
Product_Sale_Price: Price of each product sold on the

transaction
Store_Location: State in which the store is located
State Sales Tax table:
State: the state abbreviation
State Tax Rate: the tax rate for each state
There are two columns which match in these two tables:

Store_Location (from the Sales_Transaction table) and State (from
the State Sales Tax table). These two tables are placed next to one
another to make the VLOOKUP function easier to manage.
Step 1: We will add a new column to the Sales_Transactions
table to bring in the State_Sales_Tax associated with every
Customer_St listed in the transactions table.
1. Add a new column to the right of the Store_Location named
Sales_Tax (cell G1).
In cell G2, we will create a VLOOKUP Function. VLOOKUP

Functions have four arguments:
Cell_reference: the cell in the current table that has a match in the
related table. In this case, it is a reference to the row’s
corresponding Store_Location. Excel will match that state with the
corresponding state in the sales tax table.
Table_array: an entire table reference to the table that contains the
descriptive data that you wish to be returned. In this case, it is the
entire State Sales Tax table.
Column_number: the number of the column (not the letter!) that
contains the descriptive data that you wish to be returned. In this
case, State Sales Tax Rate is in the second column of the State
Sales Tax table, so we would type 2.
True or False: there are two types of VLOOKUP functions, true and
false. TRUE is for looking up what Excel calls “approximate” data –
this is useful for looking up tax rates based on income buckets or
discount rates based on buckets corresponding with dates that the
customer pays by. In our case, we’ll use FALSE. A FALSE
VLOOKUP will only return matches for when there is an exact
match between the two tables (whenever your data is ‘relational,’
structured data, a perfect match should be easily discoverable).
Step 2: Because the ranges of data are not only formatted as
tables, but also named properly, the VLOOKUP Function should be
relatively easy to follow:
page 452
2. Type in the following function (using cell references will be easier

than typing manually):
3. Once you click Enter, the formula should copy all the way down—
once again exhibiting the benefits of working with Excel tables
instead of ranges.
page 453
Appendix D
SQL Part 1
SQL can be used to create tables, delete records, or edit databases,

but in Data Analytics, we primarily use SQL to extract data from the
database—that is, not to edit or manipulate the data, but to create
different views of the data to help us answer business questions.
SQL extraction queries are also referred to as SELECT queries,
because they each begin with the word SELECT.
Throughout this appendix, all the examples and the practice
problems refer to Appendix D Data.accdb. This is a very small
database to help you be able to immediately visualize what a query’s
results would look like.
INTRODUCTION TO SQL CLAUSES

Every SQL SELECT query must have two key SQL clauses,
SELECT and FROM.
Introduction to SELECT
SELECT indicates which attributes you wish to view. For example,
the Customers table contains a complete customer list with several
descriptive attributes for each of the company’s customers. If you
would like to see a full customer list, but you just want to see
FirstName, LastName, and State, you can just select those three
attributes in the first line of your query:
SELECT FirstName, LastName, State
Introduction to FROM
FROM lets the database management system know which table(s)
contains the attributes that you are selecting. For instance, in the
query begun above, the three attributes in the SELECT clause come
from the Customers table. So that query can be completed with the
following FROM clause:
FROM Customers
Try putting that query all together to see the results:
SELECT FirstName, LastName, State
FROM Customers
page 454
This returns the result in Figure D-1:
FIGURE D-1
If you wish to view the same three columns, but you want to see
the LastName column as the first column, so that the results more
closely resemble a phone book, you can change the order of the
attributes listed in your SELECT statement:
SELECT LastName, FirstName, State
FROM Customers
Now the query returns the same number of records, but with a
different order of attributes (columns), seen in Figure D-2:
FIGURE D-2
Source: Microsoft Excel, 2016
SELECT FROM Practice
1. Create a query that will return only the Inventory_Description and
Price from the Inventory table.
2. Create a query that will show only the Order_date and CustomerID
from the Sales_Order table.
3. Create a query that will show the City and State from the Customers
table.
page 455
After you get the hang of creating simple SELECT FROM

queries, you can begin to bring in some of the SQL clauses that can
make our queries even more interesting. The next two SQL clauses
we will cover are WHERE and ORDER BY. They follow FROM, to
make a query in this order:
SELECT
FROM
WHERE
ORDER BY
One more bit of SELECT information—how to SELECT all of the
attributes:
If you wish to view every attribute in the same order as they exist
in the table, you can use a shortcut to select all:
SELECT *
FROM Inventory
A simple SELECT FROM query with SELECT * isn’t very
interesting on its own, but when we begin filtering records, SELECT *
can be a quick way to view how many records fit a certain criteria.
We filter records with the WHERE clause.
Introduction to WHERE
WHERE behaves like a filter in Excel. An example of using WHERE
to modify the query in is the following:
FROM Customers
WHERE State = “Arkansas”
That query would return only the customers who were from
Arkansas, the result is shown in Figure D-3:
FIGURE D-3
Source: Microsoft Excel, 2016
The syntax of a simple WHERE clause is the following:

WHERE [attribute_name] = [criteria]
The attribute_name needs to spelled exactly the way it is in the
database without any formatting (for example, do not place the
attribute name in quotes).
Formatting criteria in a WHERE clause, or “So, what’s the deal
with the quotes around Arkansas?”
There are three main datatypes that you will work with in SQL:
text, numbers, and dates. Every attribute in a database is stored as
one of those datatypes. Let’s look at the Inventory table in the
Ch2_SQL_Tutorial database. It has three fields:
InventoryID (example entries: I-1, I-2)
Inventory_Description (example entries: Dalton Dress Boot, Ray-
Ban Wayfarer)
Price (example entries: 495, 1250)
Text Datatypes
Both InventoryID and Inventory_Description are text datatypes. Most
text datatypes are descriptive or categorical elements in the
database. When you filter for criteria from a text attribute, the criteria
must be surrounded in quotes. Examples:
WHERE State = “Arkansas”
WHERE Inventory_Description = “Dalton Dress Boot”
page 456
A word of caution! Programs like Microsoft Word apply

formatting to quotes by turning them into “curly quotes.” Microsoft
Access and other relational database systems cannot read the curly
quotes. If you draft your queries in Word and then copy and paste
them into a SQL editor, the quotes will need to be re-typed in the
query editor for the database application to be able to read the
criteria appropriately.
Number Datatypes
Price, on the other hand, is a number datatype. You could sum or
average the contents of that attribute and arrive at a meaningful
value. Another example of number datatypes are Quantity_Sold in
the Sales_Order table. When you filter for criteria from a number
attribute, there is no need to format the criteria at all. Examples:
WHERE Price = 395
WHERE Quantity = 2
Date Datatypes
For an example of the third datatype, date, look at the Sales_Order
table to find the Order_Date attribute. When you filter for criteria from
a date attribute, the date should be enclosed in # signs and follow
the following format #mm/dd/yyyy# Examples:
WHERE Order_Date = #01/02/2019#
WHERE Order_Date = #12/31/2018#
Date formats in other database management systems: Date
formatting in SQL is variable across relational database
management systems.
In SQLite, date format is ‘yyyy-mm-dd’ For example, ‘2019-01-02’ or
‘2018-12-31’
In SQL Server, date format is ‘yyyymmdd’ For example ‘20190102’
or ‘20181231’
Other methods of filtering, or, do we always have to filter for an
exact match?
Each of the WHERE examples we’ve seen so far have used the
equals sign operator. But there are many other ways to filter other
than for exact matches. For now, we’ll just start with a few other
operators, shown in Exhibit D-1:
EXHIBIT D-1
Operator Description
> used Returns all records that have numbers in that field
with a greater than the criteria specified.
number
datatype
> used Returns all records that follow the criteria
with a text alphabetically (a–z).
datatype
> used Returns all records that have numbers in that field less
with a than the criteria specified.
number
datatype
> used Returns all records that precede the criteria
with a text alphabetically (a–z).
datatype
Operator Description
>= and <= Similar to the above criteria, but will also include
numbers or text that is an exact match to what is listed
in the criteria.
<> functions as the inverse as the exact match (=) filter, it
will return all of the records except those that match
the criteria listed in the WHERE clause.
More SELECT FROM WHERE Examples

To extract all of the records from the Inventory table that have prices
greater than $1,000:
SELECT *
FROM Inventory
WHERE Price > 1000
page 457
That query returns the following records shown in Figure D-4:
FIGURE D-4
To extract all of the records from the Customers table that follow
the last name “jones” alphabetically:
SELECT *
FROM Customers
WHERE LastName > “Jones”
FIGURE D-5
If you wanted to include any employees with the last name of

Jones in the list, you would change the operator from > to >= :
SELECT *
FROM Customers
WHERE LastName >= “Jones”
The revised output is shown in Figure D-6:
FIGURE D-6
You can see that Jeremy Jones is included in this output.
SELECT FROM WHERE Practice

1. Write a query that will return all of the records from the Sales_Order
table that have a quantity greater than 4.
table that have a quantity less than or equal to 3.
table that had exactly one item on it.
4. Write a query that will return all of the records from the Customers
table of the customers from Texas.
5. Write a query that will return all of the records from the Customers
table of the customers from Baton Rouge
page 458
Introduction to ORDER BY
In Figure D-6 above, when you added Jeremy Jones to the output,
you might have been surprised that the order of the records didn’t
change. The default order of SQL queries is ascending based on the
first column selected. When you SELECT *, the default will be in the
order of the Primary Key, which is the order of the records in the
original table.
If you would like to sort the records in a query output based on
any other column, you can do so with the ORDER BY clause.
The syntax of an ORDER BY clause is the following:
To sort the records in ascending order (1 to infinity or A to Z):

ORDER BY [attribute_name] ASC
To sort the records in descending order (infinity to 1 or Z to A):
ORDER BY [attribute name] DESC
Similar to WHERE, the attribute_name needs to spelled exactly

the way it is in the database without any formatting (for example, do
not place the attribute name in quotes).
ORDER BY is always the last line in any query, no matter how
complex the query is.
Example Queries with ORDER BY

To revise the first query in the appendix from Figure D-6, order the
output by state, ascending:
FROM Customers
ORDER BY State ASC
FIGURE D-7
Notice how the two figures have the same information, the same
order of attributes, and the same number of records, but the ordering
of the records has changed.
To revise the same query, but this time to order the results by
both Last Name and First Name (ascending):
FROM Customers
ORDER BY LastName ASC, FirstName ASC
FIGURE D-8
page 459
EXPANDING THE USAGE OF SELECT

So far, you have learned the SQL keywords SELECT, FROM,
WHERE, and ORDER BY. These words are useful for creating views
of a limited amount of data from one table, but other than limiting the
amount of rows and columns returned (with WHERE and SELECT,
respectively) and changing the order in which they are returned, the
table data that returns from these queries is still presented in the
same format as it is stored in the database.
We can extend the usage of SELECT to manipulate the data that
returns from a query with aggregates.
More Actions You Can Take in SELECT:

Aggregates and Aliases
To aggregate means to form a group or a cluster. In SQL, aggregate
data represents grand totals or subtotals of data. For example, if you
did not want to simply view all of the individual orders in a
Sales_Orders table, but you wanted instead to see a total count of
how many orders were in a table, or you wanted to see the grand
total quantity of products ever sold, you would want to aggregate the
data.
The following functions are commonly used in the SELECT
clause to aggregate data:
SUM(attribute)
COUNT(attribute)
AVG(attribute)
The following query uses an aggregate function to create a query
that would show the total count of orders in the Sales_Orders table:
SELECT COUNT(Sales_Order_ID)
FROM Sales_Orders
The output of that query would produce only one column and one
row, shown in Figure D-9:
FIGURE D-9
The problem with this output, of course, is the lack of description.

The column is titled Expr1000, which is not very descriptive. This
title is produced because there isn’t a column named page 460
COUNT(Sales_Order_ID), so the database
management system doesn’t know what to title the column in the
output.
To make this column more meaningful, we can use aliases. An
alias simply re-names a column. It is written as AS. To re-name the
column COUNT(Sales_Order_ID) to Count_Total_Orders, the query
would look like the following:
SELECT COUNT(Sales_Order_ID) AS Count_Total_Orders
FROM Sales_Orders
The output is more meaningful with the alias added in, shown in
Figure D-10:
FIGURE D-10
To create a query that would show the grand total quantity of

products ever sold (as stored in the Sales_Orders table) with a
meaningful column name, we could run the following:
SELECT SUM(Quantity_Sold) AS Total_Quantity_Sold
FROM Sales_Orders
Which returns the following output, shown in Figure D-11:
FIGURE D-11
Aggregates and Aliases Practice

1. Create a query that would show the average price of our inventory
items (use the Inventory table). Re-name the column in the output
Avg_Price.
2. Create a query that would show the total number of Customers we
have stored in the Customers table. Re-name the column in the
output Num_Customers.
EXTENDING THE QUERY WITH

GROUP BY AND HAVING CLAUSES
Aggregates are extremely useful to return grand totals of the data
that is stored in a database. But sometimes, we would prefer to view
that data by subtotals, as well.
Introduction to GROUP BY
In the introduction to aggregates, we worked through an example
that provided the grand total count of orders in the Sales Order table:
FROM Sales_Orders
That query results in a grand total of 10, but what if we would like
to see how that data splits up among customers who have ordered
from us? This is where GROUP BY comes in. GROUP BY works as
the “engine” that powers subtotaling the data. After the key word
GROUP BY, you indicate the attribute by which you would like to
slice the data. In this case, we want to slice the grand total by
CustomerID.
page 461
FROM Sales_Orders
GROUP BY CustomerID
The problem with this query, is that it does slice the data by
customer, but it doesn’t actually show us the CustomerID associated
with each subtotal. The output is shown in Figure D-12:
FIGURE D-12
If we want to actually view the CustomerID that is associated with
each subtotal, we need to not only put the attribute in the GROUP
BY field, but also add it to the SELECT field.
Remember from earlier in this tutorial, that the order in which you
place the attributes in the SELECT clause indicates the order that
those columns will display in the output. For this output, it would
make the most sense to see CustomerID before Count_Total,
because CustomerID is acting as a label for the totals. We can
modify the query to include CustomerID in the following way:
SELECT CustomerID, COUNT(Sales_Order_ID) AS
Count_Total_Orders
FROM Sales_Orders
GROUP BY CustomerID
This provides the following output, shown in Figure D-13:
FIGURE D-13
Similarly, we can extend the second example provided in the
Aggregates section that created a grand total of the quantity sold
from the Sales_Order table. If we would prefer to not see the grand
total quantity sold, but instead slice that total by InventoryID in order
to see the subtotal of the quantity of each inventory item sold, we
can create the following query:
SELECT InventoryID, SUM(Quantity_Sold) AS
Total_Quantity_Sold
FROM Sales_Orders
GROUP BY InventoryID
page 462
Which produces the following query, shown in Figure D-14:
FIGURE D-14
Notice that InventoryID needs to be added in two places: you

must place it in the GROUP BY clause to provide the “engine” that
subtotals a grandtotal (or slices it), and then you must also place
InventoryID in the SELECT clause so that you can see the labels
associated with each subtotal.
GROUP BY Practice
1. Create a query that would show the total quantity of items sold each
day. Re-name the aggregate Total_Quantity_Sold.
have stored in the Customers table, and group them by the State
the customers are from. Re-name the aggregate column in the
output Num_Customers.
Introduction to HAVING
Occasionally when running a query to gather subtotals (using a
GROUP BY clause), you do not want to see all of the results, but
instead would rather filter the results for certain subtotals.
Unfortunately, SQL cannot filter aggregate measures in the WHERE
clause, but fortunately, we have a different clause that can—
HAVING.
Any time you wish to filter your query results based on aggregate
values (e.g., SUM(Quantity_Sold), you can do so in the HAVING
clause.
For example, in the previous section about GROUP BY, we
created a query to see the total count of orders each customer had
been on. The output showed that the vast majority of our customers
had participated in only one order. But what if we wanted to only see
the customer(s) who had participated in more than one order?
We can create the following query to add in this filter:
SELECT CustomerID, COUNT(Sales_Order_ID) AS
Count_Total_Orders
FROM Sales_Orders
GROUP BY CustomerID
HAVING COUNT(Sales_Order_ID) > 1
As it turns out, there is only one customer who participated in
more than one order, as we can see in the query output, shown in
Figure D-15:
FIGURE D-15
The format of the HAVING clause is similar to WHERE:
HAVING aggregate(attribute) = number
page 463
The aggregate can be any of our aggregate values, SUM(), AVG(),

or COUNT().
The attribute is the field that you are aggregating, SUM(Quantity) or
COUNT(CustomerID).
The = can be replaced with any operator, =, <, >, =<, =>, <>.
The number is the value that you are filtering your results on.
Let’s work through another example. The second example in the
GROUP BY section showed the quantity sold of each inventory item.
If we want to view only those items that have sold less than 5 items,
we can create the following query:
SELECT InventoryID, SUM(Quantity_Sold) AS
Total_Quantity_Sold
FROM Sales_Orders
GROUP BY InventoryId
HAVING SUM(Quantity_Sold) < 5
This query produces the following output, shown in Figure D-16:
FIGURE D-16
HAVING Practice
1. Create a query that would show the total quantity of items sold each
day. Re-name the aggregate Total_Quantity_Sold. Show only the
days on which more than 6 items were sold.
have stored in the Customers table, and group them by the State
the customers are from. Re-name the aggregate column in the
output Num_Customers. Show only the states that more than one
customer is from.
EXTENDING THE USE OF THE FROM

CLAUSE: SELECTING DATA FROM
MORE THAN ONE TABLE
Some of the real power of SQL extends beyond relatively simple
SELECT FROM WHERE clauses. Since relational databases are
focused on reducing redundancy, there are often important details
that we would like to use for analysis stored across two or three
different tables.
For example, in our sample database, we may be interested to
know the phone number of the customer associated with each order.
Each order is stored in the Sales_Order table, but the details about
our customers (including their phone numbers) are stored in the
Customers table. To retrieve data from both tables, we need to first
make sure that the tables are related. We can do that by looking at
the database design:
The call-out circle and boxes in the figure can help us find how
these two tables are related. First, we can see the circle that
indicates the relationship connecting the Customers and
Sales_Orders table. This shows us that the two tables are indeed
related. The next step is to identify how they are page 464
related. The two red boxes in Figure D-17 indicate
the related fields, CustomerID is the Primary Key in the Customers
table, and CustomerID is the Foreign Key in the Sales_Order table.
Since these two tables are related, we can retrieve them fairly easily
with a JOIN clause.
FIGURE D-17
In order to retrieve data from more than one table, we need to

use SQL JOINs. There are three types of JOINs, but for much of our
analysis, an INNER JOIN will suffice. JOINs are technically part of
the FROM clause. They follow the following template:
FROM table1
INNER JOIN table2
ON table1.matching_key = table2.matching_key
The order of the tables does not matter, you could place the
Customers table in either the FROM or the INNER JOIN clause, and
the order of the tables does not matter in the ON clause. It just
matters that you indicate both tables you want to retrieve data from,
and that you indicate the two different tables with their matching keys
in the ON clause.
To select all of the data from the Customers table and the
Sales_Orders table, you can run the following query:
SELECT *
FROM Customers
INNER JOIN Sales_Orders
ON Customers.CustomerID = Sales_Orders.CustomerID
If you want to only select the Sales_Order_ID and the
Order_Date from the Sales_Orders table, but also select the State
attribute from the Customers table, you could run the following
query:
SELECT Sales_Order_ID, Order_Date, State
FROM Customers
INNER JOIN ON Practice

1. Create a query that will show the Customer’s First and Last Names,
as well as the Quantity_Sold and Price of each order the customer
was on.
2. Create a query that will show the Order_Date and Quantity_Sold on
each order, as well as the Inventory_Description of the items
associated with each order.
page 465
Parentheses Are Key to Joining More Than

One Table
Sometimes you will want to not only join two tables, but three or
more. When you join more than two tables together, you need to
nest the extra joins in parentheses.
In more detail, if you define the number of tables you’re trying to
join as “n,” then the number of parentheses you need after the word
FROM is n-2, and you need to have one right parentheses before
the start of each new join clause.
For example, if you are joining three tables, you need a
parentheses after the word FROM and after the first ON clause.
Then you can proceed with the query as normal. To join all three
tables in our example database, it would look like the following:
SELECT *
FROM (Customers
ON Customers.CustomerID = Sales_Orders.CustomerID)
INNER JOIN Inventory
ON Sales_Orders.InventoryID = Inventory.InventoryID
Note: there are other types of joins! Beyond INNER JOINs, we
can also create LEFT and RIGHT JOINs to get slightly different
results, depending on our data and our needs. There is a deep dive
to LEFT and RIGHT JOINs in Appendix H.
Putting It All Together

This tutorial has introduced you to the majority of the SQL keywords
you will need to extract data for data analysis or even to answer
simple data analysis questions directly in the database. If you were
to use all of the SQL words that we have discovered in this tutorial in
one query, the keywords must go in the following order:
SELECT
FROM
INNER JOIN
ON
WHERE
GROUP BY
HAVING
ORDER BY
I typically indent the INNER JOIN and ON clauses when drafting
queries to help to remember that those clauses are technically part
of the FROM clause, this helps with remembering the order of all of
the clauses.
page 466
Appendix E
SQLite
SQLite is a free open source database browser that may be installed

on any Windows or Mac computer, including labs, without
administrative privileges. It is a simple, open-source alternative to
Microsoft Access.
To get started, save the file Appendix E Data to a folder on your
computer. This way you can access it through SQLite.
Windows:
1. Go to https://sqlitebrowser.org/dl/.
2. Choose the DB Browser for SQLite - .zip (no installer) for 64-bit
Windows option.
3. In your Downloads folder, open the zip file and drag the DB
Browser for SQLite folder to the desktop.
4. Open the folder on the desktop and open DB Browser for
SQLite.exe.
Source: SQL
page 467
Mac:
1. Go to https://sqlitebrowser.org/dl/
2. Choose the DB Browser for SQLite option under macOS.
3. In your Downloads folder, open the DMG and drag DB Browser for
SQLite to your Applications or Desktop folder.
4. Open your Applications or Desktop folder and double-click DB
Browser for SQLite. If you receive a message indicating that the
app you are trying to install is not a Microsoft-verified app, select
Install Anyway.
Source: SQL
NAVIGATING SQLITE AND

EXECUTING SQL QUERIES
1. Once you have DB Browser for SQLite open, you can select Open
Database to open an existing database file.
Source: SQL
page 468
2. Navigate to the Appendix E Data file to open it.
3. The first view will be of the Database Structure. You can expand the
tables to get a glimpse of the attributes that are stored in each table
and their datatypes.
Source: SQL
4. Clicking into the Browse Data tab will allow you to switch views
between tables to view the data that is stored in each table.
Source: SQL
page 469
5. To execute a SQL query, click the Execute SQL tab.
Source: SQL
6. Add the text of your query in the open box.

7. To execute your query, click the blue arrow, or you can click F5 on
your keyboard as a keyboard shortcut.
8. Once the query has executed, you will see the results, as well as a
window describing that your query ran successfully. If there was an
error in your query, the third window would indicate where the error
was.
Source: SQL
Microsoft applications (such as Excel or Power BI) do not have a

default mechanism for connecting to SQLite files, but if you just want
to explore data using SQL, SQLite is a great tool for working with
queries and exploring data.
page 470
Appendix F
Power Query
Excel’s Get and Transform tools are a part of the Power BI suite that
is integrated into Excel 2016. These tools allow you to connect
directly to a dataset stored in a variety of locations, including other
Excel files; .csv files; the web; and a multitude of relational
databases, including Microsoft Access, SQL Server, Teradata,
Oracle, PostGreSQL, and MySQL.
Throughout this text, the majority of the times we analyze the
Dillard’s dataset in the Comprehensive Labs, we will load the data
from SQL Server into Excel using this Get and Transform tool.
When we extract the data, we may want to extract entire tables,
or we may want to extract only a portion via a SQL query.
In this appendix, we will connect to the Dillard’s data. The
Dillard’s data is stored on the University of Arkansas’ remote
desktop, so make sure to log in to the desktop in order to work
through these steps. Ask your instructor for login information if you
do not have it already.
CONNECT TO SQL SERVER
THROUGH EXCEL’S GET AND
TRANSFORM TOOL
1. Open Excel in the University of Arkansas remove desktop. From the
Data tab on the ribbon, click New Query. Then select From
Database > From SQL Server Database.
2. The following box will pop up, into which you should page 471
provide the name of the Server and the Database
name that your instructor provides you. For the majority of the
exercises, we use the Database name WCOB_DILLARDS. Labs 3-
4, 3-5, and 4-4 use a different database, titled
UA_DILLARDS_2016.
Source: SQL
If prompted to input your credentials, select Use my Current

Credentials. Once you have input the Server and Database name,
you have two options:
3. Extract entire tables by clicking OK. Continue to Step 5.
4. Extract only a portion of data from one or more tables based on the
criteria of a SQL query. To do so, click Advanced Options. Skip to
Step 7.
To Extract Entire Tables

5. Click OK.
Source: SQL
6. Select the table(s) that you would like to load into page 472
Excel. If you would like to select more than one table,
place a checkmark in the box next to Select multiple items.
Source: SQL
To Extract a Portion of the Data

7. Click Advanced Options to input your SQL query. Input your query
in the space provided and click OK.
Source: SQL
page 473
EDITING THE DATA IN POWER

QUERY
Regardless of whether you extracted entire tables or extracted data
based on a query, you can either Load the data directly into Excel, or
you can Edit the data in Power Query first.
Clicking Load will load the data directly into an Excel table.
Clicking Edit will open the Power Query window for you to
transform the data before it is loaded into Excel (add or delete
columns, remove or transform null values, aggregate data, etc.)
8. To Edit (Transform) the data, click Edit.
9. The Power Query ribbon has several tabs that provide useful ways
to transform the data. A few of the buttons that we use throughout
the text are called out for the Home tab and the Transform tab on
the ribbon below.
Home tab on the ribbon:
Click the Close & Load button when you are finished transforming
the data to load it into Excel.
The Remove Rows button provides options to remove rows with
nulls in selected columns, with duplicates in selected columns, or
based on other criteria
Transform tab on the ribbon:
page 474
Replace Values functions the same way in Power Query as it does

in Excel, except the transformation is stored and thus repeatable
when created in Power Query.
Pivot Column creates two new columns out of an existing category
column (for example, we can pivot the Transaction_Type column by
the transaction amount.
The Date button will allow you to transform an existing date column
into a date part (year, month, day, etc.) or change the date format. It
is also useful to create duplicates of existing date columns, then
transform the copies into the date parts.
AFTER LOADING THE DATA INTO

EXCEL: HOW TO CONTINUE
WORKING WITH YOUR DATA
Once you are finished transforming your data in the Power Query
Editor and you click the Close & Load button, your data will begin to
load into a worksheet.
What to Do if the Load to the Worksheet Fails

This process may take several minutes depending on how large of a
datafile you are loading.
Sometimes, the query load will fail—this will often occur due to an
attempt to load more than 1,048,576 records. If the load fails, you
can hover over the error message to view what caused the error. In
this case, the query result was too large, so we can select Load to
Data Model.
Loading the data to the data model will allow us to work with a
large dataset in a PivotTable, even though the dataset itself is too
large for the worksheet.
page 475
How to Return to the Power Query Window

after Closing It
If you wish to further transform your data using the Power Query
window, you can do so by double-clicking on the Query label in the
Workbook Queries window (if the Workbook Queries pane is not
showing, you can click the Data tab on the ribbon, then click Show
Queries).
page 476
Appendix G
Tableau
Before jumping into the labs, you may wish to introduce yourself to
Tableau through this appendix if you have never used the Tableau
tool.
To access Tableau, you can use the University of Arkansas’
remote desktop (see your instructor for instructions on how to access
it), or you can download a free academic usage license of Tableau
by following this URL: https://www.tableau.com/academic/students.
Tableau will work on a PC or a Mac. The images in this textbook will
reflect Tableau for PC, but it is very similar to Tableau for Mac.
Tableau can connect to a variety of datatypes, including Excel,
Access, and SQL Server. We will connect to the dataset Appendix
G Data.xlsx. If you worked through Appendix C about PivotTables,
this is the same dataset that you worked with previously.
1. Open Tableau.
2. Immediately upon opening Tableau, you will see a list of file types
that you can connect to. We’ll connect to an Excel file, so click
Microsoft Excel.
page 477
3. Navigate to where your file is stored and click Open.
Tableau automatically detects the data types of the attributes
you import. In this dataset, the attributes probably all imported as
the data type you would expect. Notice that the first two, Invoice #
and Customer # imported as number. Continue looking at the
attributes, and you will notice the globe icon above Zip Code. This
is Tableau showing you one of its best features, it shows that the
Zip Code data was imported as geographic data. This will allow
you to create maps.
4. To begin working with the data, click Sheet 1 in the bottom left.
Here is a quick introduction to pieces of the Tableau canvass:
page 478
5. To begin working with the data, double-click on the
measure Gross Margin.
Immediately you will see how Tableau interacts with data

differently than Excel because it has defaulted to displaying a bar
chart. This isn’t a very meaningful chart as it is, but you can add
meaning by adding a dimension.
6. Double-click Description from the dimensions.
Similar to the analysis found in Appendix C regarding PivotTables,
you find the same numbers: Apples with a gross margin of 140.4,
apricots with a gross margin of 78.0, and so on.
page 479
7. To make this data easier to interpret, you can sort it.
Click the Sort Descending icon to sort the data.
You can continue adding pills to the columns or rows shelves

and changing the method of visualization using the Show Me tab to
further familiarize yourself with the tool.
page 480
Appendix H
SQL Part 2
In Appendix D, you learned about many key terms in SQL, including

how to join tables. The purpose of joining tables is to enable you to
retrieve data that is stored in more than one table all at once. The
join type that you learned about in Appendix D is an INNER JOIN.
There are two other popular join types, though, LEFT and RIGHT.
We will work with the same Access Database that you used in
Appendix D. Although it contains the same data, you can access it
through the Appendix H Data.accdb.
We’ll start with bringing this data into Tableau.
1. Open Tableau.
2. Select Microsoft Access to connect to the file and navigate to
where you have stored the file, then click Open.
3. In the Data Source view, drag both the Customers table and the
Sales_Orders tables to the Drag tables here section.
EXPLANATION OF INNER JOINS

Notice the Venn diagram that appears connecting the two tables:
And if you click the Venn diagram, you can see the following
details about how the tables are related:
Tableau has defaulted to joining these two tables with an INNER

join, and it has accurately identified the two keys that are related
between the two tables, Customer ID in the Customers table, and
CustomerID in the Sales Order table.
page 481
This is very similar to how you would write a query to gather the
same information directly in the Access database, where one of the
tables is indicated in the FROM clause, the second table is indicated
in the INNER JOIN clause, and the keys that are common between
the two tables are indicated with an equal sign between them in the
ON clause:
SELECT *
FROM Customers
As the Venn diagram suggests, an INNER join will show all of the
data for which there is a match between the two tables. However, it
is important to notice what that means it leaves out – it will not return
any of the data for which there is NOT a match between the two
tables.
In this instance, there is actually one customer held in the
Customers table that is not included in the Sales_Orders table
(Customer 3, Edna Orgeron). Why would this happen? Perhaps this
fictional company records data on potential customers, so even
though someone may have not actually purchased anything yet, the
company can still contact them. Whatever the reason might be—the
fact that CustomerID 3 does not exist in both tables—CustomerID 3
will not be included in the results.
If the above SQL query were to be run, the following result would
return:
Source: SQL
Notice that the red box surrounding the records for customers 2
and 4 do not include anything for customer 3.
page 482
EXPLANATION OF LEFT JOINS

If we wanted to see all of the data from the Customers table, even if
there isn’t a match in the Sales_Order table, then we need to change
our join type.
Back to the Tableau image—if we click into the Venn Diagram,
we can change the join type to a LEFT join.
The red box indicates an important change that occurs as soon

as you we made the change to a LEFT join—Customer 3 is included!
But not only that, while we see Customer 3’s name and contact
information, we see null values for any attributes from the
Sales_Order table. That is because there isn’t any corresponding
information for Customer 3 in the Sales_Order table.
To replicate this query in Access, the only change that needs to
be made is swap the word INNER with LEFT:
SELECT *
FROM Customers
LEFT JOIN Sales_Orders
It is easier to visualize how joins are created in Tableau, but they

work the same way in SQL. The table that you place in the FROM
clause is the “left” table, and the table that you place in the JOIN
clause is the “right table.”
EXPLANATION OF RIGHT JOINS

Looking back to the Venn diagram, we can see that a RIGHT join
would return the opposite result of a left join. In this specific instance,
if there were any Sales_Orders that had a CustomerID on them that
were not associated with a CustomerID in the Customers table, the
only we would see them is if we created a RIGHT join.
We dive more deeply into this concept in the text when we
discuss audit analytics, but perhaps you can already imagine how
this type of join would be useful for detecting errors or fraud—we
definitely would want to isolate any sales orders that had Customer
information on them that didn’t align with our verified customer
listing!
page 483
Appendix I
Power BI
Power BI Desktop is a Microsoft tool that combines ETL tools with

reporting tools. When we work with Power Query or PowerPivot in
Excel, we’re actually working with Power BI tools. If you will
ultimately want to run statistical tests such as hypothesis testing or
regression analysis, it’s best to work within Excel directly and use the
Power Query add-in. However, if you need to transform your data
using Power Query prior to creating reports or dashboards or even if
you just want to explore your data, Power BI Desktop can be a great
alternative to other reporting and visualization tools such as Tableau.
When it comes to creating visualizations in Power BI, you can
create extremely similar results to what you can create in Tableau,
but the path to getting there is different. Power BI defaults to a report
mode (similar to Tableau’s Dashboard mode), so that as you create
visuals, they appear as tiles that you can resize and rearrange
around the canvas.
When you open Power BI Desktop, you will be greeted with a
startup screen similar to the following:
Source: Powe BI Desktop
page 484
The tutorials and other training resources on the right of the

startup screen are helpful for getting started with the tool.
The Get Data button on the left of the startup screen will bring
you into Power BI’s Power Query tool. It is set up exactly like the
Power Query tool is set up in Excel, so you can use it to connect to a
variety of sources (Excel, SQL Server, Access, etc.).
To familiarize yourself with Power BI, we will use the Appendix I
data.xlsx. It is a modified version of the Slainte_Subset file that you
might work with in Lab 2-2, Lab 4-2, or Lab 6-2. The data are a
subset of the sales data related to a fictional brewery named Slainte.
1. Click Get Data on the startup screen.
2. Select Excel from the list of possible data sources, then click
Connect.
3. Browse to the file location for Power BI_Appendix.xlsx and Open

the file.
4. Because there are three spreadsheets in the file, the Navigator
provides you the option to select 1, 2, or all of the spreadsheets.
Place checkmarks in each.
5. You are also given an option to either Load or Edit the data. If you
click Edit, you will enter the Power Query window with the same
ribbon and options to transform the data as you are familiar with
from the Excel version of the tool (add columns, split columns, pivot
data, etc.). This data does not need to be transformed, so we will
click Load.
page 485
6. Once the data is loaded, you will see a blank canvas
on which you can build a report. There are three key elements
bordering the canvas:
a. To the left of the blank canvas, you are presented with three
options:
Report Mode: The first option, represented with an icon that looks
like a bar chart, is for Report mode. This is the default view and is
where you can build your visualizations and explore your data.
Data Mode: The second option, represented with an icon that
looks like a table or a spreadsheet, is for Data mode. If you click
into this icon, you can view the raw data that you have imported
into Power BI. You can also create new measures or new columns
from this mode.
Model Mode: The third option, which looks like a database
diagram, is for Model mode. If you click into this icon, you enter
PowerPivot. From this mode, you can edit the table and attribute
names or edit relationships between tables.
page 486
b. To the right of the blank canvas is your Fields list and
your options for Visualizations.
Visualizations: You can drag any of these options over into the
canvas to begin designing a visualization. Once you have tiles on
your report, you can change the type of visualization being used to
depict a set of fields by clicking the tile then selecting any of the
visualization options to change the way the data is presented.
Fields: This section is similar to your PivotTable field list. You can
expand the tables to see the attributes that are within each and
placing a check mark in the fields will add them to an active tile.
Values, Filters, etc.: this section will vary based on the tile and
the fields you are actively working with. Any time you add a field to
a visualization, that field gets automatically added to the filters,
which cuts out the need to manually add filters or slicers to your
PivotTable.
c. Immediately above the canvas is the familiar ribbon that you can
expect from Microsoft applications. The four tabs—Home, View,
Modeling, and Help—stay consistent across the three different
modes (report, data, and model), but the options that you can
select will vary based on the mode in which you are working.
page 487
7. To begin working with the data, Expand the Customer
table to place a checkmark in the State field.
8. Power BI will default to creating a tile with a map visualization. This

is similar to how Tableau defaults to working with geographic data.
To make the map more interesting, expand the Sales_Orders table
to place a checkmark in the Quantity Sold field.
page 488
This will make the tile more interesting by changing
the size of the symbol associated with each state—the larger the
symbol, the higher the quantity sold in that state.
9. You can also change the way the data is presented by selecting a
different visualization type. Select the first option to view the data in
a horizontal bar chart.

0. One of the most exciting offerings from Power BI is its natural
language processing for data exploration. From the Home tab in the
ribbon, click Buttons. In the drop-down, select Q&A.
1. The following icon will appear as a separate tile. If the placement

defaults to being on top of the bar chart, you can click and drag it to
somewhere else on the canvas:
page 489
2. To activate the Q&A, ctrl + click the icon. The
following window will pop-up, and you can select from the list of
questions that Power BI has come up with, or you can type directly
into the “Ask a question about your data” box.
page 490
3. You can also add a question directly to the canvas by
selecting Ask a Question from the Home tab on the ribbon. In the
screenshot that follows, I typed “sum of quantity sold by state” to get
a replicated visualization of the first visualization we made by
manually selecting the fields State and Quantity Sold:
There are many other exciting benefits that Power BI can do,
but with this introduction you should have the confidence to jump in
and explore more that Power BI has to offer. To get more step-by-
step help with Power BI, you can work through the final portion of
Lab 4-3, which replicates a Tableau lab in Power BI.
page 491
Appendix J
Dillard’s ER Diagram
page 492
Appendix K
Data Dictionaries
This appendix contains the data dictionary and validation statistics

for the datasets used in this textbook.
CollegeScorecard
Lending Club
OK PCard
COLLEGESCORECARD
UNITID A unique identifier for the institution
INSTNM Institution name
CITY City
STABBR State postcode
CONTROL 1 = Public. 2 = Private nonprofit. 3 = Private for-
profit
CCBASIC Carnegie Classification, basic:
—2 Not applicable
0 (Not classified)
1 Associate’s Colleges: High Transfer-High
Traditional
2 Associate’s Colleges: High Transfer-Mixed
Traditional/Nontraditional
3 Associate’s Colleges: High Transfer-High
Nontraditional
4 Associate’s Colleges: Mixed
Transfer/Vocational & Technical-High Traditional
Transfer/Vocational & Technical-Mixed
Traditional/Nontraditional
Transfer/Vocational & Technical-High
Nontraditional
7 Associate’s Colleges: High Vocational &
Technical-High Traditional
Technical-Mixed Traditional/Nontraditional
Technical-High Nontraditional
10 Special Focus Two-Year: Health Professions
11 Special Focus Two-Year: Technical
Professions
12 Special Focus Two-Year: Arts & Design
13 Special Focus Two-Year: Other Fields
14 Baccalaureate/Associate’s Colleges:
Associate’s Dominant
15 Doctoral Universities: Highest Research
Activity
16 Doctoral Universities: Higher Research Activity
17 Doctoral Universities: Moderate page 493
Research Activity
18 Master’s Colleges & Universities: Larger
Programs
19 Master’s Colleges & Universities: Medium
Programs
20 Master’s Colleges & Universities: Small
Programs
21 Baccalaureate Colleges: Arts & Sciences
Focus
22 Baccalaureate Colleges: Diverse Fields
23 Baccalaureate/Associate’s Colleges: Mixed
Baccalaureate/Associate’s
24 Special Focus Four-Year: Faith-Related
Institutions
25 Special Focus Four-Year: Medical Schools &
Centers
26 Special Focus Four-Year: Other Health
Professions Schools
27 Special Focus Four-Year: Engineering Schools
28 Special Focus Four-Year: Other Technology-
Related Schools
29 Special Focus Four-Year: Business &
Management Schools
30 Special Focus Four-Year: Arts, Music &
Design Schools
31 Special Focus Four-Year: Law Schools
32 Special Focus Four-Year: Other Special Focus
Institutions
33 Tribal Colleges
ADM_RATE Admission rate
SAT_AVG Average equivalent SAT of students admitted
UGDS Enrollment of undergraduate certificate/degree-
seeking students
UGDS_WHITE Total share of enrollment of undergraduates who
are White
UGDS_BLACK Total share of enrollment of undergraduates who
are Black
UGDS_HISP Total share of enrollment of undergraduates who
are Hispanic
UGDS_ASIAN Total share of enrollment of undergraduates who
are Asian
UGDS_AIAN Total share of enrollment of undergraduates who
are American Indian/Alaska Native
UGDS_NHPI Total share of enrollment of undergraduates who
are Native Hawaiian/Pacific Islander
UGDS_2MOR Total share of enrollment of undergraduates who
are two or more races
UGDS_NRA Total share of enrollment of undergraduates who
are nonresident aliens
UGDS_UNKN Total share of enrollment of undergraduates
whose race is unknown
PPTUG_EF Share of undergraduate degree/certificate-
seeking students who are part-time
NPT4_PUB Average net price for Title IV institutions (public)
NPT4_PRIV Average net price for Title IV institutions (private
for-profit and nonprofit)
COSTT4_A Average cost of attendance
TUITFTE Net tuition revenue per full-time equivalent
student
INEXPFTE Instructional expenditures per full-time equivalent
student
PFTFAC Proportion of faculty that is full-time
PCTPELL Percentage of undergraduates who receive a Pell
Grant
C150_4 Completion rate for first-time, full-time students at
four-year institutions (six-year)
PFTFTUG1_EF Share of undergraduate students who are first-
time, full-time, degree seeking undergraduates
RET_FT4 First-time, full-time student retention rate at four-
year institutions
PCTFLOAN Percent of all federal undergraduates receiving a
federal student loan
page 494
LENDING CLUB
LoanStats
LoanStatNew Description
acc_now_delinq The number of
accounts on which the
borrower is now
delinquent
acc_open_past_24mths Number of trades
opened in past 24
months
addr_state The state provided by
the borrower in the
loan application
all_util Balance to credit limit
on all trades
annual_inc The self-reported
annual income
provided by the
borrower during
registration
annual_inc_joint The combined self-
reported annual
income provided by
the co-borrowers
during registration
application_type Indicates whether the
loan is an individual
application or a joint
application with two
co-borrowers
avg_cur_bal Average current
balance of all
accounts
bc_open_to_buy Total open to buy on
revolving bankcards
bc_util Ratio of total current
balance to high
credit/credit limit for all
bankcard accounts.
chargeoff_within_12_mths Number of charge-offs
within 12 months
collection_recovery_fee post charge off
collection fee
collections_12_mths_ex_med Number of collections
in 12 months
excluding medical
collections
delinq_2yrs The number of 30+
days past-due
incidences of
delinquency in the
borrower’s credit file
for the past 2 years
delinq_amnt The past-due amount
owed for the accounts
on which the borrower
is now delinquent.
desc Loan description
provided by the
borrower
dti A ratio calculated
using the borrower’s
total monthly debt
payments on the total
debt obligations,
excluding mortgage
and the requested LC
loan, divided by the
borrower’s self-
reported monthly
income
dti_joint A ratio calculated
using the co-
borrowers’ total
monthly payments on
the total debt
obligations, excluding
mortgages and the
requested LC loan,
divided by the co-
borrowers’ combined
self-reported monthly
income
earliest_cr_line The month the
borrower’s earliest
reported credit line
was opened
emp_length Employment length in
years. Possible values
are between 0 and 10
where 0 means less
than one year and 10
means 10 or more
years.
emp_title The job title supplied
by the borrower when
applying for the loan.*
fico_range_high The upper boundary
range the borrower’s
FICO at loan
origination belongs to
fico_range_low The lower boundary
FICO at loan
origination belongs to
funded_amnt The total amount
committed to that loan
at that point in time
page 495
funded_amnt_inv The total amount
committed by
investors for that loan
at that point in time
grade LC assigned loan
grade
home_ownership The home ownership
status provided by the
borrower during
registration or
obtained from the
credit report; our
values are RENT,
OWN, MORTGAGE,
OTHER
id A unique LC assigned
ID for the loan listing
il_util Ratio of total current
balance to high
credit/credit limit on all
install accounts
initial_list_status The initial listing
status of the loan;
possible values are
W, F
inq_fi Number of personal
finance inquiries
inq_last_12m Number of credit
inquiries in past 12
months
inq_last_6mths The number of
inquiries in past 6
months (excluding
auto and mortgage
inquiries)
installment The monthly payment
owed by the borrower
if the loan originates
int_rate Interest rate on the
loan
issue_d The month in which
the loan was funded
last_credit_pull_d The most recent
month LC pulled
credit for this loan
last_fico_range_high The upper boundary
last FICO pulled
belongs to
last_fico_range_low The lower boundary
last FICO pulled
belongs to
last_pymnt_amnt Last total payment
amount received
last_pymnt_d Last month payment
was received
loan_amnt The listed amount of
the loan applied for by
the borrower. If at
some point in time,
the credit department
reduces the loan
amount, then it will be
reflected in this value.
loan_status Current status of the
loan
max_bal_bc Maximum current
balance owed on all
revolving accounts
member_id A unique LC assigned
Id for the borrower
member
mo_sin_old_il_acct Months since oldest
bank installment
account opened
mo_sin_old_rev_tl_op Months since oldest
revolving account
opened
mo_sin_rcnt_rev_tl_op Months since most
recent revolving
account opened
mo_sin_rcnt_tl Months since most
recent account
opened
mort_acc Number of mortgage
accounts
mths_since_last_delinq The number of
months since the
borrower’s last
delinquency
mths_since_last_major_derog Months since most
recent 90-day or
worse rating
mths_since_last_record The number of
months since the last
public record.
mths_since_rcnt_il Months since most
recent installment
accounts opened
mths_since_recent_bc Months since most
recent bankcard
account opened.
mths_since_recent_bc_dlq Months since most
recent bankcard
delinquency
mths_since_recent_inq Months since most
recent inquiry.
mths_since_recent_revol_delinq Months since most
recent revolving
delinquency.
next_pymnt_d Next scheduled
payment date
page 496
num_accts_ever_120_pd Number of accounts
ever 120 or more
days past due
num_actv_bc_tl Number of currently
active bankcard
accounts
num_actv_rev_tl Number of currently
active revolving trades
num_bc_sats Number of
satisfactory bankcard
accounts
num_bc_tl Number of bankcard
accounts
num_il_tl Number of installment
accounts
num_op_rev_tl Number of open
revolving accounts
num_rev_accts Number of revolving
accounts
num_rev_tl_bal_gt_0 Number of revolving
trades with balance
>0
num_sats Number of
satisfactory accounts
num_tl_120dpd_2m Number of accounts
currently 120 days
past due (updated in
past 2 months)
num_tl_30dpd Number of accounts
currently 30 days past
due (updated in past 2
months)
num_tl_90g_dpd_24m Number of accounts
90 or more days past
due in last 24 months
num_tl_op_past_12m Number of accounts
opened in past 12
months
open_acc The number of open
credit lines in the
borrower’s credit file
open_acc_6m Number of open
trades in last 6
months
open_il_12m Number of installment
accounts opened in
past 12 months
open_il_24m Number of installment
accounts opened in
past 24 months
open_act_il Number of currently
active installment
trades
open_rv_12m Number of revolving
trades opened in past
12 months
open_rv_24m Number of revolving
trades opened in past
24 months
out_prncp Remaining
outstanding principal
for total amount
funded
out_prncp_inv Remaining
outstanding principal
for portion of total
amount funded by
investors
pct_tl_nvr_dlq Percent of trades
never delinquent
percent_bc_gt_75 Percentage of all
bankcard accounts >
75% of limit
policy_code Publicly available
policy_code=1;new
products not publicly
available
policy_code=2
pub_rec Number of derogatory
public records
pub_rec_bankruptcies Number of public
record bankruptcies
purpose A category provided
by the borrower for
the loan request.
pymnt_plan Indicates if a payment
plan has been put in
place for the loan
recoveries Post charge-off gross
recovery
revol_bal Total credit revolving
balance
revol_util Revolving line
utilization rate, or the
amount of credit the
borrower is using
relative to all available
revolving credit.
sub_grade LC assigned loan
subgrade
tax_liens Number of tax liens
term The number of
payments on the loan;
values are in months
and can be either 36
or 60
title The loan title provided
by the borrower
tot_coll_amt Total collection
amounts ever owed
page 497
tot_cur_bal Total current balance
of all accounts
tot_hi_cred_lim Total high credit/credit
limit
total_acc The total number of
credit lines currently in
the borrower’s credit
file
total_bal_ex_mort Total credit balance
excluding mortgage
total_bal_il Total current balance
of all installment
accounts
total_bc_limit Total bankcard high
credit/credit limit
total_cu_tl Number of finance
trades
total_il_high_credit_limit Total installment high
credit/credit limit
total_pymnt Payments received to
date for total amount
funded
total_pymnt_inv Payments received to
date for portion of
total amount funded
by investors
total_rec_int Interest received to
date
total_rec_late_fee Late fees received to
date
total_rec_prncp Principal received to
date
total_rev_hi_lim Total revolving high
credit/credit limit
url URL for the LC page
with listing data
verification_status Indicates if income
was verified by LC,
not verified, or if the
income source was
verified
verified_status_joint Indicates if the co-
borrowers’ joint
income was verified
by LC, not verified, or
if the income source
was verified
zip_code The first 3 numbers of
the zip code provided
by the borrower in the
loan application.
revol_bal_joint Sum of revolving
credit balance of the
co-borrowers, net of
duplicate balances
sec_app_fico_range_low FICO range (high) for
the secondary
applicant
sec_app_fico_range_high FICO range (low) for
the secondary
applicant
sec_app_earliest_cr_line Earliest credit line at
time of application for
the secondary
applicant
sec_app_inq_last_6mths Credit inquiries in the
last 6 months at time
of application for the
secondary applicant
sec_app_mort_acc Number of mortgage
accounts at time of
application for the
secondary applicant
sec_app_open_acc Number of open
trades at time of
application for the
secondary applicant
sec_app_revol_util Ratio of total current
balance to high
credit/credit limit for all
revolving accounts
sec_app_open_act_il Number of currently
active installment
trades at time of
application for the
secondary applicant
sec_app_num_rev_accts Number of revolving
accounts at time of
application for the
secondary applicant
sec_app_chargeoff_within_12_mths Number of charge-offs
within last 12 months
at time of application
for the secondary
applicant
sec_app_collections_12_mths_ex_med Number of collections
within last 12 months
excluding medical
collections at time of
application for the
secondary applicant
sec_app_mths_since_last_major_derog Months since most
recent 90-day or
worse rating at time of
application for the
secondary applicant
page 498
hardship_flag Flags whether or not
the borrower is on a
hardship plan
hardship_type Describes the
hardship plan offering
hardship_reason Describes the reason
the hardship plan was
offered
hardship_status Describes if the
hardship plan is
active, pending,
canceled, completed,
or broken
deferral_term Amount of months
that the borrower is
expected to pay less
than the contractual
monthly payment
amount due to a
hardship plan
hardship_amount The interest payment
that the borrower has
committed to make
each month while they
are on a hardship plan
hardship_start_date The start date of the
hardship plan period
hardship_end_date The end date of the
hardship plan period
payment_plan_start_date The day the first
hardship plan
payment is due—for
example, if a borrower
has a hardship plan
period of 3 months,
the start date is the
start of the three-
month period in which
the borrower is
allowed to make
interest-only
payments
hardship_length The number of
months the borrower
will make smaller
payments than
normally obligated
due to a hardship plan
hardship_dpd Account days past
due as of the hardship
plan start date
hardship_loan_status Loan status as of the
hardship plan start
date
orig_projected_additional_accrued_interest The original projected
additional interest
amount that will
accrue for the given
hardship payment
plan as of the
Hardship Start Date;
this field will be null if
the borrower has
broken his/her
hardship payment
plan
hardship_payoff_balance_amount The payoff balance
amount as of the
hardship plan start
date
hardship_last_payment_amount The last payment
amount as of the
hardship plan start
date
disbursement_method The method by which
the borrower receives
his/her loan; possible
values are CASH,
DIRECT_PAY
debt_settlement_flag Flags whether or not
the borrower, who has
charged-off, is
working with a debt-
settlement company.
debt_settlement_flag_date The most recent date
that the
Debt_Settlement_Flag
has been set
settlement_status The status of the
borrower’s settlement
plan; possible values
are COMPLETE,
ACTIVE, BROKEN,
CANCELLED,
DENIED, DRAFT
settlement_date The date that the
borrower agrees to
the settlement plan
settlement_amount The loan amount that
the borrower has
agreed to settle for
settlement_percentage The settlement
amount as a
percentage of the
payoff balance
amount on the loan
settlement_term The number of
months that the
borrower will be on
the settlement plan
page 499
RejectStats
RejectStats
Description
File
Amount The total amount requested by the borrower
Requested
Application The date which the borrower applied
Date
Loan Title The loan title provided by the borrower
Risk_Score For applications prior to November 5, 2013, the risk
score is the borrower’s FICO score; for applications
after November 5, 2013, the risk score is the
borrower’s Vantage score
Debt-To- A ratio calculated using the borrower’s total monthly
Income debt payments on the total debt obligations,
Ratio excluding mortgage and the requested LC loan,
divided by the borrower’s self-reported monthly
income
Zip Code The first 3 numbers of the zip code provided by the
borrower in the loan application
State The state provided by the borrower in the loan
application
Employment Employment length in years; possible values are
Length between 0 and 10, where 0 means less than one
year and 10 means 10 or more years.
Policy Code publicly available policy_code=1;new products not
publicly available policy_code=2
OK PCARD
SourceYearMonth Year and month of the transaction
SourceType Data source, e.g. Bank
CardholderLastName Cardholder’s last name
CardholderFirstInitial Cardholder’s first initial
ItemDescription Description of item purchased
Amount Purchase amount in U.S. dollars
BusinessUnit Identifier for business unite
MerchantName Supplier’s name, provided by the issuing
bank
TransactionDate Date of purchase
PostedDate Date of transaction posting to account
MCCDescription Merchant Category Code, provided by the
issuing bank
page 500
Glossary
2018 Tax Cuts and Jobs Act Tax Reform (366) Tax legislation
offering a major change to the existing tax code.
A
audit data standards (ADS) (205) The audit data standards define
common tables and fields that are needed by auditors to perform
common audit tasks. The AICPA developed these standards.
B
Balanced Scorecard (271) A particular type of digital dashboard
that is made up of strategic objectives, as well as KPIs, target
measures, and initiatives, to help the organization reach its target
measures in line with strategic goals.
Benford’s law (103) An observation about the frequency of
leading digits in many real-life sets of numerical data. The law states
that in many naturally occurring collections of numbers, the
significant lending digit is likely to be small.
Big Data (4) Datasets that are too large and complex for
businesses’ existing systems to handle utilizing their traditional
capabilities to capture, store, manage, and analyze these datasets.
C
causal modeling (106) A data approach similar to regression, but
used when the relationship between independent and dependent
variables where it is hypothesized that the independent variables
cause or are associated with the dependent variable.
classification (9, 106) A data approach that attempts to assign
each unit in a population into a few categories potentially to help with
predictions.
clustering (10, 106) A data approach that attempts to divide
individuals (like customers) into groups (or clusters) in a useful or
meaningful way.
co-occurrence grouping (10) A data approach that attempts to
discover associations between individuals based on transactions
involving them.
common data model (10, 103) A tool used to map existing
database tables and fields from various systems to a standardized
set of tables and fields for use with analytics.
common size financial statement (328) A type of financial
statement that contains only basic accounts that are common across
companies.
composite primary key (43) A special case of a primary key that
exists in linking tables. The composite primary key is made up of the
two primary keys in the table that it is linking.
computer-assisted audit techniques (CAATs) (228) Automated
scripts that can be used to validate data, test controls, and enable
substantive testing of transaction details or account balances and
generate supporting evidence for the audit.
continuous auditing (208) A process that provides real-time
assurance over business processes and systems.
continuous data (143) One way to categorize quantitative data, as
opposed to discrete data. Continuous data can take on any value
within a range. An example of continuous data is height.
continuous monitoring (208) A process that constantly evaluates
internal controls and transactions and is the chief responsibility of
management.
continuous reporting (208) A process that provides real-time
access to the system status and accounting information.
D
Data Analytics (4) The process of evaluating data with the
purpose of drawing conclusions to address business questions.
Indeed, effective Data Analytics provides a way to search through
large structured and unstructured data to identify unknown patterns
or relationships.
data dictionary (14, 44) Centralized repository of descriptions for
all of the data attributes of the dataset.
data mart (365) A subset of the data warehouse focused on a
specific function or department to assist and support its needed data
requirements.
data reduction (10, 98) A data approach that attempts to reduce
the amount of information that needs to be considered to focus on
the most critical items (i.e., highest cost, highest risk, largest impact,
etc.).
data request form (47) A method for obtaining data if you do not
have access to obtain the data directly yourself.
data warehouse (204, 365) A data warehouse is a repository of
data accumulated from internal and external data sources, including
financial data, to help management decision making.
page 501
decision boundaries (109) Technique used to mark the split

between one class and another.
Decision support system (112) An information system that
supports decision-making activity within a business by combining
data and expertise to solve problems and perform calculations.
decision tree (109) Tool used to divide data into smaller groups.
declarative visualizations (144) Made when the aim of your
project is to “declare” or present your findings to an audience. Charts
that are declarative are typically made after the data analysis has
been completed and are meant to exhibit what was found in the
analysis steps.
descriptive analytics (94, 228) Procedures that summarize
existing data to determine what has happened in the past. Some
examples include summary statistics (e.g. Count, Min, Max,
Average, Median), distributions, and proportions.
descriptive attributes (43) Attributes that exist in relational
databases that are neither primary nor foreign keys. These attributes
provide business information, but are not required to build a
database. An example would be “Company Name” or “Employee
Address.”
diagnostic analytics (94, 228) Procedures that explore the current
data to determine why something has happened the way it has,
typically comparing the data to a benchmark. As an example, these
allow users to drill-down in the data and see how it compares to a
budget, a competitor, or trend.
digital dashboard (270) An interactive report showing the most
important metrics to help users understand how a company or an
organization is performing. Often created using Excel or Tableau.
discrete data (143) One way to categorize quantitative data, as
opposed to continuous data. Discrete data are represented by whole
numbers. An example of discrete data is points in a basketball game.
DuPont ratio (330) Ratios developed by the DuPont Corporation
to decompose return on equity (ROE) into its component ratios:
Profit margin x Asset turnover x Equity multiplier.
E
ETL (45) The extract, transform, and load process that is integral to
mastering the data.
exploratory visualizations (144) Made when the lines between
steps P (perform test plan), A (address and refine results), and C
(communicate results) are not as clearly divided as they are in a
declarative visualization project. Often when you are exploring the
data with visualizations, you are performing the test plan directly in
visualization software such as Tableau instead of creating the chart
after the analysis has been done.
F
financial statement analysis (328) Used by investors, analysts,
auditors, and other interested stakeholders to review and evaluate a
company’s financial statements and financial performance.
flat file (41, 204) A means of storing data in one place, such as in
an Excel spreadsheet, as opposed to storing the data in multiple
tables, such as in a relational database.
foreign key (43) An attribute that exists in relational databases in
order to carry out the relationship between two tables. This does not
serve as the “unique identifier” for each record in a table. These
must be identified when mastering the data from a relational
database in order to extract the data correctly from more than one
table.
fuzzy match (124) A computer-assisted technique of finding
matches that are less than 100 percent perfect by finding
correspondencies between portions of the text of each potential
match.
fuzzy matching (229) Process that finds matches that may be less
than 100 percent matching by finding correspondences between
portions of the text or other entries.
H
heterogeneous systems approach (204) Heterogeneous
systems represent multiple installations or instances of a system. It
would be considered the opposite of a homogeneous system.
homogeneous systems approach (203) Homogeneous systems
represent one single installation or instance of a system. It would be
considered the opposite of a heterogeneous system.
horizontal analysis (331) An analysis that shows the change of a
value from one period to the next.
I
index (332) A metric that shows how much any given subsequent
year has changed relative to the base year.
interval data (143) The third most sophisticated type of data on
the scale of nominal, ordinal, interval, and ratio; a type of quantitative
data. Interval data can be counted and grouped like qualitative data,
and the differences between each data point are meaningful.
However, interval data do not have a meaningful 0. In interval data, 0
does not mean “the absence of” but is simply another number. An
example of interval data is the Fahrenheit scale of temperature
measurement.
K
key performance indicator (KPI) (269) A particular type of
performance metric that an organization deems the most important
and influential on decision making.
page 502
L
link prediction (10, 106) A data approach that attempts to predict
a relationship between two data items.
M
mastering the data (40) The second step in the IMPACT cycle; it
involves identifying and obtaining the data needed for solving the
data analysis problem, as well as cleaning and preparing the data for
analysis.
monetary unit sampling (MUS) (234) Allows auditors to evaluate
account balances. MUS is more likely to pull accounts with large
balances (higher risk and exposure) because it focuses on dollars,
not account numbers.
N
nominal data (142) The least sophisticated type of data on the
scale of nominal, ordinal, interval, and ratio; a type of qualitative
data. The only thing you can do with nominal data is count, group,
and take a proportion. Examples of nominal data are hair color,
gender, and ethnic groups.
normal distribution (143) A type of distribution in which the
median, mean, and mode are all equal, so half of all the
observations fall below the mean and the other half fall above the
mean. This phenomenon is naturally occurring in many datasets in
our world, such as SAT scores and heights and weights of newborn
babies. When datasets follow a normal distribution, they can be
standardized and compared for easier analysis.
O
ordinal data (142) The second most sophisticated type of data on
the scale of nominal, ordinal, interval, and ratio; a type of qualitative
data. Ordinal can be counted and categorized like nominal data and
the categories can also be ranked. Examples of ordinal data include
gold, silver, and bronze medals.
P
performance metric (268) Any calculation measuring how an
organization is performing, particularly when that measure is
compared to a baseline.
predictive analytics (94, 228) Procedures used to generate a
model that can be used to determine what is likely to happen in the
future. Examples include regression analysis, forecasting,
classification, and other predictive modeling.
predictor (or independent or explanatory) variable (9) A
variable that predicts or explains another variable, typically called a
predictor or independent variable.
prescriptive analytics (95, 228) Procedures that model data to
enable recommendations for what should be done in the future.
These typically include developing more advanced machine learning
and artificial intelligence models to recommend a course of action
based on a current problem.
primary key (42) An attribute that is required to exist in each table
of a relational database and serves as the “unique identifier” for each
record in a table.
production or live systems (204) Production (or live systems) are
those active systems that collect and report and are directly affected
by current transactions.
profiling (10, 100) A data approach that attempts to characterize
the “typical” behavior of an individual, group, or population by
generating summary statistics about the data (including mean,
standard deviations, etc.).
proportion (142) The primary statistic used with quantitative data.
Proportion is calculated by counting the number of items in a
particular category, then dividing that number by the total number of
observations.
Q
qualitative data (142) Categorical data. All you can do with these
data are count and group, and in some cases, you can rank the data.
Qualitative data can be further defined in two ways: nominal data
and ordinal data. There are not as many options for charting
qualitative data because they are not as sophisticated as quantitative
data.
quantitative data (143) More complex than qualitative data.
Quantitative data can be further defined in two ways: interval and
ratio. In all quantitative data, the intervals between data points are
meaningful, allowing the data to be not just counted, grouped, and
ranked, but also to have more complex operations performed on
them such as mean, median, and standard deviation.
R
ratio analysis (328) A tool that attempts to evaluate relationships
among different financial statement items to help understand a
company’s financial and operating performance.
ratio data (143) The most sophisticated type of data on the scale
of nominal, ordinal, interval, and ratio; a type of quantitative data.
They can be counted and grouped just like qualitative data, and the
differences between each data point are meaningful like with interval
data. Additionally, ratio data have a meaningful 0. In other words,
once a dataset approaches 0, 0 means “the absence of.” An
example of ratio data is currency.
regression (9, 106) A data approach that attempts to estimate or
predict, for each unit, the numerical value of some variable using
some type of statistical model.
relational database (41) A means of storing data in order to
ensure that the data are complete, not redundant, and to help
enforce business rules. Relational databases also aid in
communication and integration of business processes across an
organization.
page 503
response (or dependent) variable (9) A variable that responds to,

or is dependent on, another.
S
similarity matching (10, 106) A data approach that attempts to
identify similar individuals based on data known about them.
sparkline (333) A small visual trendline or bar chart that efficiently
summarizes numbers or statistics in a single spreadsheet cell.
standard normal distribution (144) A special case of the normal
distribution used for standardizing data. The standard normal
distribution has 0 for its mean (and thus, for its mode and median, as
well), and 1 for its standard deviation.
standardization (144) The method used for comparing two
datasets that follow the normal distribution. By using a formula, every
normal distribution can be transformed into the standard normal
distribution. If you standardize both datasets, you can place both
distributions on the same chart and more swiftly come to your
insights.
standardized metrics (338) Metrics used by data vendors to allow
easier comparison of company reported XBRL data.
structured data (100) Data that are organized and reside in a
fixed field with a record or a file. Such data are generally contained
in a relational database or spreadsheet and are readily searchable
by search algorithms.
summary statistics (97) Describe the location, spread, shape,
and dependence of a set of observations. These commonly include
the count, sum, minimum, maximum, mean or average, standard
deviation, median, quartiles,correlation covariance, and frequency
that describe a specific measurable value.
supervised approach/method (103) Approach used to learn more
about the basic relationships between independent and dependent
variables that are hypothesized to exist.
support vector machines (110) A discriminating classifier that is
defined by a separating hyperplane that works first to find the widest
margin (or biggest pipe).
systems translator software (204) Systems translator software
maps the various tables and fields from varied ERP systems into a
consistent format.
T
tax data mart (365) A subset of a company-owned data
warehouse focused on the specific needs of the tax department.
tax planning (369) Predictive analysis of potential tax liability and
the formulation of a plan to reduce the amount of taxes paid.
test data (109) A set of data used to assess the degree and
strength of a predicted relationship established by the analysis of
training data.
training data (109) Existing data that have been manually
evaluated and assigned a class, which assists in classifying the test
data.
U
unsupervised approach/method (103) Approach used for data
exploration looking for potential patterns of interest.
V
Vendor address - Warehouse address Software such as Tableau
can calculate distances, but it requires a little more work. See
http://www.vizwiz.com/2012/01/tableau-tip-calculating-distance.html
to learn how.
vertical analysis (328) An analysis that shows the proportional
value of accounts to a primary account, such as Revenue
W
what-if scenario analysis (370) Evaluation of the impact of
different tax scenarios/alternatives on various outcome measures
including the amount of taxable income or tax paid.
X
XBRL (99, 336) XBRL stands for eXtensible Business Reporting
Language and is a type of XML (extensible markup language) used
for organizing and defining financial elements.
XBRL (eXtensible Business Reporting Language) (99) A global
standard for exchanging financial reporting information that uses
XML.
XBRL taxonomy (336) Defines and describes each key data
element (like cash or accounts payable). The taxonomy also defines
the relationships between each element (like inventory is a
component of current assets and current assets is a component of
total assets).
XBRL-GL (340) Stands for XBRL-General Ledger; relates to the
ability of enterprise system to tag financial elements within the firm’s
financial reporting system.
page 504
Index
A
Access. See Microsoft Access
Accountants, skills for analytic, 12
Accounting
analytic models for, 94–97
auditing and, 5–6
Data Analytics and, 5–7
data reduction and, 98–100
decision support systems and, 112–113
regression approach, 108–109
profiling example in management, 102
summary statistics, 97–98
Accounting data, using/storing, 40–41
Account organization, XBRL taxonomy, 337
Accounts receivable
Question 1.1: How Effectively Are We Collecting Our Cash?, 392–
396
Question 1.3: What Is the Total Revenue and Balance in Accounts
Receivable, 398–399
ACL software, 230
Activity ratios, 329–330
audit data analytics and, 230
Lab 2-2: Use PivotTables to Denormalize and Analyze the Data,
67
Lab 2-6: Comprehensive Case: Dillard’s Store Data: How to
Create an Entity-Relationship Diagram, 78
Preview Data from Tables In a Query, 80
Lab 2-8: Comprehensive Case: Dillard’s Store Data: Connecting
Excel to a SQL Database, 89–90
Lab 3-3: Classification, 130
Lab 3-4: Comprehensive Case: Dillard’s Store Data: Data Abstract
(SQL) and Regression (Part 1), 135
(SQL) and Regression (Part II), 137
Lab 4-1: Use PivotCharts to Visualize Declarative Data, 166
Lab 4-3: Comprehensive Case: Dillard’s Store Data: Create
Geographic Data Visualizations in Tableau and in Power BI,
188–196
Lab 4-4: Comprehensive Case: Dillard’s Store Data: Visualizing
Regression in Tableau, 199
Lab 6-1: Evaluate the Master Data for Interesting Addresses, 250
Lab 6-4: Comprehensive Case: Dillard’s Store Data: Hypothesis
Testing (Part I), 263
Testing (Part II—Data Visualization), 265
Lab 7-2: Create a Balanced Scorecard Dashboard in Tableau, 292
Lab 7-6: Comprehensive Case: Dillard’s Store Data: Creating KPIs
in Excel (Part IV—Putting It All Together), 317–319
Lab 8-1: Create a Horizontal and Vertical Analysis Using XBRL
Data, 349
Lab 8-4: Use SQL to Query an XBRL Database, 360
Lab 9-3: Comprehensive Case 2: Dillard’s Store Data: Calculate
Sales Tax for Dillard’s States Part 2—Compare Year over Year,
385–386
Lab 9-5: Comprehensive Case 4: Dillard’s Store Data: Does a
State’s Tax Rate Affect Dillard’s Decision to Open Stores
There?, 389
LendingClub, 17–19
management accounting, 275–276
Advanced Environmental Recycling Technologies (AERT), 102
Age analysis, descriptive analytics, 229, 231–232
Aggregates/aliases, expand SELECT SQL, 459–460
Ahmed, A. S., 108n3
Alarms, continuous monitoring, 208–209
Alibaba, 3, 10, 96, 340
Alphabet, 340
Alternative stacked bar chart, 153
Amazon, 3, 10, 96, 114, 279–280, 345, 409
Amazon (AMZN), 345
Amazon Prime, 281
Amazon RDS, 41
American Institute of Certified Public Accountants (AICPA), 47, 205
Analytics mindset, 12
Analyze and Refine Results, 314–315 (lab)
Analyze the Financial Ratios, 354 (lab)
Apple, Inc. (AAPL), 29, 329, 331, 332, 344, 346, 347, 353
Applied statistics, predictive analytics and, 229, 243
Arelle.org, 356
Artificial intelligence (AI), prescriptive analytics and, 96, 113–114,
228, 229, 243
Asset turnover ratio, 330
Audience, effective communication and, 159–160
Audit Data Analytics
Address and Refine Results, 230. See also Address and Refine
Results
Benford’s law, 236–239
communicate insights, 230
descriptive analytics and, 230–235, 239–242
diagnostic analytics, 235–236
examples of, 229
identify the problem, 226
Lab 6-1: Evaluate the Master Data for Interesting Addresses, 248–
250
Lab 6-2: Perform Substantive Tests of Account Balances, 250–256
Lab 6-3: Finding Duplicate Payments, 256–257
Testing (Part I), 257–263
nature/extent/timing of, 226
perform the test plan, 226–228
predictive analytics and, 242–243
sentiment analysis, 243
track outcomes, 230
See also Auditing
Audit Data Standards (ADS), 47, 205, 210
Auditing
automated systems for, 210
clustering approach in, 105–106
Lab 1-3: Data Analytics in Auditing, 33–34
Lab 5-5: Identify Audit Data Requirements, 221–222
predictive analytics, regression and, 108
profiling in, 103
remote, 209–210
tax compliance and, 364
workflow, working papers, 209–210
See also Audit Data Analytics
Auditing standards (PCAOB), 207
Auditors
Question Set 1: Order-To-Cash (O2C), 392–400
Question Set 2: Procure-to-Pay (P2P), 400–406
Audit plan
automating, steps for, 208
characteristics of, 207
Lab 5-6: Prepare Audit Plan, 222–223
methodology/standards, 207
page 505
Aura, PwC tool, 210, 211

Automating data analytics
Automating data analytics, computer-assisted audit techniques
(CAATs), 228–230
Automation, 202
audit plan, 207–208
scripts, 228–230
Average collection period ratio, 330
Average days in inventory ratio, 330
B
Background information, select Fortune 100 companies, 350 (lab)
Balanced Scorecard
components of, 271–272
defined, 271, 277
example, 267, 272
key performance indicators, 270–274
Lab 7-2: Create a Balanced Scorecard Dashboard in Tableau,
284–292
strategy map template, 275
See also Scorecard
Balance sheet composition
sunburst diagram, visualize, 334
XBRL data, 339
Bar charts, 145, 154
Bay Area Rapid Transit (BART), 93
Benchmarks, financial statements and, 331
Benford’s law, 103–104, 115
diagnostic analytics and, 229
predicting distribution, 236–239
Berinato, Scott, 141
Big Data, 4, 20
Bjerrekaer, J. D., 39
Boeing Co., 339
Boundaries, support vector machine, 111
Box, cloud computing, 219
Box and whisker plots, 148
Box Chart, 152
Bristol-Myers Squibb (BMY), 353
Bullet graph, 270
Business, Data Analytics effects on, 4–5
Business process
defining, 392
Order-To-Cash (O2C) sales cycle, 392–400
Procure-To-Pay (P2P), 400–406
C
Calcbench data vendor, 338, 339, 342
Cash, accounts receivable and, 392–396
Cash tag, XBRL and, 337
Categorical data, 142–143
Causal modeling, 106, 115
Central tendency, describing sample by, 434–435
Certified management accountant (CMA), 328
Certified public accountant (CPA), 328
Change amount, 332
Change in value relative to base year, 332
Change percent, 332
Charting data
create good, study bad, 151–154
qualitative, 145–147
quantitative, 147–148
refining charts, 156–158
types of charts, summary of, 148
See also Data visualization
Chevron (CVX), 353
Chick-Fil-A, 434
Chief audit executive (CAE), 203
Citigroup (C), 353
Class, 106
Classification
defined, 9, 115
evaluating, 111
goals of, predictive analytics and, 109
Lab 3-3: Classification, 127–130
lease, flowchart, 113
model, trade-off, complexity v., 112
overfitting, 111–112
predictive analytics and, 96, 229, 242
steps of, 109
terminology of, 109–111
Classification analysis, setup/conduct, 119–120
Classification model, 106
Clean data, 50
Cloud folder, 219–220 (lab)
Cluster analysis
auditing and, 105–106
high volume stores, 320–322 (lab)
unsupervised approach, 103
Clustering
defined, 10, 115
diagnostic analytics and, 96, 229, 242
COLLEGESCORECARD data, 492–493
College Scorecard data, 57, 125–126 (lab)
Color
charts and, 157–158
visualizing financial data, 333–334
See also Data visualization
Column charts, 146
Columns, tables and, 42–43
Combine Visualizations into a Dashboard, 218 (lab)
Committee of Sponsoring Organization (COSO), 207
Common data model
defined, 210
Lab 5-1: Create a Common Data Model, 215–217
Lab 5-2: Create a Dashboard Based on a Common Data Model,
217–219
AICPA and, 204–206
Common size financial statement, 328–329, 332, 341, 349–352 (lab)
Common table functions, 356 (lab)
Communicate Findings/Results
charting data, 156–158
content/organization, 158–159
67
Lab 4-1: Use PivotCharts to Visualize Declarative Data, 166–168
Lab 4-2: Use Tableau to Perform Exploratory Analysis and Create
Dashboards, 176–177
in Excel (Part III), 315
revising message, 160
See also Charting data; Data visualization
LendingClub, 19
tracking outcomes and, 1, 11–12
Complexity of model, classification of v., 112
Composite primary key, 43, 53
Comprehensive Case
Lab 1-4: Dillard’s Store Data, 34–37
Lab 2-6: Dillard’s Store Data: How to Create an Entity-Relationship
Diagram, 75–78
Lab 2-7: Dillard’s Store Data: How to Preview Data from Tables In
a Query, 78–80
Lab 2-8: Dillard’s Store Data: Connecting Excel to a SQL
Database, 80–90
Lab 2-9: Dillard’s Store Data: Joining tables, 90–91
Lab 3-4: Dillard’s Store Data: Data Abstract (SQL) and Regression
(Part 1), 130–135
Lab 3-5: Dillard’s Store Data: Data Abstract (SQL) and Regression
(Part II), 135–137
Lab 4-3: Dillard’s Store Data: Create Geographic Data
Visualizations in Tableau and in Power BI, 177–196
Lab 4-4: Dillard’s Store Data: Visualizing Regression in Tableau,
196–199
Lab 6-4: Dillard’s Store Data: Hypothesis Testing (Part I), 257–263
Lab 6-5: Dillard’s Store Data: Hypothesis Testing (Part II—Data
Visualization), 263
Lab 7-3: Dillard’s Store Data: Creating KPIs in Excel (Part I), 293–
299
Lab 7-4: Dillard’s Store Data: Creating KPIs in Excel (Part II), 299–
307
page 506
Lab 7-5: Dillard’s Store Data: Creating KPIs in Excel (Part III),
307–315
Lab 9-2: Case 1: Dillard’s Store Data: Calculate Sales Tax for
Dillard’s States, 379–381
Dillard’s States Part 2—Compare Year over Year, 381–386
Dillard’s States Part 3—Calculate City Tax and Compare Tax
Owed Year over Year, 386–397
Lab 9-5: Dillard’s Store Data: Does a State’s Tax Rate Affect
Dillard’s Decision to Open Stores There?, 387–389
Computer-assisted audit techniques (CAATs), 228–230, 229, 244
Conceptual chart, 142
Conceptual data, 143
Confidence interval, 437
Confidence level, 234
Connect, PwC tool, 210, 211
ConocoPhillips (COP), 353
Content, data visualization and, 158–159
Continuous auditing, 103, 208, 211
Continuous data, 143, 161
Continuous monitoring, 208
alarms/exceptions, 208–209
defined, 211
Continuous reporting, 208, 211
Co-occurrence grouping, 10, 21
cluster analysis and, 103
defined, 115
Corptax, 362
Cost behavior, 269–270
Costco (Cost), 353
Coughlin, Tom, 103
Create Four Visualizations, 218 (lab)
Current ratio, 329
D
Daily Mail, 152
Dashboards, 101, 141
balanced scorecard dashboard, 275
digital, 270, 277
218–219
Testing (Part II—Data Visualization), 263–265
284–292
Data
Big, 4, 20
cleaning, 50
ethics breach and, 39
gather/review, 8–9
quality, 50–52
relationships, relational databases and, 42–43
storing, 48–49
validate, 49–50
variability/spread, describing, 435
Data Analysis ToolPak, Excel add-in
accessing, 440–441
tutorial, 440–441
Data Analytics
auditing and, 5–6
automating. See Automating data analytics, 206–208
business and, 4–5
defined, 4, 21
financial reporting and, 6–7
four categories of, 94–97
IMPACT cycle, 8
Lab 1-1: Data Analytics in Financial Accounting, 28–31
Lab 1-2: Data Analytics in Managerial Accounting, 31–33
Lab 2-1: Create a Request for Data Extraction, 59–60
taxes and, 7
Database maps, 209
Databases
computer languages for, 48–49
data dictionary, 44–45
ETL process, 45–49
management, software, 41
relationships, relational, 42–43
table attributes, 42–43
types of, 29
Database Schema, 41
Data dictionary, 14, 21
COLLEGESCORECARD, 492–493
defined, 53
LendingClub, 44–45, 494–498
LoadStats, 494–498
OK PCARD, 499
RejectStats, 499
Data-driven chart, 142
Data environment, modern, 202–203
Data management, taxes and, 363–364
Data marts, 363, 370
Data profiling. See Profiling data
Data quality, XBRL and, 338–339
Data reduction, 10–11, 21
defined, 98, 115
descriptive analytics and, 95–96, 98–100
Lab 3-1: Data Reduction Using Fuzzy Matching, 122–125
Data request, 46–47, 59 (lab)
Data request form, 47, 53
Data scale, charting data, 157
Datasets
ethics and, 39
See also Data dictionary; Dillard’s Stores Inc.; LendingClub;
Sláinte
Data types, 142–143
Data visualization, 12
audience/tone, effective communication, 159–160
bar charts, 154
box charts, 152
categorical data, choosing, 142–143
chart types, 142
color, charts and, 157–158
content/organization and, 158–159
create good charts, study bad, 151–154
data scale/increments and, 157
declarative, explanatory v., 144–145
designing, 275
heat map, 139
218–219
Lab 7-7: Comprehensive Case: Dillard’s Store Data: Advanced
Models in Tableau, 320–324
Lab 9-1: State Sales Taxes and Create a Data Visualization, 375–
379
normal distribution, 143–144
pie chart, stacked bar chart v., 153
purpose of, determine, 141–142
qualitative data, charts for, 145–147
quantitative data, charts for, 147–148
Question 1.2: How Can We Explore the Product Hierarchy
Through Data Visualization?, 412–422
Question 3.1: By Looking at Line Charts for 2014 and 2015, Does
the Average Percentage of Sales Returned in 2014 Seem to
Be Predictive of Returns in 2015, 427–429
Question Set 1: Descriptive and Exploratory Analysis, 410–422
rank-ordered bar chart, 153
refining charts, 156–158
relative size of accounts, 333–334
revising message, 160
sparklines/heat maps, 333
stacked bar chart, 154
tracking KPS, tax data and, 365–366
trends, sparkline/trendlines, 333
USA Drug Overdose, 140
visuals, tools for choosing, 149–151
See also Charting data
page 507
Data warehouse, 204, 211, 363, 364

Date datatypes, SQL WHERE clause, 456
Dates, data quality and, 50–51
DB Browser, SQLite, 215 (lab). See also Web browser
Debt-to-equity ratio, 330
Debt-to-income ratio, 15, 16, 19
Decision boundaries, classification, 109, 110, 115
Decision support systems, prescriptive analytics and, 96, 112–113,
115
Decision trees, classification, 109, 110, 115
Declarative chart, 142
Declarative visualization, 144–145, 161
Deductions, tax planning what-if scenarios and, 369–370
Delivery process, 396–398
Dependent variables, 9, 21
Descriptive analytics
age analysis, 231–232
defined, 94, 115, 228, 244
example, 229
financial, 328
sampling, 233–235
sorting, 232–233
summary of approaches, 95–96
summary statistics, 97–98, 233
Descriptive attributes, 43, 53
Diagnostic analytics
Benford’s law and, 236–239
cluster analysis, 103–106
defined, 94, 100, 115, 228, 244
drill down, 239
exact and fuzzy matching, 239–241
example, 229
financial, 331
profiling data and, 100–103
Question 2.1: Is the Percentage of Sales Returned Significantly
Higher in January After the Holiday Season?, 422–426
Question 2.2: Is the Percentage of Sales Returned Significantly
Different in Arkansas Than the Rest of the Country?, 426–427
sequence check, 241
stratification/clustering, 242
summary of, 95–96
z-score, 235–236
Digital dashboard, 270, 277
Dillard’s Stores Inc.
ER Diagram, 491
estimating sales returns, question sets for, 410–432. See also
Estimating sales returns, question sets for
Lab 1-4: Comprehensive Case: Dillard’s Store Data, 35
Lab 1-4: Dillard’s Store Data, 34–37
Create an Entity-Relationship Diagram, 75–78
Lab 2-7: Dillard’s Store Data: How to Preview Data from Tables In
a Query, 78–80
Lab 2-8: Dillard’s Store Data: Connecting Excel to a SQL
Database, 80–90
in Excel (Part I), 293–299
in Excel (Part II), 299–307
in Excel (Part III), 307–315
Sales Tax for Dillard’s States, 379–381
381–386
Sales Tax for Dillard’s States Part 3—Calculate City Tax and
Compare Tax Owed Year over Year, 386–397
There?, 387–389
Discrete data, 143, 161
Distribution
predicting, Benford’s law and, 236–239
probability, 435–436
Documents
create using OneDrive (lab), 27–28
See also Electronic working papers
Drill-down, diagnostic analytics and, 229, 239
Dropbox, 219
DuPont ratio, 330, 340, 341
E
eBay, 3
EDGAR database (SEC), 29
Effective tax rate (ETR), 365–366
Electronic working papers, 209–210
Lab 5-3: Set up a Cloud Folder, 219–220
platforms for, 210
ELT process, loading data, 52
Encoding, data quality and, 51
English dictionary (H4N-INF), 335
Enterprise data
common data model, 204–206
systems approaches to, 203–204
Enterprise Risk Management (COSO), 207
Entity-relationship diagram (ERD), 75–78 (lab)
Equifax, 21
Equity multiplier ratio, 330
ER Diagram, Dillard’s Stores Inc., 491
Errors, data quality and, 52
Estimated misstatement, 234
Estimating sales returns, question sets for
Q. 1.1: Which Attributes Could Help Predict Percentage of
Returned Sales?, 410–412
Q. 1.2: How Can We Explore the Product Hierarchy Through Data
Visualization?, 412–422
Q. 2.1: Is the Percentage of Sales Returned Significantly Higher in
January After the Holiday Season?, 422–426
Q. 2.2: Is the Percentage of Sales Returned Significantly Different
in Arkansas Than the Rest of the Country?, 426–427
Q. 3.1: By Looking at Line Charts for 2014 and 2015, Does the
Average Percentage of Sales Returned in 2014 Seem to Be
Predictive of Returns in 2015, 427–429
Q. 3.2: Using Regression, What Can We Predict for Returns as a
Percentage of Sales Based on Historical Transactions?, 429–
432
Ethics, breach of, datasets and, 39
ETL process
automating data analytics and, 206–207
defined, 53
extract, 46–47
61
Lab 2-5: College Scorecard Extraction and Data Preparation, 73–
75
Regression in Tableau, 197–198
page 508
Lab 5-1: Create a Common Data Model, 216–217

load data, 52
transform, 49–50
European Union, 364
Evaluate data, 221–222 (lab)
Exact matching, diagnostic analytics and, 229, 239
Excel. See Microsoft Excel entries
Exception report, 208–209
Exceptions, generate/follow up, 101
Experian, 21
Explanatory variables, 9, 21
Exploratory analysis
Question 1.2: How Can We Explore the Product Hierarchy
Through Data Visualization?, 412–422
Exploratory chart, 142
Exploratory visualization, 144–145, 161
eXtensible Business Reporting Language (XBRL), 29, 99–100, 116
cash tag, 337
data quality, 338–339
defined, uses of, 336–338, 341
financial ratios on Google Sheets, 353
financial statement analytics using, 340
instance document, 336–338
Data, 346–349
Lab 8-2: Create Dynamic Common Size Financial Statements,
349–352
Lab 8-3: Analyze Financial Statement Ratios, 352–353
Lab 8-4: Use SQL to Query an XBRL Database, 355–360
standardized metrics, 338
standardized tags, 339
taxonomy, 336
XBRL-Global Ledger, real-time financial reporting, 340
Extensible markup language (XML), 336
External auditing, data reduction and, 98–99
Extracting data. See ETL process
ExxonMobile (XOM), 353
F
Facebook, 7, 11, 21
Facebook (FB), 21, 353
False positive, 209
FASB taxonomy, 338
Favorable variances, 269
Fawcett, T., 9n
Filled geographic maps, 148
Financial accounting, data analytics in, 28–31 (lab)
Financial Accounting Standards Board (FASB), 29
Financial analysis, text mining/sentiment analysis, 334–335
Financial dictionary (Fin-Neg), 335
Financial reporting
real-time, XBRL/XBRL-GL, 340
Financial Statement Analysis
common size financial statement, 328–329, 331–332
defined, 341
descriptive, 328
diagnostic financial analytics, 331
EXBRL examples, 340
index showing change in value relative to base year, 332
Data, 346–349
Lab 8-2: Create Dynamic Common Size Financial Statements,
349–352
predictive financial analytics, 331–332
ratios, comparison among three companies, 331
ratio types, 329–330
vertical/horizontal analysis, 328–329, 331–332
Financing ratio, 330
Findings. See Address and Refine Results; Communicate
Findings/Results
FinDynamics, 346, 347
Flat file, 41, 53, 204, 211
Forbes.com, 361
Forbes Insights/KPMG report, 5–6
Foreign keys, 43, 53
Fortune 100, 346, 350
Frequency distribution, 375–377 (lab)
FROM, SQL clause, 453–454, 463–464
Fujitsu, 338
Fuzzy Lookup add-in, 121 (lab), 239–241
Fuzzy match, 98, 99
defined, 244
diagnostic analytics and, 229, 239–241
250
G
Gartner Magic Quadrant for Business Intelligence and Analytics
Platform, 149
Generalized audit software (GAS), 230
General Motors (GM), 340
Generate a Request for Data, 285 (lab)
Good classification, 109
Google, 7, 114
Google Account, 352–354 (lab)
Google Drive, 219
Google Sheets
Data, 346
GROUP BY, SQL clause, 460–462
H
Halo, PwC tool, 201, 210, 211
Harriott, J. S., 8n, 40n
Harvard Business Review, 141
HAVING, SQL clause, 462–463
Heat map, data visualization, 139, 146, 333
Heterogeneous systems approach, data structure, 204, 211
Hewlett-Packard Co. (HP), 225
Histogram, data visualization, 375–377 (lab)
Homogeneous systems approach, data structure, 203–204, 211
Horizontal financial statement analysis, 328–329, 331–332, 341,
346–349 (lab)
Hyperion, 362
Hypothesis testing
Q. 2.1: Is the Percentage of Sales Returned Significantly Higher in
Q. 2.2: Is the Percentage of Sales Returned Significantly Different
in Arkansas Than the Rest of the Country?, 426–427
statistics and, 436–437
I
IBM, 338
IBM DB2, 41
IDEA software, 230
Benford’s law, 238
fuzzy address match, 241
250
monetary unit sample (MUS), 235
sample size, 234
sorting, 233
summary statistics, 233
page 509
Identify Questions
Lab 1-4: Dillard’s Store Data (comprehensive case), 35–36
61
Lab 2-3: Resolve Common Data Problems in Excel and Access,
68
Lab 2-5: College Scorecard Extraction and Data Preparation, 74
Preview Data from Tables In a Query, 78–79
Lab 2-9: Comprehensive Case: Dillard’s Store Data: Joining
tables, 90
Lab 3-2: Regression in Excel, 125
Lab 3-3: Classification, 127
Dashboards, 169
178
Regression in Tableau, 197
Lab 6-2: Perform Substantive Tests of Account Balances, 251
Lab 6-3: Finding Duplicate Payments, 256
Lab 7-1: Evaluate Management Requirement and Identify Useful
KPIs from a List, 284
284–285
in Excel (Part I), 295
in Excel (Part II), 300
in Excel (Part IV—Putting It All Together), 316
Data, 346
Lab 9-1: State Sales Taxes and Create a Data Visualization, 375
Sales Tax for Dillard’s States, 379
382
Compare Tax Owed Year over Year, 386
There?, 388
Identify the Location of the Data, 222 (lab)
Identify the Problem
auditing data analytics, 226
Lab 3-1: Data Reduction Using Fuzzy Matching, 121
IMPACT cycle
data, gather/review, 8–9
data visualization, 141
outcomes, track, 11–12
questions, identify, 8
results, address/refine, 11
test plan, perform, 9–11, 94–97
IMPACT model
audit data analytics and, 226–230
hands-on, Lendingclub example. See LendingClub entries
KPIs for decision-making and, 272, 366
mastering the data, 40
tax data and, 363
Import and Validate the Data
Lab 5-1: Create a Common Data Model, 215
218
Income tax liability, 365
Increments, charting data, 157
Independent variables, 9, 21
Index, 332, 341
INNER join, 480–481
Inner join, 239
INNER JOIN ON, SQL clause, 464–465
Input Ticker Symbols, 354 (lab)
Instagram, 7, 21
Instance document, XBRL and, 336
Internal auditing
importance of, 203
profiling and, 103
International characters, data quality and, 51
Interval data, 143, 161
Inventory turnover ratio, 330
Invoices, paying, Question Set 2: Procure-To-Pay (P2P), 400–406
IRS and tax analytics, 362
Isson, J. P., 8n, 40n
iXBRLAnalyist script, 346 (lab)
J
James, LeBron, 361
JD Edwards, 203, 213
Johnson & Johnson (JNJ), 353
JPMorgan Chase (JPM), 353
K
Kaplan, Robert S., 271
Kenya Red Cross, 267
Key performance indicators (KPIs), 141, 268–269
balanced scorecard and, 270–274
defined, 277
financial performance/operational, 273
KPIs from a List, 282–283
monitor/track, tax data analytics, 365–366
page 510
Kirkegaard, E., 39
KPIs. See Key performance indicators (KPIs)
L
Labs
1-0: How to Complete Labs in This Text, 27–28
1-1: Data Analytics in Financial Accounting, 28–31
1-1: Data Analytics in Financial Accounting, 28–31
1-2: Data Analytics in Managerial Accounting, 31–33
1-3: Data Analytics in Auditing, 33–34
1-4: Dillard’s Store Data (comprehensive case), 35–37
2-1: Create a Request for Data Extraction, 58–60
2-2: Use PivotTables to Denormalize and Analyze the Data, 60–67
2-3: Resolve Common Data Problems in Excel and Access, 67–71
2-4: Generate Summary Statistics in Excel, 71–73
2-5: College Scorecard Extraction and Data Preparation, 73–75
2-6: Comprehensive Case: Dillard’s Store Data: How to Create an
Entity-Relationship Diagram, 75–78
2-8: Comprehensive Case: Dillard’s Store Data: Connecting Excel
to a SQL Database, 80–90
2-9: Comprehensive Case: Dillard’s Store Data: Joining tables,
90–91
3-1: Data Reduction Using Fuzzy Matching, 121–125
3-2: Regression in Excel, 125–126
3-3: Classification, 127–130
3-4: Comprehensive Case: Dillard’s Store Data: Data Abstract
(SQL) and Regression (Part 1), 130–135
3-5: Comprehensive Case: Dillard’s Store Data: Data Abstract
(SQL) and Regression (Part II), 135–137
4-1: Use PivotCharts to Visualize Declarative Data, 166–168
4-2: Use Tableau to Perform Exploratory Analysis and Create
4-3: Comprehensive Case: Dillard’s Store Data: Create
177–196
4-4: Comprehensive Case: Dillard’s Store Data: Visualizing
5-1: Create a Common Data Model, 215–217
5-2: Create a Dashboard Based on a Common Data Model, 217–
219
5-3: Set up a Cloud Folder, 219–220
5-4: Review Changes to Working Papers, 220–221
5-5: Identify Audit Data Requirements, 221–222
5-6: Prepare Audit Plan, 222–223
6-1: Evaluate the Master Data for Interesting Addresses, 248–250
6-2: Perform Substantive Tests of Account Balances, 250–256
6-3: Finding Duplicate Payments, 256–257
6-5: Comprehensive Case: Dillard’s Store Data: Hypothesis
7-1: Evaluate Management Requirement and Identify Useful KPIs
from a List, 282–283
7-2: Create a Balanced Scorecard Dashboard in Tableau, 284–
292
7-3: Comprehensive Case: Dillard’s Store Data: Creating KPIs in
Excel (Part I), 293–299
Excel (Part II), 299–307
Excel (Part III), 307–315
Excel (Part IV—Putting It All Together), 315–319
7-7: Comprehensive Case: Dillard’s Store Data: Advanced Models
in Tableau, 320–324
8-1: Create a Horizontal and Vertical Analysis Using XBRL Data,
346–349
8-2: Create Dynamic Common Size Financial Statements, 349–
352
8-3: Analyze Financial Statement Ratios, 352–354
8-4: Use SQL to Query an XBRL Database, 355–360
9-1: State Sales Taxes and Create a Data Visualization, 375–379
9-2: Comprehensive Case 1: Dillard’s Store Data: Calculate Sales
Tax for Dillard’s States, 379–381
Tax for Dillard’s States Part 2—Compare Year over Year, 381–
386
Tax for Dillard’s States Part 3—Calculate City Tax and
9-5: Comprehensive Case 4: Dillard’s Store Data: Does a State’s
Tax Rate Affect Dillard’s Decision to Open Stores There?, 387–
389
Question Set 2: Procure-To-Pay (P2P), 392–400
Question Set 3: Predictive Analytics, 427–432
Question set 1 Order-To-Cash (O2C), 392–400
Languages
data quality and, 51
SQL, 48
text mining, sentiment analysis and, 335
XML, EBRL and, 336
Left join, 239
LEFT join, SQL clause, 482
Legislation, tax planning what-if scenarios and, 369–370
LendingClub
Credit Risk Scores, customer, 17, 19
credit score data, 116
data, 494–498
data, gather/review, 13–15
data dictionary, 44–45
datasets/data dictionary, website, 14
debt-to-income ratio, declined loan applications by, 16, 19
debt-to-income ratio, loans declined by, 19
declined loan data, 14
employment length, declined loan applications by, 16
insights, communicate, 19
67–71
Lab 2-4: Generate Summary Statistics in Excel, 71–73
loan rejection, predicting, 112
loans issued, 13
outcomes, tracking, 19
pivot tables, 16, 18, 19
questions, identify, 13
regression, predictive analytics and, 108–109
RejectStatsA dataset, 15
reported loan purpose, 13
results, address/refine, 17–19
test plan, 15–19
Linear classifiers, 110
Line charts, 148, 427–429
page 511
Link prediction
defined, 10, 21, 115
example, 11
predictive analytics and, 96, 106
Liquidity ratio, 329
Load data. See ETL process, 52
Look for Audit Procedures that Evaluate Data, 221–222 (lab)
Loughran, Tim, 335, 345
Lyft, 334
M
Machine learning, prescriptive analytics and, 96, 113–114, 228, 243
Magic quadrant, 149
Management accounting
balanced scorecard/KPIs, 270
cost behavior, 269–270
data visualization, 275
drill-down/refine results, 275–276
evaluate data quality, 274–275
identify questions, 268–270
key performance indicators (KPIs)/variance analysis and, 268–269
KPIs for, 273
KPIs from a List, 282–283
284–292
predictive analytics, regression and, 108
profiling in, 102
relevant costs, 268
Management discussion, word frequency, 334
Managerial accounting, data analytics in, 32–33 (lab)
Managers
Manipulating data, 12
Mapping data, 215–217 (lab)
Marr, Bernard, 282
Mastering the Data and Performing the Analysis
Master the Data and Prepare for Analysis
Lab 8-3: Analyze Financial Statement Ratios, 353
Master the Data and Prepare for Analysis Using Excel
Master the Data
6-5: Comprehensive Case: Dillard’s Store Data: Hypothesis
defining, 40, 53
extract, 45–49
Lab 1-4: Dillard’s Store Data (comprehensive case), 36–37
Lab 2-1: Create a Request for Data Extraction
61–66
68–71
Lab 2-5: College Scorecard Extraction and Data Preparation, 74–
75
tables, 90–91
178–185
THERE APPEARS TO BE A STAGE MISSING
KPIs from a List, 284
Data, 346–348
382–383
There?, 388
load the data, 52
Question Set 1: Descriptive and Exploratory Analysis, 419
transform, 49–50
page 512
Master the Employee and Vendor Data

249
McDonald, Bill, 335, 345
McKinsey Global Institute, 5
Measures, data quality and, 51
Merck (MRK), 353
Merge data, 63–65 (lab)
Metadata, Lab 1-4: Dillard’s Store Data (comprehensive case), 37
Microsoft Access
defined, 41
60
67–71
Lab 6-2: Perform Substantive Tests of Account Balances, 251
Microsoft BI suite, 149
Microsoft Corp. (MSFT), 114, 329, 331, 332, 333, 353
Microsoft Excel
add-ins, 308–309
age analysis, 231
Benford’s law, predicting distribution, 237–238
Benford’s law percentages, 238–239
formatting, income statement using SUM(), 442–450
Fuzzy Lookup add-in, 239–241
Get and Transform tool, 470–472
Lab 6-1: Evaluate the Master Data for Interesting Addresses, 248,
249
379
382–386
There?, 387–389
monetary unit sample (MUS), 234
PivotTables, 448–449. See also Microsoft Excel PivotTable
sample size, 234
sorting, 232
spreadsheets, tax data and, 363
Tableau and, 476–478
tutorial, 442–452
VLookup function, 450–452
z-score calculation, 236
Data Analysis Toolpak tutorial, 440–441
Microsoft Excel Data Analysis Toolpak
accessing, 440–441
tutorial, 440–441
Microsoft Excel database
Lab 1-3: Data Analytics in Auditing, 33
60, 61–66
67–71
Lab 2-4: Generate Summary Statistics in Excel, 71–73
Lab 3-1: Data Reduction Using Fuzzy Matching, 121
Lab 3-2: Regression in Excel, 125–126
storing data, 48–49
summary statistics in, 72–73
VLookup function, 48–49
Microsoft Excel Internal Data Model, 62–63 (lab)
Microsoft Excel PivotTable
60–67
383–384
LendingClub, 15, 16, 18, 19
Question 3.2: Using Regression, What Can We Predict for Returns
as a Percentage of Sales Based on Historical Transactions?,
429–432
tools for, 448–449
Microsoft Excel PowerPivot add-in
Microsoft Excel Query Editor, 63–65 (lab)
Microsoft OneDrive
Lab 1-0: How to Complete Labs in This Text, 27–28
Lab 5-6: Prepare Audit Plan, 223
Microsoft SQL Server Management Studio
connect, Excel’s Get and Transform Tool, 470–472
defined, 41
page 513

tables, 90–91
(SQL) and Regression (Part II) 135–137
177
382–386
Question 1.1: Which Attributes Could Help Predict Percentage of
Middle value, describing sample by, 434–435
Monetary unit sample (MUS), 234–235, 244
Mystery ratios, 350 (lab)
N
NASDAQ, 327
New York Stock Exchange, 327
Nike (NKE), 346
Nominal data, 142, 161
Normal distribution, 143–144, 161, 435–436
Norton, David P., 271
Number datatypes, SQL WHERE clause, 456
Numbers, data quality and, 51
O
Object-relational mapping (ORM), 215
Obtain data
data request, 46–47
personally, 48
Office 363, 210
Office.com, 27–28, 27–28 (lab)
Office of National Statistics, 152
OkCupid, 39
OK PCARD data, 498, 499
Online sales, analyzing, 409
Open Science Framework, 39
Oracle, 41, 340
ORDER BY, SQL clause, 458–459
Order-To-Cash (O2C), 392–400. See also Question Set 1: Order-To-
Cash (O2C)
Ordinal data, 142, 161
Organization, data visualization and, 158–159
Outcomes. See Track Outcomes
Outer join, 239
Overfitting, classification, 111–112
Overlap method, text mining and, 335
P
Parameters, statistics v., 434
Payments, Procure-to-Pay (P2P), 400–406 (lab)
185–188
285–292
Data, 348–349
379
Performance metrics, 268–269
balanced scorecard/KPIs and, 270–274
defined, 277
Perform the Analysis
66–67
tables, 90–91
(SQL) and Regression (Part 1), 131–135
250
383–384
There?, 388–389
Perform the Test Plan
page 514
Pie charts, 145, 146, 153

PivotTable. See Excel PivotTable
Poisson distribution, probability, 436
Population, sample v., 434
PostGreSQL, 41
Post-pruning, decision tree, 109
Power BI
analytics tool, 391
ask a question, 490
177–196
load data, 485
mode, choose, 485
opening, startup screen, 483–484
tutorial, 483–490
visualizations/fields/values, 486–489
PowerPivot. See Excel PowerPivot add-in
Power Query
editing data in, 473–474
load data into Excel, 474
return to window after closing, 475
SQL server, Excel Get and Transform, 470–472
tutorial, 470–475
worksheet failure, workaround, 474
Power Query Editor, 426–427
Predictive analytics
applied statistics and, 229, 243
approaches to, 106
auditing and, 108
classification and, 109–111, 242
defined, 94, 115, 228, 244
examples, 229
financial, 331–332
managerial accounting, regression and, 108
overfitting data, 111–112
probability and, 229, 242–243
Question 1.1: Which Attributes Could Help Predict Percentage of
Question 3.1: By Looking at Line Charts for 2014 and 2015, Does
the Average Percentage of Sales Returned in 2014 Seem to
Be Predictive of Returns in 2015, 427–429
429–432
regression and, 107–108, 242
sentiment analysis, 243
summary of, 95, 96
Predictor variables, 9, 21
Prepare for Analysis, 121–122 (lab)
Pre-pruning, decision tree, 109
Prescriptive analytics
decision support systems, 112–113
defined, 95, 115, 228, 244
examples, 229
machine learning/artificial intelligence, 113–114, 228, 229, 243
summary of, 95, 96
uses of, 112
Primary keys, 42–43, 53
Probability
auditing and, 242–243
predictive analytics and, 229
Probability distributions
normal, 435–436
uniform/Poisson, 436
Procure-To-Pay (P2P), 400–406. See also Question Set 2: Procure-
To-Pay (P2P)
Production or live systems, 204, 211
Profiling, 21
defined, 10
IRS and, 362
Profiling data, 21
defined, 10, 115
internal audit, 103
steps of, 101
structured data and, 100
Profitability ratio, 330
Profit margin on sales ratio, 330
Proportion, quantitative data and, 142, 161
Provost, F., 9n
Pruning, decision tree, 109, 110
Public Company Accounting Oversight Board (PCAOB), 207
Purchasing cycle processes, 400–406. See also Question Set 2:
Procure-To-Pay (P2P)
P-value
There?, 389
statistical testing, 437
PwC, 4–5, 201, 210, 211, 280
Python, 70 (lab), 206
Q
Qualified research expenditures (QREs), 369
Qualitative chart, 142
Qualitative data
charts for, 145–147
defined, 161
Qualtrics, 434
Quantitative chart, 142
Quantitative data
charts for, 147–148
defined, 161
normal distribution, 143–144
Query, preview data from tables in, 78–80 (lab)
Questions, identify problems. See Identify Questions
Question Set 1: Descriptive and Exploratory Analysis
1.1: Which Attributes Could Help Predict Percentage of Returned
Sales?, 410–412
1.2: How Can We Explore the Product Hierarchy Through Data
Visualization?, 412–422
Question Set 1: Order-To-Cash (O2C)
1.1: How Effectively Are We Collecting Our Cash?, 392–396
1.2: Is the Delivery Process Following the Expected Procedure?,
396–398
1.3: What Is the Total Revenue and Balance in Accounts
Receivable, 398–399
1.4: What Else Can You Determine about the O2C Process?, 400
processes, 392
Question Set 2: Diagnostic Analytics—Hypothesis Testing
2.1: Is the Percentage of Sales Returned Significantly Higher in
2.2: Is the Percentage of Sales Returned Significantly Different in
Arkansas Than the Rest of the Country?, 426–427
Question Set 2: Procure-To-Pay (P2P)
2.1: How Long Are We Taking to Pay Our Invoices?, 400–404
2.2: Are There Any Erroneous Payments?, 404–405
2.3: Are We Missing Out on Discounts by Paying Late?, 405–406
2.4: What Else Can You Determine about the P2P Process?, 406
purchasing cycle processes, 400
Question Set 3: Predictive Analytics
3.1: By Looking at Line Charts for 2014 and 2015, Does the
Average Percentage of Sales Returned in 2014 Seem to Be
Predictive of Returns in 2015, 427–429
3.2: Using Regression, What Can We Predict for Returns as a
Percentage of Sales Based on Historical Transactions?, 429–
432
Quick (acid test) ratio, 329
R
R, 206
R. R. Donnelley, 338
Rankandfiled.com, 334
Rank-ordered bar chart, 153
Ratio analysis, 328, 341, 352–354 (lab)
Ratio data, 143, 161–162
Ratio types, 329–330
page 515
R&D Tax Credit, 362, 369–370

Real-time financial reporting, XBRL/EBRL-GL and, 340
Red Cross, 267
Refine the Query, 381 (lab)
Regression, 21
accounting/auditing, predictive analytics and, 108–109
defined, 9, 115
predictive analytics and, 96, 106–108, 229, 242
429–432
statistical output from, interpreting, 439
Regression analysis
cost behavior, 270
mixed costs, 270
RejectStats, data, 498, 499
Relational database
defined, 42, 53
65–66
Relational Database Management Systems (RDBMS), 41
Relationships, relational databases and, 42–43
Relevant costs, 268
Remote audit work, 209–210
Research, transparency and, 39
Resource planning systems (RPS), 203
Response variables, 9, 21
Results, communicate. See Address and Refine Results;
Communicate Findings/Results
Return on assets ratio, 330
Return on equity ratio, 330
Return on equity ratio (ROE), 330
Returns, estimating sales, 410–432
Reviewable turnover ratio, 330
Review Document Revision History, 221 (lab)
Revising message, 160
Right join, 239
RIGHT join, SQL clause, 482
Risk, accounts receivable and, 392–396
Risk scores, 17, 19
Robotics process automation, 202
R-Square, interpret, 389 (lab)
S
Sales, analyzing, 409
Sales returns, estimating. See Estimating sales returns, question
sets for
Sales returns, predicting percentage, 410–412
Sales tax liability, evaluate, 365
379
381–386
There?, 387–389
Sample
describing, 434–435
population v., 434
Sampling, descriptive analytics and, 229, 233–235
Samsung, 21
SAP, 340
Scale, charting data, 157
Scatter plots, 148, 269–270
Scorecard, 267. See also Balanced Scorecard
Screen capture tool
Lab: 1-0: How to Complete Labs in This Text, 27–28
Lab 1-1: Data Analytics in Financial Accounting, 28
Lab 1-4: Dillard’s Store Data (comprehensive case), 35
60
68
Lab 2-4: Generate Summary Statistics in Excel, 71
Dashboards, 169
Lab 5-1: Create a Common Data Model, 215, 216
218
Lab 5-2:: Create a Dashboard Based on a Common Data Model,
218
Lab 5-3: Set up a Cloud Folder, 220
Lab 5-4: Review Changes to Working Papers, 221
Lab 5-5: Identify Audit Data Requirements, 222
250
289, 290, 291, 292
in Excel (Part IV—Putting It All Together), 317, 318, 319
Models in Tableau, 321, 323, 324
Data, 349
Lab 8-2: Create Dynamic Common Size Financial Statements, 351
Lab 8-3: Analyze Financial Statement Ratios, 354
Lab 8-4: Use SQL to Query an XBRL Database, 358, 360
Lab 9-1: State Sales Taxes and Create a Data Visualization, 377,
379
Scripting language, 206
Scrubbing data, 12
Security and Exchange Commission (SEC), 29, 99
SELECT, SQL clause, 453, 459–460
SELECT FROM practice, SQL clause, 454–455
SELECT FROM WHERE practice, SQL clause, 456–457
Sentiment analysis, 243
predictive analytics and, 229, 243
text mining and, 334–335
Sequence check, 229, 241
Shared folder, create, 219–220 (lab)
Similarity matching, 10, 21
defined, 115
predictive analysis, 106
Simsion, G. C., 42n
Singleton, T., 46n
Sláinte Sales Subset, 160
Snapchat, 21
page 516
Snow, John, 139

Software
auditing and, 230
See also IDEA software; Microsoft Access; Microsoft Excel
entries; Tableau
Software needs
database, 41
exact and fuzzy matching, 239–241
Lab 1-0: How to Complete Labs in This Text, 27
Lab 1-2: Data Analytics in Managerial Accounting, 31
60
68
Lab 2-4: Generate Summary Statistics in Excel, 71
Preview Data from Tables In a Query, 78
tables, 90
Dashboards, 169
177
217
Lab 5-5: Identify Audit Data Requirements, 221
Models in Tableau, 320
Data, 346
382
There?, 387
Lab 7-2 Create a Balanced Scorecard Dashboard in Tableau, 284
sampling, 234–235
sorting, 232–233
storing data, 48–49
visualization tools, 149
Software translators, 204
Solvency ratio, 330
Sorting, descriptive analytics, 229, 232–233
Sparkline, visualizing trends, 333, 341
SQL clauses
FROM, 453–454
aggregates/aliases, expand SELECT, 459–460
example queries, ORDER BY, 458–459
FROM, select data from more than one table, 463–464
GROUP BY, 460–462
HAVING, 462–463
INNER join, 480–481
INNER JOIN ON practice, 464–465
LEFT join, 482
ORDER BY, 458
parenthesis, joining tables, 465
RIGHT join, 482
SELECT, 453
SELECT FROM practice, 454–455
SELECT FROM WHERE, 456–457
SELECT FROM WHERE practice, 457
tutorials, 453–465, 480–482
WHERE, 455–456
SQLite
defined, 41
download/install, 466–467
execute/navigate in, 467–469
60
tutorial, 466–469
SQL queries
65–66
SQL (Structured Query Language), 48. See also Microsoft SQL
Server Management Studio
Stacked bar chart, data visualization, 145, 147, 153, 154
Stacked Bar Chart of Monthly Store Performance, 322–323 (lab)
Standardization, 144, 162
Standardized metrics, 338–339, 341
Standardized tags, XBRL and, 339
Standardizing distributions, Z-scores, 144
Standard normal distribution, 144, 162
Statistical testing
confidence interval, 437
p-value, 437
Statistics
describing sample, 434–435
hypothesis testing and, 436–437
output from sample t-test difference of means of two groups,
interpreting, 438
parameters v., 434
population, sample v., 434
probability distribution and, 435–436
regression, interpreting statistical output from, 439
statistical testing, 437
tutorial, 434–439
StockSnips (app), 327
Storing data, 48–49
Strategy Management Group Company, 272
Stratification, diagnostic analytics, 229, 242
Structured data, 98, 101, 115
Structured Query Language (SQL), 48. See also Microsoft SQL
Server Management Studio; SQL entries
Summary statistics
defined, 115
descriptive analytics and, 95, 97–98
page 517

view in excel, 72
Sunburst diagram, balance sheet composition, 334
Supervised approach, predictive analysis, 106, 115
Support vector machine, classification, 110, 111, 115
SurveyMonkey, 434
Sweet spot, 112
Symbol maps, 146
Systems translator software, 204, 211
T
Table attributes, databases and, 42–43
Tableau software, 149–151
accessing, connect to Excel, 476–478
analytics tool, 391
data storage, 48–49
data visualizations, 410–422
joins in, 480–482
177–196
217–219
284–292
379
tutorial, 476–479
Tableau Workbook
Table Import Wizard, 312
Tables
tables, 90
Takeda, C., 108n3
Target, 106
Target (TGT), 353
Tax analytics
compliance and liability, 364
data for planning and, 367–370. See also Tax planning, data
analytics
data management and, 363–364
income tax liability, 365
IRS and, 362
379
381–386
There?, 387–389
sales tax liability, 365
uses for, 362–363
visualizations, monitor/track KPIs, 365–366
Tax cost, 365–366
Tax credits, tax planning what-if scenarios and, 369–370
Tax Cuts and Jobs Act Reform of 2018, 364, 370
Tax data mart, 363, 370
Tax efficiency/effectiveness, 366
Taxes, Data Analytics and, 7
Tax planning
defined/uses for, 367, 370
legislation, deductions, credits, What-if scenarios for, 369–370
what-if scenario analysis, 368–369
Tax risk, 366
Tax sustainability, 366
TeamMate, 210
TeamMate Analytics, 230, 243
Teradata, 41
Tesla, 280, 282
Test data, classification, 109, 116
Testing data, classification, 112
Test plan
LendingClub, 15–19
performing, 9–11, 94–97, 226–228
Text datatypes, SQL WHERE clause, 455–456
Text editor, 74 (lab)
Text mining, sentiment analysis and, 334–335
Thomas, S., 108n3
Times interest earned ratio, 330
Tolerable misstatement, 234
Tone, effective communication and, 159–160
Total revenue, 398–399
Track Outcomes
LendingClub, 19
tracking insights and, 11–12
Trade-off, 112
Training data, classification, 109, 112, 116
Transform data
429–432
Question Set 1: Descriptive and Exploratory Analysis, 419
See also ETL process, 49–50
TransUnion, 21
Tree maps, 146, 332–324 (lab)
Trendlines, visualizing, 333
T-test, interpreting output from sample, 438
Turnover ratio, 330
Twitter, 7
Typical value, describing sample by, 434–435
U
Uber, 334
UML diagram, 169
Underfitting data, 111
Unfavorable variances, 269
UNICODE, 51
Unified Modeling Language (UML), 41
Uniform distribution, probability, 436
Unique identifier, 42
Unsupervised approach, clustering, 103, 116
Upload files
Lab 5-4: Review Changes to Working Papers, 220–221
U.S. GAAP Financial Reporting Taxonomy, 336–337
U.S. Supreme Court, 365
V
Validate data, 49–50
Value, describing sample by middle or typical, 434–435
Variability of data, describing, 435
Variables, types of, 9
page 518
Variance analysis, data profiling, 102

Variance analysis, KPIs and, 268–269
Vertical analysis, 341
Vertical financial statement analysis, 328–329
Vertical financial statement analysis, XBRL Data, 346–349 (lab)
Visualizing data. See Data visualization
VLookup function, Excel, 48–49, 450–452
W
Walmart (WMT), 103–106, 116, 274, 345, 353
Washington.cbslocal.com, 362
Wayfair, 281
Wayfair decision, SCOTUS, 365
Web browser
Lab 5-5: Identify Audit Data Requirements, 221–222
See also DB Browser, SQLite
Weka software, 33 (lab)
Wells-Fargo (WFC), 353
What-if analysis
prescriptive analytics and, 229
scenario, 368–370
WHERE, SQL clause
date datatypes, 456
number datatypes, 456
text datatypes, 455–456
Whole Foods, 345
Wikipedia, 130, 135, 196
Witt, G. C., 42n
Word Cloud, 146
Word frequency, text mining and, 334–335
Word processor
Lab 1-2: Data Analytics in Managerial Accounting, 31
Lab 2-1: Create a Request for Data Extraction, 58
Workflow, audit, 209–210
Working capital ratio, 329
Working papers
audit workflow and, 209
Lab 5-4: Review Changes to Working Papers, 220–221
remote audit work, 209–210
See also Electronic working papers
Workpapers. See Working papers
Write a SQL Query to Transform Your Data, 216–217 (lab)
Write-off classification, 109
X
XBRL. See eXtensible Business Reporting Language (XBRL), 29
XBRLAnalyst, 339, 342
XBRL-GL (global ledger), 340, 341
XBRL instance document, 336–338
XBRL taxonomy, 336, 341
XBRL-US Center for Data Quality, 338
Xbrlview.fasb.org, 336–337
Xero, 210
Y
Young, Steve, 435
Z
Zobel, J., 158n2
Zobel, Justin, 160
Z-score, 100–101, 103
diagnostic analytics, 235–236
outlier detection, 229
standardizing distributions with, 144
page 519
page 520

Data Analytics For Accounting, 2nd Edition Vernon Richardson PDF

Uploaded by

Data Analytics For Accounting, 2nd Edition Vernon Richardson PDF

Uploaded by

page i

Data Analytics for Accounting

DATA ANALYTICS FOR ACCOUNTING

Published by McGraw-Hill Education, 2 Penn Plaza, New

Some ancillaries, including electronic and print components,

This book is printed on acid-free paper.

Cover Image: Rawpixel.com/Shutterstock

The Internet addresses listed in the text were accurate at the

My wonderful daughter, Melissa, for your

My wife, Erin, and children, Sylvia and

To my co-author, friend, and colleague,

About the Authors

Vernon J. Richardson is a Distinguished Professor of Accounting

Ryan A. Teeter is a Clinical Assistant Professor of Accounting in the

Amelia Annette Baldwin

Main Text Features

Multiple Choice Questions

Data Analytics for Accounting,

Chapter by Chapter Updates

Connect for Data Analytics for

With Connect for Data Analytics in Accounting, your

Labs: Select labs are assignable in Connect but will require

Test Bank: The test bank includes auto-graded multiple

You’re in the driver’s seat.

Less Time Grading

They’ll thank you for it.

Make it simple, make it affordable.

Padlock: Jobalou/Getty Images

Solutions for your challenges.

Checkmark: Jobalou/Getty Images

Effective, efficient studying.

Study anytime, anywhere.

“I really liked this app—it made it easy to

Calendar: owattaphotos/Getty Images

Learning for everyone.

Brief Table of Contents

Examples of the Regression Approach in Managerial

Address and Refine Results 230

Lab 8-2 Create Dynamic Common Size Financial

A Look at This Chapter

The Chinese e-commerce company Alibaba is perhaps the biggest online

LO 1-1 Define Data Analytics.

HOW DATA ANALYTICS AFFECTS

There is little question that the impact of data analytics on business

HOW DATA ANALYTICS AFFECTS

Data Analytics is expected to have dramatic effects on auditing and

Data Analytics also expands auditors’ capabilities in services like

THE DATA ANALYTICS PROCESS

Data Analytics is a process to identify business questions and

In addition, to give us some idea of the data questions, we may

Data reduction —A data approach that attempts to reduce the

While we don’t believe that accountants need to become data

HANDS-ON EXAMPLE OF THE

Here we provide a complete, hands-on example of the IMPACT

Identify the Questions

Master the Data

Source: Accessed March, 2019. https://www.lendingclub.com/info/statistics.action.

Borrowers borrow money for a variety of reasons, including

Source: Accessed March, 2019.

Amount Total requested loan amount

Application Date of borrower application

Loan Title Loan title

Risk_Score Borrower risk (FICO) score

Dept-To- Ratio of borrower total monthly debt payments

Zip Code The first 3 numbers of the borrower zip code

State Two digit State Abbreviation provided from loan

Employment Employment length in years, where 0 is less than

Policy Code policy_code= 1 if publicly available.

Amount Application Risk

Perform Test Plan

The second analysis was on the length of employment and its

Source: Microsoft Excel, 2016.

The third analysis we perform is to consider the credit or risk

Address and Refine Results

Source: Microsoft Excel, 2016.

Develop an analytics mindset.