Episode 95: Data Analytics with Matt Storlie, CFE, CIDA
In this season of The Data Sleuth Podcast, titled "Conversations about Fraud," guest host Justin Burns, managing partner of Space Coast Forensics in Brevard County, Florida, tackles topics including embezzlement, collaborative divorce, economic damages, construction fraud, and more. In each episode, Justin is joined by an industry expert to help tell the story behind the numbers and explore the latest in fraud detection and prevention.
In today’s episode we discuss data analytics with Matt Storlie, CFE, CIDA. Our conversation includes:
How artificial intelligence is affecting data analytics.
Why forensic accountants need to be creative when planning their analytics.
A case story about using data analytics in a class action lawsuit.
The 3 C’s of data analytics.
GUEST BIO
With 30 years of experience in business and accounting, fraud examinations, data analysis, internal audit, and IT systems, Matt Storlie assists clients with forensic services, internal investigations, computer forensics, e-discovery and litigation readiness, and fraud risk assessments. As one of the firm’s subject matter experts on data analytics, he utilizes his deep experience with IT systems, internal controls, and data analytics to develop proactive, risk-based anti-fraud models and programs, and is a certified data analyst of IDEA® software.
Email: matt.storlie@wipfli.com
LinkedIn: Matt Storlie, CFE, CIDA
RESOURCES MENTIONED IN TODAY’S EPISODE
To learn more about the Investigation Game Education Edition, visit: workmanforensics.com/tig-educators
Order your copy of Leah’s book, Data Sleuth: Using Data in Forensic Accounting and Fraud Investigations today on Amazon!
CONNECT WITH WORKMAN FORENSICS
Youtube: @WorkmanForensics
Facebook: @wforensics
Twitter: @wforensics
Instagram: @wforensics
LinkedIn: @workmanforensics
CONNECT WITH JUSTIN BURNS, CPA, CFE
Website: www.spacecoastforensics.com
LinkedIn: @space-coast-forensics
LinkedIn: Justin Burns, CPA, CFE
Subscribe and listen to this and more episodes of The Data Sleuth® Podcast on Apple Podcasts, Spotify, Android, or anywhere you listen.
Transcript
Leah Wietholter:
Hi, I'm Leah Wietholter, CEO and founder of Workman Forensics, and this is the Data Sleuth Podcast.
For the 2024 season, we are changing it up a bit. I've invited Justin Burns, a forensic accounting professional and fraud fighter, to guest host the episodes. He's one of the friendliest and most personable professionals I've met in this space and I'm excited for you all to meet him and his guests.
Justin Burns:
Hi everyone. As Leah mentioned, I'm Justin Burns, Managing Partner of Space Coast Forensics in Brevard County, Florida, and I'm your host for this season of the Data Sleuth Podcast. I'm excited to host this season titled Conversations about Fraud, where we'll tackle topics including embezzlement, collaborative divorce, economic damages, construction fraud, and more. In each episode, I'll be joined by an expert to help tell the story behind the numbers.
In today's episode of the Data Sleuth Podcast, Matt Storlie and I are going to talk about data analytics. Matt Storlie is a Senior manager in the Forensic and Litigation Services Department of Wipfli out of Minneapolis, Minnesota. Nearly 30 years of experience in business and accounting, fraud examinations, data analysis, internal audit, and IT systems, Matt Storlie assists clients with forensic services, internal investigations, computer forensics, e-discovery and litigation readiness, and fraud risk assessments. As one of the firm's subject matter experts on data analytics, he utilizes his deep experiences with IT systems, internal controls, and data analytics to develop proactive risk-based anti-fraud models and programs and is a certified data analyst of IDEA Software.
Hi Matt, thanks for joining me on today's episode.
Matt Storlie:
Hey Justin. Thanks for having me.
Justin Burns:
So Matt, I know you work on all kinds of engagements including forensic accounting engagements, internal investigations, and e-discovery and litigation readiness, but today I want to focus on the data analysis portion of your job. I know you're one of your firm's subject matter experts on data analytics. Can you tell me what it is about data analytics that you find so interesting?
Matt Storlie:
Sure, yeah. I think I've actually been doing data analytics for a long time, longer than I've been a fraud investigator. I became a certified fraud examiner in, well, let's just say 17 or so years ago, but I've actually been doing data analytics longer than that, a good 20 plus years. I started in internal audit doing data analytics for a large Fortune 500 retailer here in the Twin Cities. And I think what I found most fascinating about data analytics is the ability to take data, because every organization has data, and create something with it, make something useful out of the data, tell a story with the data, if you will.
You hear a lot about data is the new oil, different little quotes like that, and I think there really is a lot of power in data. And as long as you have good data to work with and you are creative enough to tell a really good story, it can be very powerful. In internal audit, especially when I was working at internal audit, I just noticed that if we had a problem, let's say, or an issue that we had identified and we were trying to figure out how big of an issue it really was, using data analytics to really quantify that issue, I found very powerful because we didn't know if it was a small or big issue. A lot of times that can feel like a gut feeling, like is this a big issue or not? We use that data to actually quantify it and is it a $5,000 issue, is a $5 million issue? Once we got the data and figured it out, we kind of knew exactly how big of an issue it was.
In forensics, I've used data analytics the entire time I've been a fraud investigator. They've gone hand in hand. And I've used analytics to find the needles in a haystack and to basically identify all of the additional fraud that we may know there's some fraud but the analytics can really help kind of fill out the picture and identify all the other potential fraud as well, as we're doing the investigation. So I just find a lot of power in data and analyzing it. I did have a computer science background in college, so I think that kind of affected me a little bit as well. I do enjoy working with programs and programming and so data analytics also just feels like kind of another natural outgrowth from that as well.
Justin Burns:
Yeah, and I mean I feel like as a forensic accountant we may not say what I'm doing is data analytics, but a lot of what we do is data analytics, whether we call it that or not.
Matt Storlie:
Right.
Justin Burns:
And you mentioned working with different programs and things like that with your computer science background. It feels like every conference you go to or every continuing ed thing you go to, lately there's sessions on AI. The most recent conference I went to, three and a half day conference, it felt like one whole day was all sessions about AI and how AI is going to change accounting and ways you can implement it. How do you see AI being used to assist with data analytics or how do you see data analytics changing with the implementation of AI?
Matt Storlie:
Sure. I mean it is, it's coming up on, I think two years now, since kind of ChatGPT exploded, I think it was November of '22, and it fascinating to see the quick adaptation of it. Everyone talks about that, and you're right, every conference we've attended or been a part of it seems like, and even the Institute of Internal Auditors had a AI analytics and automation conference earlier this year. So it is the talk everywhere.
As far as accounting and data analytics specifically, we have to remember it is only a couple of years old for the most part. I really see AI as more of an exploratory thing right now. And when it comes to accounting and data analytics, I think we need to be careful as far as how much we're exploring within AI and utilizing AI for that portion of it. And I see data analytics as a very finite contained sandbox that we're doing all my analytics in. When it comes to artificial intelligence, again, I think it feels more exploratory. I think maybe if you've got something that's maybe more autonomous where you want to utilize something on a recurring basis and things are set pretty well as far as all of the analytics and the metrics and stuff, then it could be a very effective tool for that, kind of speeding up processing, if you will.
But I think we're still in the early stages of it. I still see so many clients that still struggle even with some basic data analytics. I just think it's going to be a long time for that maturity curve to really grow and for AI to be fully integrated into accounting and analytics. But I think utilizing the power of machine learning, generative AI, and all these buzzwords out there, I think we will continue to utilize it as much as we can going forward. But I think it's going to take a little while for us to get there.
Justin Burns:
Yeah, it's useful for here just churn some numbers, things like that. But it sounds like what you're hinting at is you still need a human touch to it. You still need a practiced professional to have some judgment in there. And that's one of the things that we talked about getting ready for this podcast was one thing I think you brought up multiple times, I kept hearing you say data correlation, data correlation, and having the human ability is not really something that AI is able to replicate yet. Can you talk more about that?
Matt Storlie:
Yeah, I mean, just think about, I'm sure a lot of folks have tried AI and they've maybe generated an image or something and you know how it just may not look quite right, something's a little off.
Justin Burns:
It can never get hands right for some reason.
Matt Storlie:
So six fingers or whatever it is, right?
Yeah, I mean it's just not quite there yet. And I think it does. I kind of feel like the work that I've been doing over 20 years in data and analytics has been human intelligence and not artificial intelligence. And you do need that human touch. You need to set the boundaries, you need to set the parameters, you need to ensure that the data, that whatever program you're using that you are ingesting, you think of AI too, you think of it massive amounts of data, but how much of that is garbage in that's going to be garbage out once it goes through the other side? So ensuring that the data that you have coming in is we'll say cleansed or normalized, is in a good spot.
And then the correlation piece is kind of the next piece of it. Think of machine learning as kind of human learning too. So if I want to identify something, I do some exploratory kind of analytics and profiling of the data, but then once I start trying to correlate the data and make more sense of it, imagine that you don't get a flat file with every possible field you can imagine and every row imaginable. You don't just get one big massive file. You have to realize when you're starting to try to correlate data, and I've done multiple presentations on this, is you have to think more like a computer system. And the fact that you have multiple databases, you have to think with relational databases, and realize if you're looking at an account's payable file or purchasing file, you're going to be missing some data. You might have a vendor ID but you're not going to have all the other information from the vendor included in that file necessarily.
So correlating data really gets to the point of having multiple databases and correlating the data between the databases so that you can really start to build more robust testing and analytics from that. That is just something that it takes a human touch, it takes iterations, and we can try to speed it up potentially with AI, but I think for now it's still we need to be the drivers of that, generating those developing models, generating the information, whatever we need to do, the formulas, everything to build that at this point. But yeah, I think that's just the human aspect of it. And there's a creativity piece too, which I think maybe we'll talk about as well, but creatively, there's only so much that an AI type model might be able to do. As a fraud investigator the sky's the limit as far as what you can come up with as far as fraud tests and stuff. And that's something again, I don't think an AI type model is necessarily going to have. You're going to have to feed that into the model for it to know how to actually process all that data.
Justin Burns:
And also you're talking about, you put the data in, you're getting the results from it, but I don't know that I'd feel comfortable relying on AI to do the interpretation piece. Sure, do my calculation pull, ones that based on my parameters, here's the transactions that meet your parameters that you say look funny or whatever, however I term it. But let me do the interpretation on my side. And there's other AI programs out there that attorneys are always like, "Well, let me just pop this into ChatGPT and it'll fix all my wording for me." And there's other ones to help you with the writing if that's something that you want, but it should all come from your ideas. Like I use an AI program for my email for it to a plugin in a word, Grammarly, because not the greatest with punctuation and spelling, all that stuff. And it helps me that way, but the idea is everything's still mine. I don't really feel comfortable with an AI program running my calculations and then telling me how it's interpreted it yet.
Matt Storlie:
Right. I think it's first drafts of something.
Justin Burns:
Right.
Matt Storlie:
And it's kind of an idea of something, but you still need to ensure that that idea comes to fruition in the right way and you wouldn't want to present upper management something that you ran through AI, or if I'm doing a fraud investigation, I wouldn't want to utilize it and say, "Well, yep, ChatGPT said this, so it must be fraud." I mean there's always the validation piece, the verification piece, that has to come after that to ensure that what you're relaying is accurate.
Justin Burns:
Yeah, because you're still responsible for the end product. You can't just go back and be like, "Oh, well I plugged this in and this is what it gave me." That doesn't sound great.
Matt Storlie:
Right, right. And I can't stress enough the importance of the quality of the data that you're utilizing as well, even before you get to the fun analytics. I spend a lot of my time at my firm on the front end process, the cleansing process, the normalizing process and that's also where it gets kind of fuzzy for me with AI, because it's like, "Can we completely rely on all this data coming in?" There's a lot of unstructured data out there. But I'm typically working in structured data for the most part. Probably 90% of the work I do is in structured data. And so I want to ensure that that data that I'm pulling in and putting into my model or doing whatever with is accurate, is complete and accurate, or otherwise it is going to be garbage in, garbage out, GIGO, the old acronym. And in my experience I've seen it's typically been 65, 75% of the effort is really that cleansing upfront effort before I even get into what I call doing the fun analytics stuff.
So I think I just to make that a point as well. It's like that front part is really important before we get into the, because what I find too often too are firms where they will decide, let's say they're in the middle of an audit and they'll decide halfway through the audit, "Oh, let's run some analytics on this." And you and I both know at that point it's going to be tough because you don't know if you're going to get data, if it's going to be good data, and there's this whole process that you go through. So as much as you can push that to the front end planning part of whatever project you're on, the better off you're going to be. But yeah, I just wanted to kind of talk about that too because I feel like sometimes it's glossed over, but it's a real critical, important aspect of any data analytic project that you might be performing.
Justin Burns:
Right. And I just want to ask you about a couple of times you use there, structured versus unstructured data. Can you tell us the difference?
Matt Storlie:
Sure. I mean, structured data, you really think of a database. You think of columns and rows. And that's typically what we're working with in data analytics, but we are seeing more and more application of unstructured data, which is emails, it's whatever unstructured data you can think of that you can't put into a box necessarily. And with a lot of the AI type products, I think they do utilize a lot of that information. I tend to stick primarily with the structured data because again, I feel like I could put structure around it and I know exactly what it is I'm looking at, even if I have multiple tables that I'm working with, most of the time that information is structured and I can correlate it and do all the analytics that I want.
Justin Burns:
Perfect. All right and I think that gets us to our ad break. So we're going to take a quick break and we'll be right back.
Speaker 4:
Breaking into the field of forensic accounting can be overwhelming, leaving you wondering where to begin. The best way to gain experience is by working cases. But how do you work cases to learn without the results impacting the outcome of a real case? Leah's book Data Sleuth: Using Data and Forensic Accounting and Fraud Investigations contains case studies and solutions for you to practice planning a case, from identifying the client's concerns and connecting them to the appropriate analyses and finding the most relevant data to perform the analysis. Build your confidence in addressing client needs through reliable analyses using the Data Sleuth book and free practice downloads as a guide by ordering your copy today on Amazon.
Justin Burns:
All right, we're back from break with Matt Storlie talking about data analytics. Before the break, Matt, you talked about getting creative in your data analytics and it's something as forensic accountants we have to do just to solve whatever issue we're running up against. Sometimes we got to get creative and it's part of the job that I love. Do you have any case examples for us where you've had to be really creative with your analytics to help solve a case?
Matt Storlie:
Yeah, sure. I mean, I've got a lot of examples. I've been doing this a long time. And one case that kind of strikes me as being able to utilize it more than any other, I've done a lot of fraud investigation, forensic cases, embezzlements, whatnot. And one case I worked on a few years ago was actually was turning into a class action lawsuit out of California. And we were representing the company, essentially the defendant, so it was an individual who was starting to bring a class action lawsuit against the company. California, of course it's pay, it's wage and labor type stuff. And it was a wage and labor case where they were claiming that the time systems that the company was using was rounding down, affecting the calculations and rounding down. So it was basically they were claiming that they were getting underpaid and were also not getting paid for say, meal breaks and lunches and stuff like that because of the way the systems were operating within the company.
And so it was interesting because it was really a data analytics heavy case. And I worked on the case and I was basically queuing up to be potentially even a testifying expert explaining all the data analytics that was the driving force behind the case for the defendant, in this case the company. And of course it was a situation where it was a long time period, almost five years maybe, and so there were multiple systems involved as well. They had one system, they switched to a different system, then they went back to the other system. So it was a case where I was dealing with tens of thousands, hundreds of thousands, of records, time records, and multiple systems, multiple formats. And we got creative because we wanted to kind of illustrate because they were trying to make it a class action lawsuit, illustrate how the system was I? Actually decrementing that individual's time, was it maybe potentially assisting others? Maybe some of the time was going down. Some of it was going up.
So basically we worked with a sample, but then we also looked at all the records to see on a time basis to see how exactly everything was being calculated. And there was a lot of data and there was a lot of slicing and dicing trying to correlate that data and figure out exactly what was going on. And at the end of the day we identified that some people were benefiting from this, some people were not benefiting from this. The reality was on an individual basis overall it was actually pretty darn close. And so what we were able to do through the case was identify basically what this individual was trying to bring forward and try to turn into a class action lawsuit, they were going to have a hard time proving in court because we had hard numbers and data analysis that really illustrated that it was not necessarily the case.
It didn't go to court. It went to mediation. And I think they identified what was going on in the analysis and could kind of see that they were going to have a hard time proving this because we had already kind of done a lot of the work to prove the other side of it. And they ended up settling in mediation. So as opposed to this becoming a very large class action lawsuit, multi, hundreds or tens of millions of dollars, they were able to settle in mediation, my client was. And it was interesting because the day that they were in mediation, I was ready all day just in case they were to call to have me talk about exactly how I had performed the analysis. So it was just kind of an interesting first time I had done a case like that, that was really kind of not fraud per se, but it was using the analytics to actually prove the case.
Justin Burns:
Yeah. And I mean, you just get handed this puzzle and you're talking about the different systems, you've got all these pieces that don't really jive with each other, and you've got to figure out how to make them make sense along with each other. So, yeah, certainly some creativity involved. And I feel like that's a lot of our job is, "Okay, how can we answer the question that our clients have with what we're able to get?"
Matt Storlie:
Right. Right. And I like to say too, Justin, whenever I'm presenting, "We do have a puzzle to put together, but we don't have the box as a reference." That's what makes it so hard. There's no linear progression. You're kind of all over the place and you have to be willing to go to those spots and connect those dots to try to figure it out.
Justin Burns:
Yeah, you have no idea what it's going to look like at the end. You just got to figure out how to get it all put together and see what the ending is.
Matt Storlie:
Yes. Yep. Exactly.
Justin Burns:
So do you have any other case stories where maybe the data analytics really shined through? You talked about in the California class action case where you did the data analytics and the other side didn't really have as big of a case as they thought they did.
Matt Storlie:
Right.
Justin Burns:
Is there another one like that where maybe they thought it was bigger or smaller and the data analytics got run and it really shined through for you?
Matt Storlie:
Yeah. I worked on a financial statement fraud case many years ago. So I have several models that I built using analytics to identify potential fraud, waste, and abuse. And the model is really used, and I present on it quite frequently as well, to the Institute of Internal Auditors, the Association of Certified Fraud Examiners, whatever. I do a lot of presentations on this because it kind of combines a lot of what we're talking about today, about cleansing the data, correlating the data, and then being creative with the analytics.
So this model, but it's more of a proactive model, but in this case, it was a fraud investigation that I worked on. And it was the head CFO, controller, whatever you want to call the individual, had committed this financial statement fraud and was booking a lot of journal entries trying to essentially hide what he had done. And the way it got stumbled upon was one month he forgot to do the entries or he's on vacation or whatever it was, and all of the amounts were sitting in one balance sheet account, which got stumbled up in consolidations to overseas to the corporate headquarters. And they noticed this balance sheet account with $15 million and they're like, "What is this?" And they found out that this guy had been committing financial statement fraud.
And I used my model that I built, and it has a lot of anti-fraud flags that it looks for, it uses kind of a risk methodology as well, so this is where the correlation comes in and getting creative like, "Okay, if you do identify something, how big of a deal is it? Is it a big deal or not? Within each one of these tests?" Because I knew that this guy, when he did admit to doing the fraud, he said that there were 10 accounts that he had affected. So because I knew that, I integrated that into the model, and again, this is where the human interaction comes in. So I'm like, "All right, let's just take my model. Let's create some additional fields here. Let's look for these 10 accounts and see what else we can find." When all was said and done, the model identified 15 additional accounts that he had affected that he had not admitted to. So in working with the internal auditors at the company, they were reconciling those 10 accounts thoroughly. They had to expand their scope to 25 accounts to add these additional accounts in there.
The other thing that was interesting was it was one of the first times, because the external auditors were on the hook for this fraud because they didn't identify it. They were concerned about it, they wanted to know, "All right, well how did you identify? What are you using?" So they actually tested my model, which was good because it had never been tested before. They had me tweak it a couple of times so that was interesting. But in the end, it ended up holding up. So I felt really good about it. I'm like, "Okay, this thing really does stand the test. It works. Somebody else tested it." So that was good to find out as well.
But the interesting side note to this case is, along with all the analytics, was this was one of those situations I guess you'd call effective altruism. So it was an interesting case where a CFO, controller, whatever, was doing this to make the division look not as bad as it actually was, better than it was, because it was about ready to potentially be sold off in this company. And so he was trying to, in essence, keep the division looking fairly profitable. Everyone's jobs were on the line. The lawyers were not thrilled about that. They didn't think that could be the case. Had me run a bunch of additional analytics to look through payroll, look through purchasing to see, "Okay, this guy had to have taken money out somehow, taken cash out of the company somehow." But with all the tests we ran, we really could not see that. We couldn't see bonuses tied to performance or anything like that.
So it truly was a case where he was just trying to keep the division afloat, basically, not get sold off from the company. But it was interesting to not only use the analytics, identify additional accounts that needed to be looked at, that he didn't admit to right away, maybe just forgot them because he had so many different entries he was trying to make, and then ultimately have the model kind of get tested and proven out as well.
Justin Burns:
So he didn't benefit at all other than getting to keep his job. It was nothing extra?
Matt Storlie:
Right. Nothing extra. And we spent a lot of time running a lot of different tests. And I was a forensic investigator at that time. I knew what kind of tests to run. I knew what to look for, to look for ghost employees, fictitious vendors. I did all kinds of tests and could not identify cash going out. And the lawyers still were not happy about it. They wanted something else to pin on this guy, but they did have him commit financial statement fraud and that was still an offense, and he still paid the price for it. But yeah, he did not. You think the fraud triangle, it's like, "All right, how's he benefiting from this?" But we really had a hard time improving that piece of it.
Justin Burns:
Wow.
Matt Storlie:
Yeah.
Justin Burns:
Wow. Yeah, that's surprising. You think, going back to the fraud triangle, certainly that rationalization of like, "Well, I'm keeping this thing afloat by my acts. Surely I'm entitled to something." But really it was, "I'll just keep everybody employed and keep this division going."
Matt Storlie:
Yeah, and you hear more and more I think about some of these effective altruism cases where it's for the benefit of others potentially kind of a thing. And there's not a lot of them out there, but this was one of them and it was from quite a while ago, so it's interesting.
Justin Burns:
Yeah, that one is really, really, really interesting. Wow. So Matt, this has been a great conversation. We've had a great podcast recording. Is there anything in particular that you want our listeners to walk away from this?
Matt Storlie:
Yeah, yeah. Thanks Justin. I've mentioned a few times over the course of this podcast, anytime I do a project or I do a presentation, which I do a lot of them, at the beginning I talk about three C's, and that's kind of the last thing I want everyone to remember because a there's lot of stuff during my presentation or whatever, there's a lot of information that I tackle. But cleanse. So anytime you're working on a data analytic project, make sure you're cleansing the data. Don't skip over that. That's incredibly important. Three C's, cleanse, correlate, we talked about correlating the data and make it a much more robust analysis. And then finally create, be creative. And so yeah, those are the three C's, cleanse, correlate, and create. So if you kind of apply those to whatever data analytic project you work on, I think you probably be successful if you stick with those three C's.
Justin Burns:
Yeah, if you stick with those three C's, your end product is going to be a lot better. Your data analytics are going to be a lot more useful.
Matt Storlie:
Absolutely. Definitely.
Justin Burns:
All right, so our listeners want to connect with you after the podcast. What is the best way for them to do that?
Matt Storlie:
You're welcome to go on LinkedIn. I've got a LinkedIn page. Look for Matt Storlie and you'll find me. Company is Wipfli that I work for and yeah, I'm happy to connect with you there. And then if we connect there, I'm happy to send an email address. My email address is matt.storlie at Wipfli, W-I-P-F-L-I.com. And either one of those ways is a great way to get a hold of me.
Justin Burns:
Perfect. We'll include those links in the show notes.
Matt Storlie:
Excellent.
Justin Burns:
Thank you for being our guest today. Thank you for talking with us.
Matt Storlie:
Thanks. Thank you for having me. It's been great. Happy to be a part of the Data Sleuth Podcast. Thanks for having me, Justin.
Justin Burns:
All right.
Leah Wietholter:
Thank you for listening to the Data Sleuth Podcast. If you enjoyed this episode, please leave us a review wherever you listen. The Data Sleuth Podcast is a production of Workman Forensics. To learn more about our investigation services and resources, please visit workmanforensics.com.