Vernon J. Richardson is Distinguished Professor of Accounting and the Glezen Chair in the Sam M. Walton College of Business at the University of Arkansas. He served as Accounting department chair. He is also a visiting professor at the International Business School Suzhou, Xi'an Jiaotong Liverpool University.
Katie 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. Katie has taught at universities around the world and is a member of the American Accounting Association.
00:08 Matt Waller: Hi. I'm Matt Waller, Dean of the Sam M. Walton College of Business. Welcome to Be EPIC, the podcast where we explore excellence, professionalism, innovation and collegiality, and what those values mean in business, education and your life today.
00:28 Matt Waller: I have with me today Professor Vern Richardson, who is distinguished Professor of Accounting, and Katie Terrell, who is an instructor in the accounting department. And they are co-authors on a new textbook, it's the first of its kind, in Data Analytics for Accounting. And this is such a big topic in every discipline. It seems like there aren't textbooks in data analytics and many other disciplines, and you all have the first and only one, and we're thrilled that it's coming out of the Walton College. That's great. And I know that there's many features of the book. You actually came up with a framework, that we'll talk about later, to think about data analytics and accounting. And from what I have learned about it, I'm really impressed by it. But you've also used data from Dillard's department stores, we had access to a huge amount of data from them, and other companies, so that's pretty impressive. Congratulations on the accomplishment.
01:37 Katie Terrell: Thank you.
01:37 Vern Richardson: Thank you. We're glad it's done.
01:38 Katie Terrell: That's right.
01:40 Matt Waller: Well, seriously flipping through it, I thought, "Oh, my gosh, this must have been a huge amount of work." I can't imagine it, but thank you for doing it. But the other thing that occurred to me about this book is that... I know it's a textbook, but I would think businesses would want to learn how to do this as well.
02:03 Vern Richardson: Yeah. We've talked to basically all the big accounting firms and just tried to get an idea of how they approach it, and I think it actually took us some time to figure out what data analytics means to them. And I think we tried to go first to really say what we think the minimum needed to be in a textbook, a theoretic frame, some experiential data, the analysis that you actually have to go through and get your hands wet or dirty, and try to solve some questions that people have, and accounting questions. A lot of times we get feedback like, "The text is good but especially because it deals directly with accounting." And there's business analytics books, and there's other analytics books out there, but accounting seems to be the one since it's an accounting question that we address that seems to resonate with students, and instructors, and the profession.
02:57 Matt Waller: Knowing the demand for data analytics in general, if you Google data analytics and there's a lot of results that are generated, but it's clear the interest is very high. What made you decide, "Yeah, I wanna go into this, I wanna write a textbook in data analytics"?
03:17 Katie Terrell: I would mostly blame Vern for that, but it is a class that we've been teaching here for four years now. And we've been teaching advanced accounting information systems, and I began teaching it without a book. I had to create the material or cobble together material that already existed online. And you mentioned that there are business analytics textbooks and other data analytics textbooks for other fields, and so you can kind of gather all of that together and try and explain how it fits into the accounting perspective. And that takes time, and it takes effort, and it's worth doing, but then you see that you're not the only one doing that. There are other universities that have professors and instructors who are trying to do the same thing. And then there's just a real wall for anyone who doesn't know how to do that or doesn't have the time to do that, and it prevents the subject from being taught. And so we just thought, "You know what, we've been teaching it, we've been finding these materials, we've been researching what needs to be taught in this class, let's just write the book for it, let's make it happen." And because the book exists, this class is being able to be taught all over the place by people who wouldn't have the first idea how to create a SQL query in the first place, but they have a book that will teach them so they can teach their students. So, it's a really, really valuable resource.
04:31 Matt Waller: Are other schools adopting it yet?
04:34 Vern Richardson: Last number I heard was 4000 copies sold in the first nine months.
04:39 Matt Waller: Wow.
04:39 Katie Terrell: Right.
04:41 Vern Richardson: So, we're very pleased where it stands, and it just means people are using it. A lot of times it's, "Can the instructor learn it?" 'Cause what usually happens, the department chair would go to the instructor and say, "We need data analytics, you're teaching it." And so they immediately say, "Oh, no, I'm not trained in this." And actually a lot of our research training helps a lot in this area. But a lot of them say, "Okay, what materials are there?" And they often start heading toward our book, and at least will look at it and consider using it.
05:11 Matt Waller: I wanna just talk for a moment about the model you have, the impact cycle, you call it. And one feature of the book is this theoretic framework, and I wanted to just say a lot of times people in practice, of course, think of theoretic as not practical but, of course, as we know, a good theory is very practical because it describes, explains and predicts whatever it is you're doing. So by having this framework that you're using throughout the book, it benefits the students, and it helps you predict good outcomes and explain why you're teaching it this way, why the book is laid out this way. How did you come up with the model?
05:55 Vern Richardson: The impact model actually comes from another textbook provider, and we've pretty much stolen it and cited it, but then we try to use it throughout. So, in every chapter, the first four chapters are kinda the foundation for the book so we'll go through the impact model. And then every lab we use, we also follow the impact model. I stands for identify the question, M for master the data, P for perform analysis, A for address and refine results, and C is communicate insights, and then T is tracking outcomes. And it's a recursive cycle. Essentially you go through this process, or maybe just do the first two steps and say, "I've got a better question now," which happens all the time and you go back through and do it round and round again until you have answered the question that's before you. And the idea is that we're supposed to be information providers to the decision-makers. The decision-makers have a question and we say, "Okay, let's see if we can address it."
06:53 Katie Terrell: Yeah, absolutely. I'll add to it that when you say the term "data analysis," we're going to do data analysis. People tend to think about the "P" of the step, the perform the test plan. "Oh, data analysis, that's regression, or that's a pivot table in Excel that's doing the work, when you can only do the work if you know what question you're asking, you can only do the work if you know how to gather the data and clean it. The work only matters if you can explain it, if you can communicate it, and then assess how often you should go back and track and look at it. So, it really puts the meat of data analysis in the context that is really critical for our students.
07:27 Matt Waller: I would like to talk just briefly about this data. Would you describe it a little bit?
07:33 Katie Terrell: The data set we are the most excited about using and that has the most prevalent presence throughout the text is from Dillard's department store, like you mentioned, and we're thrilled that we have access to that. The Walton College has provided that data, and that data is not provided just to University of Arkansas students. Anyone can access it by getting the right permissions with the dot edu, and they can get in there and they can work with this really beautiful data set that is real. And so what we love about having this access in our textbook, and we walk students through how to get to the data, how to clean the data, how to work with it and analyze it in lots of different ways and lots of different forms. It's so important because it's messy. Students are so typically provided a beautifully clean set of data. They copy it from the back of the book, or whatever it is, it's just there and it's ready for analysis. And if all our students can do, when they enter their professional lives, is work with clean data, they're a mile behind where they need to be. So, with the Dillard's data, students are learning how to join data with SQL, they're learning how to pull that data into Excel, and clean it the way they need it to be to work with and analyze it.
08:44 Matt Waller: If I were taking this course, I might be afraid that... Well, I don't know SQL, is that okay?
08:51 Vern Richardson: That's part of the course, is we realize you didn't walk into the course knowing SQL or any of the other tools we're using. So, usually what we'll do is we'll go step by step, screenshot by screenshot, click here, point there, enter this here, and that's pretty easy. What's less easy is, at the end of each lab, we'll usually say, "Okay, we did it for January now, do it for December." Or, "We found out this about sales return data, now we need to see if we can predict it for the future, how would we do that." And so sometimes it's just written what they think they should do. Other times it's jump back in, go back to step three, and start over and do it again for the new year. Then they've got to modify it left and right, and change it in order to get the new result they need.
09:38 Matt Waller: Vern, you've written a lot in your life. You've written lots of articles for academic journals. You've written other textbooks. What are other textbooks have you written?
09:50 Vern Richardson: Mostly accounting and information systems so it's fairly standard, it's a survey of accounting systems, and that's where we've decided to work. And this is kind of the next stage of that. But it's something that gets easier as you write, you put yourself in the students' views, say, "What's a good question? What's a bad question? How would I ask this in terms of multiple choice? How would I explain this? What things would I cite? What have I learned?" It's something that actually becomes easier as you do a lot of it.
10:21 Matt Waller: And this textbook, I just read a couple of random paragraphs, but I can tell it's very well written, very well organized. You have a glossary, you were talking about screenshots and so forth. Is there an online companion to go with this, or is it not necessary?
10:42 Katie Terrell: There is. Through the publisher, through McGraw-Hill, they have provided a Connect companion to it. Connect is just McGraw-Hill's online resource, and not every lab and not every problem, but many labs and many problems can be worked either just completely in the text or through the Connect component. Of course, that's where students will go to gather the data that they need to work with, and it's also where they could go when there's auto-gradable labs and auto-gradable discussion questions, and that sort of thing, that make it a little bit easier for instructors.
11:15 Matt Waller: Now we're gonna talk just briefly about one of the labs that you have, and it's called Dillard's store data: Hypothesis testing. Of course, Dillard's department store is one of the most successful department store chains in the world. They're in 29 states, have over 330 stores. Most people listening to this will have probably shopped at Dillard's before. Their headquarters is in Little Rock, Arkansas, which is three hours from our campus. But Bill Dillard II graduated from your accounting program before you all were here.
11:53 Vern Richardson: That's right, we didn't train. He's trained enough.
11:56 Matt Waller: I know him quite well. He really attributes some of his success and Dillard's success to his strength in accounting. He loved accounting, and understanding accounting is so important for business in general. Dillard's department store data, what is this... Let's review this lab as an example.
12:21 Vern Richardson: One of the requirements of accountants is we try to match everything, so the revenues have to be matched with the expenses. And related to that is a recent promulgation saying not only expenses have to be matched but also sales returns. So, you get to the end of the period, and you have to predict what sales returns are gonna been even though they might not come back for a month or six months after the period. We had the idea, okay, what we're gonna do is we're gonna just have that innocent question of our return is different around holiday season versus non-holiday season. The holiday season we sell more, but I know, soon after the holidays, I'm gonna take a couple of things back. So, we're trying to predict that. And so we had the initial question, "Could we compare holiday season and non-holiday season?" And it's particularly interesting because the close of the year is soon thereafter, and so they're gonna close their books in January and all the returns haven't come back yet.
13:14 Vern Richardson: Second part of this is, just because we're in Arkansas, we said, "Is the sales returns of Arkansas different than other states? Just because Dillards is more prevalent or, I don't know, maybe Arkansans are different, somehow could we predict this differently?" And so that's kind of the setup, and how do we address it? Well, we go to the master, the M of the impact model, and I simply say, "Okay, let's get the data and then start to ask the question." So, we get the data, and we call it ETL, extract, transform, load, and essentially have the data ready for use, and then go through it. And in this case we use Excel and we just basically run a regression analysis, or could be a T-test, where we run a regression analysis just to see if they're different.
14:01 Matt Waller: A lot of times when you're dealing with this kind of sales data, you wind up having to do a lot of data cleansing. Is this data already clean?
14:12 Katie Terrell: No, it's not. We each have different favorite part of the lab. My favorite part is the master, the data piece, the ETL. Because, no, it's not clean data and its not ready for analysis. The student has to first access a remote desktop so that they can get to the Dillard's data within that remote desktop environment. The lab walks the student through how to do all of this, but then within that environment, they can pop into Excel. And through Excel, they can connect to SQL server, which is where the data is stored, so they can extract that data using some SQL code within the Excel environment, bring that into Excel. Using Power Query, they clean that data. So, there's lots of pivoting columns and massaging that data to get it ready for analysis. Also keeping in mind that it's a massive data set, so Excel alone can't hold all the records that we want to analyze. The Power BI component of Excel can, so we can have all of those records and Power BI, clean it, get it crunched and ready to fit in Excel. And then from there they can run their T-test or the regression analysis.
15:14 Matt Waller: Well, that's so important because so often students get out of college and they know how to run regression or all kinds of statistics, but they don't know how to cleanse and harmonize data.
15:27 Vern Richardson: There is a statistic that 50% to 90% of a data analyst time is spent cleaning the data. So you imagine pushing the regression button, it might be 20% of the time and the rest of the time getting the data right. And what do you do with the zero? What do you do with the negative sales amount? Is that a return or is that a mistake? What do you do with a negative 999? Those are all questions, and you have to know how to address them. You can't just delete them. You actually have to do something with them. That means really getting to know the data, what does that mean. And once you get used to the data, and sometimes you call up a friend or call up someone at Dillard's and just say, "What's going on here?"
16:07 Matt Waller: Now, this is really interesting to me 'cause I've done lots of data cleansing in my life for analyses I've done, and I know, if you're good enough at data cleansing and data harmonization, you can unfortunately affect the outcome. Even if you just... You have no agenda, you wanna be completely honest about it, you still have these decisions to make. And I would think... So if in accounting that's required now to forecast returns, do they have any guidelines on forecasting methods like you're using regression to do the forecasting?
16:43 Vern Richardson: Usually predictive ability, even just an R squared number or how well are you predicting the future type of thing, and then continuous checks. And if you're way off, then go try something new, try some of the multi-varied fashion, or some other way. But I think that's the baseline. And one of the problems is we're not at Dillard's, we're not doing sales returns, and so we're just looking at it. But then we try to look through and say, "Hey, this model worked, or this model didn't work as well." And then that's one of the questions is, "What else should we use? What else is available to use? If you could have any data in the world, what would you use to try to explain returns?" And so those are open questions that we hope people have, students have so that they then can learn from the data and say, "Okay, this is how I do it when I do work for Dillard's, and I am trying to predict sales."
17:34 Katie Terrell: Absolutely, and looking at the strange things that they might come across in the data. And are they errors, is there a consistent error with this? And the student should then be able to think, "Okay, when I work at Dillard's, what can I do to make sure that data error isn't happening?" Taking it back to the relational database design and everything, because it's all connected.
17:58 Matt Waller: If you think about it, technology is allowing us to really do a better job of teaching. You couldn't have done this 20 years ago. Well, you could have, but it would have been very hard. First of all, it would be hard for us to have that much data in Walton College. And then the tools now that are available to clean the data, extract the data, analyze data, etcetera, etcetera, they are just so much better as well.
18:26 Vern Richardson: I think you'll know from your splicing background. But activity-based costing, usually we pick two or three drivers, and someone will say, "Here's your two or three drivers." Now we can say, "Hey, here's 10 possible drivers, which ones should we pick?" So, we run our regression analysis and say, "These three loaded, these seven did not, I guess these are our three. Maybe we'll check it again next year, but this is the three we'll use for now. And that seems to explain overhead cost or costing in some way, and so let's use those and then... " That's just a different level that you wouldn't have expected students to use 20 years ago.
19:01 Matt Waller: I'm thrilled about the book, this is really impressive. Congratulations to both of you.
19:07 Katie Terrell: Thank you.
19:08 Matt Waller: Now, this new book you're working on that's not finished yet, and it's gonna be more focused on using Excel, is that right?
19:15 Katie Terrell: It is, yes.
19:17 Matt Waller: Will you also be using Dillard's data and other companies' data?
19:21 Vern Richardson: At this point, probably not, just because we're trying to hone the data analysis skills, or even Excel skills, of our students. And because you have to go to Remote Desktop, because you've got to log in, because you've got to do SQL, and all of those, those are important skills, but you've taught at this level, sophomore level student, and just getting them to run Excel might be as far as we get. And that's okay. You've got to take them where they're at.
19:48 Katie Terrell: Right.
19:50 Matt Waller: In the Excel course, will they learn things like, I don't know, how to create pivot tables?
19:54 Katie Terrell: They will, absolutely, yeah. They'll have the slightest introduction to Excel of the hand-holding, of difference between formula and function just to make sure they're even confident working within a blank worksheet. And then it will be pivot table, it will be tables in general, regression, T-tests, that sort of thing.
20:11 Matt Waller: Will they use macros at all?
20:14 Katie Terrell: I don't think that we're planning on having macros at this point with it. I think the next elevated skill that would be involved would be moving from Excel into Tableau, so working with more data visualization and seeing else that we could do...
20:26 Matt Waller: So, you will be using Tableau?
20:28 Katie Terrell: We will be using Tableau, yes.
20:30 Matt Waller: Well, that's just my awareness of what's going on in supply chain and CPG, in retail. Excel and Tableau are huge.
20:38 Katie Terrell: It's huge. It is, it's something that employers are really excited about. The students love working in Tableau. Once you get over the initial fear factor of a new application, it's really a friendly application. You can just make such cool stuff with it. It's exciting and fun, and really marketable.
20:55 Matt Waller: It's really neat that these two textbooks that you all are creating, students become facile with Excel, and Tableau, and SQL, and other tools, but what is your real emphasis?
21:10 Vern Richardson: Well, what we really want is the critical thinking skills. A lot of what we do here can be done by computers, but asking the question or having that expertise may not ever be able to be done by computers. So, what we'd like to do is really specify the question carefully and then curate the data to make sure you can answer it, that it's the right data at the right quality level that will actually answer the question, and then provide some results and communicate them so that the decision-maker can make a better decision. But it's not really the tool. We use seven different tools, but the emphasis really is on the thinking skills and going through the whole process time and again, and saying, "Okay, I know I can now handle this even though I go work at a firm and they use Alteryx, and I've never used Alteryx and what will that look like. In three years, they'll use something completely different and better, or theoretically better." And yet if they've done this process, they kinda know what to do even though the functionality of the tool might be different.
22:10 Matt Waller: Well, I hope that other colleges of business around the country and around the world adopt this new approach that you guys have created. I'm grateful that our students are first movers on this, that they're getting to learn this material.
22:27 Vern Richardson: Well, it is nice. The Walton College is all over this textbook, between Dillard's and everything that we're trying to do, so we think Walton College has given us the tools and the time and the pat on the back to say, "This is good, and so we wanna do more of it."
22:47 Matt Waller: Thanks for listening to today's episode of the Be EPIC Podcast from the Walton College. You can find us on Google, SoundCloud, iTunes, or look for us wherever you find your podcasts. Be sure to subscribe and rate us. You can find current and past episodes by searching beepicpodcast, one word. That's B-E E-P-I-C podcast. And now, be EPIC.