Risk Scoring and Fraud Detection, with Michele Tucci of Credolab | Soar Payments LLC
Michele Tucci of Credolab; Fraud Detection

Risk Scoring and Fraud Detection, with Michele Tucci of Credolab

In this episode, Michele Tucci, Chief Strategy Officer at Credolab joins us to talk about risk scoring, fraud detection, and how alternative data provides advantages over traditional data. If you’re a business owner or a risk professional you don’t want to miss this episode.

Payments & Fintech Insights In This Episode

  • Risk Scoring, Marketing Segmentation, and Fraud Detection 
  • What’s Alternative Data and how is it the future in Fintech
  • Credolab’s technology and how it’s proven in predicting behavioral data
  • How behavioral data could be turned into a risk assessment
  • The future of cookies in tech
  • And so much more!

Today’s Guest

Michele Tucci : Credolab

Credolab is a growing B2B SaaS Fintech that develops bank-grade digital scorecards built on mobile devices and online web behavioural metadata. Their pay-per-use solutions are available to banks and neobanks, digital lenders and BNPL players, insurance companies, and any industry at the intersection with financial services (e-commerce, ride-hailing apps, travel, retailers).

Featured on the Show

About PayPod

PayPod is the leading voice in the payments and fintech industry, covering payments, risk management and new technology. Host Jacob Hollabaugh interviews leaders who are shaping the payments and fintech world, as they discuss the latest developments in the payments and fintech industry.

Episode Transcript

Jacob: Welcome to PayPad. Payments Industry Podcast. Each week we’ll bring you in-depth conversations with leaders who are shaping the payments and fintech world from payment processing to risk management and from new technology to entirely new payment types. If you want to know what’s happening in the world of fintech and payments, you’re in the right place. Hello, everyone. Welcome to PayPod. I’m your host, Jacob Hollabaugh. And today on the show we are talking behavioral data analytics in relation to credit risk fraud, marketing, all areas of the credit world. Joining me to explore this topic is Michele Tucci, managing director in North and Latin America and chief strategy officer at Credolab, a growing B2B SaaS fintech that develops bank grade digital scorecards built on mobile devices and online web behavioral metadata. Their mission is to improve people’s lives by empowering every credit decision, giving easy access to fair credit for all. Michele, welcome to the show. Thank you so much for being here.

Michele: Oh, thank you for having me.

Jacob: Yeah, Let’s start with a little overview, if that’s okay with you. Credolab has three main solutions. It’s offering risk scoring, marketing, segmentation, fraud detection. Can you walk me through just what each of those does and an example of the type of merchant or company that would benefit most from using that solution?

Michele: Yeah, sure. As you mentioned, we use behavioral data to power every analytics, every calculation, every insight that we generate. And although historically we were born as an alternative risk scoring provider, we realized that the same data has quite relevant applications across also fraud detection and marketing segmentation areas. So we expanded the platform from risk to fraud and marketing. Now when it comes to risk, we focus on using behavioral data to predict the probability of somebody to miss a payment. And in the credit world, that’s something that is quite interesting, especially in the absence of credit bureau data In a fraud use case, we highlight signals that are in common to delinquent customers or customers that have displayed similar fraudulent behaviors in the past and we look at digital footprints, we look at behavioral outliers. So we look at behaviors that are anomalous in themselves or compared to other behaviors that we know have been confirmed as fraudulent behaviors in the past. And from a marketing standpoint, this is the most recent addition to our product suite. We have identified different ways to enable marketers to make better decisions, increase conversions, increase approval rates, increase product uptake, cross-selling, upselling. And so there are a couple of things that we can do. One is a estimate of the probability of somebody to uptake an offer. So basically looking at behavioral data to detect the intent to buy a product, intent to apply for a credit card. And the most recent one is we use the same data to predict your personality type without asking questions, without having to go through a psychographic assessment or a psychographic questionnaire that usually adds friction and gets people upset and then eventually they leave whatever onboarding they were going through.

Jacob: Yeah, fascinating. Let’s dive into some questions specifically around the data sets that you’re using for They’re all powered by the data that you’ve listed out, and I’ve seen it termed alternative data in a lot of places. Can you explain to me the difference in this alternative data and the data sets you’re pulling from and where you’re getting it from versus what I guess would you call traditional data or where we’ve been pulling from previously and why these types of data sets have the potential to be so much more powerful?

Michele: Yeah. So alternative data, The easiest way to define it is everything that is not traditional in the credit industry. Traditional data is credit bureau data, past histories of repayments. Most recently, some of the bureaus expanded into using utility payments kind of data. The likes of MasterCard acquiring Vinicity, for instance, brought that type of transactional data into the picture as well for credit worthiness assessment. The reality is that most of these traditional data requires you to have access to financial products, to financial services. So it’s like a dog chasing its own tail. If you don’t have a credit bureau score, you don’t get access to data. But without a credit card, without a bank account, you don’t generate credit data. And so it’s like a self-fulfilling prophecy to some extent. So we identified a way to access data that people generate every day by simply using their smartphone or by simply interacting with a web page. So it’s data that we access through our preparatory Technology. But I think I’m digressing a bit. The question was about the alternative data. So.

Jacob: Yeah. And so it sounds like it’s the data is only new to this industry. This data maybe has been available or used by many others and now you’re widening the scope, so to speak, of formerly they use this little slice of knowledge about a consumer to base this one decision off. And like you said, it’s chasing their own tail. They’re using their own when they’ve interacted with them before, more or less. And now you’re just widening the scope to say there’s so much more that we can know and learn about a person to make these decisions better. Is that am I kind of making sense of it?

Michele: That is correct.

Jacob: Okay. Let me ask you this then. You recently wrote, I was reading a blog post you did for Finextra that said, quote, The scope and intensity of alternative data use are expanding. Its application for business purpose still lacks a deep understanding of its nature. What do you think is causing that lack of deep understanding? Is it simply just the time that’s necessary to kind of close that gap and understanding, or is there some shift in people in the markets and those using it? That needs to happen?

Michele: I think it’s a combination of the two plus some initial skepticism from people thinking can this really work? So when we brought our solution into the North American market, for instance, after six years of being present in Asia across emerging markets of LatAm and Africa, the main question in the US is we need to do a proof of concept. Now we need to test if the data works and we know the data works. So because we’ve been doing it for six years and if a bank in Singapore is willing to use our technology, why not a bank in the US? And so there is that level of skepticism and that is slowing down adoption. And the other element could be related to familiarity with data. So when Experian launched Experian Boost a few years ago, the idea of bringing transactional data, connecting your bank account with your bureau to increase your credit score was revolutionary from a US standpoint. Guess what? In Europe there was a regulation already many years ago called Psd2, which forced the bank to open up pipes to anybody who had a right to access that data. So skepticism, familiarity, perhaps a mindset. I would say also. No, it’s I’m going to try what I know works. And then if I have a really big problem and there is no other solution, perhaps I’ll try this. One other thing as well. Regulatory framework doesn’t necessarily help. And although we don’t process personal data, we cannot build bias into our assessment because we don’t profile people racially or by age, gender. We don’t do any of such things. Still, the Fair Credit Reporting Act, especially in the financial services industry, is one big regulation that concerns many. And in particular, how can you decline a customer based on a behavioral assessment of that customer rather than a credit bureau score that has reason codes? And maybe I’m going a bit technical, but to give you an idea, it’s okay to decline a customer. If he missed three payments of the credit card payments, it’s not okay, according to the FCRA, to decline a customer if she took too many selfies.

Jacob: Yeah, right.

Michele: Absolutely. And we agree with that. We appreciate it. We see why that is important to be taken into consideration. However, there are so many different layers of assessment where Credolab becomes only one of them, especially in the risk space and fraud detection space. So regulatory framework could be also one of the reasons why, especially banks, are particularly concerned about using alternative data that today is not regulated.

Jacob: Which what markets were some of your earliest adopters or that you started in and then coming here into the North American market, have who have been your first early adopters or who are the main integrations you’re trying to get set up to get into this market?

Michele: So Credolab was born in 2016, January 2016 in Singapore, and what we identified at that time was a very large addressable market of about 660 million people, of which about 78% had no access to formal credit. They were either unbanked or credit bureaus had penetration below 30%. So only three out of ten applicants could actually be scored. Traditional means. But guess what? Smartphone penetration was already high in Southeast Asia back seven years ago. So our very first country where we found the very first client was Indonesia. Indonesia is 230 million people, of which less than 60 million have a bank account. When we started. So massive opportunity, but penetration of smartphones was above 80%. So right there we had access to data, behavioral data that we with our technology could have been turned into a risk assessment.

Jacob: Yeah. Wow. And with those early markets, you kind of just referenced it. But one term I saw you used for some of that market share, the consumer base that was there was the credit invisibles, and that you’re opening up these services to this wider range. And that obviously helps the companies that are, you know, the merchants that are putting these services out there, but also inevitably then helps that consumer to who wouldn’t have had access to this before. Are there any other consumer groups like that that are going to be really benefit from these types of services being rolled out And specifically in the North American market, is it still just capturing that kind of credit, invisible group, or are there other types of consumer groups that are going to see massive benefits from adoption of these types of services?

Michele: It’s a good point because we all use smartphones, we all access websites. So the same level, same kind, same granular information is available for any customer. And so Credit Invisibles in the US alone, we have about 54 million credit invisibles, and that number comes from FICO itself. Then you have thin files. So those that have credit bureau scores between 500 and 650, which can also benefit from a technology like ours. And then you have thick files, those for which the credit bureau knows everything. And arguably they have a good standing high credit bureau credit score. And but guess what, Jacob? Imagine you and I, we have the same salary, same credit bureau score. However, I came to America only a year ago, so I’m a thin file. No matter how good is my standing. I don’t have a long enough history with the bureaus to prove my credit worthiness, so I may be rejected because of that. But imagine the other case where we have the same salary, same credit bureau score. Everything looks equal. And from an affordability point of view, you want to repay, but from a willingness to repay, I don’t. So with our technology, we can discriminate with a behavioral assessment between the ability to repay and willingness to repay. So even for the top 750 and above credit score, we can still add value to a lender, to a bank. And this has been proven by TransUnion that analyzed our data and confirmed that we do bring value across the entire value chain, not just Credit invisibles or team files.

Jacob: Yeah, absolutely. Yeah. It’s you’re opening up that whole new market, but you’re also fine tuning the market that currently exists of giving them way more information to go off of and way more accurate to be able to make those predictions. Because I certainly, as someone who’s a couple decades into having a credit score know it doesn’t always reflect exactly you know, the data set they’re pulling from. I’ve had great scores and not as good of scores at different times, and it hasn’t always really, truly told the picture of whether or not someone should have been lending me anything at all. One of the big selling points from going back one last time to the actual data itself is the fact that no personal data is collected. Can you explain what are maybe a couple examples of the exact data that is being pulled and how it’s even possible that there’s no personal data in there and why that is kind of the powerful, attractive feature that it is.

Michele: So the whole technology was built with data protection in mind really from the ground up. Even when we launched in Indonesia seven years ago, we at that time already we didn’t process personal data. So there are different reasons for that. One is reputational risks for banks or lenders buy now, pay later players that work with Credolab. We don’t want them to feel exposed to reputational risks because of a third party vendor or a partner, but also from an analytical point of view, we know that processing personal data or metadata like we do today, the uplift that we get by processing personal data. On the models doesn’t justify the reputational risk of processing personal data. And also with the GDPR in Europe or the CCPA, the California Consumer Protection Act in in California, we are future ready. Really. We don’t need to have cookies. We don’t need to process personal data. We don’t need to receive the consent of the user to receive their email or phone number because we don’t need. We are embedded in the mobile app of our clients, in the websites of our clients, and that’s how we collect first party data that is consented by the user, permissioned by the user, they personalized by our client and anonymized at the source. So there are four layers of protection that cannot be questioned. I mean, we had third party auditors also that confirmed that our technology does what we claim it does.

Jacob: Yeah, that’s absolutely amazing and fascinating. Let’s turn our attention to a few trends and kind of newer sub industries within the lending and finance world. We’ve seen in recent years the rise of Bnpl or buy now pay later and earned wage access. Those are two areas where obviously your solutions are going to have a big impact, new kind of areas within the lending space, but they’re also potentially industries that need a lot of help. We’ve seen thus far they have performed even worse than traditional lending places would as far as the accuracy of who they’re giving money to, when and how much. Can you talk to me about some of the troubles that Bnpl and have presented thus far? And then if you think a company like Credolab and its solutions could get the results in these industries that get them to a sustainable place, because currently, at least from my less limited than yourself, certainly knowledge, they don’t seem to be in a sustainable place a few years into their proliferation.

Michele: First of all, we do work with a number of these players in the buy Now pay later and early wage access vertical. And the problem that we solve for them is the same that we do solve for a bank. It is lack of data. So what I have seen, not necessarily with our clients but across the industry as a whole, there is more attention to the marketing side of the buy now pay later experience than the risk side. And some of these companies don’t even have or didn’t use to have a risk function, which is surprising to me because you are lending money. So unless you trust fully, the customers will come knock at your door and give you the money back. It’s not going to happen. So lenders to me are in the business of collecting money, not disbursing funds. Yeah.

Jacob: It’s about the last business I would think would be avoiding any risk assessment is about the riskiest thing is our business is giving money out right and hoping for it back right.

Michele: So early wage access they kind of built the entire business model on a safer ground, which is I build a relationship with the employer, so I know the history of that employee. And underwriting models are built by ingesting that kind of data that otherwise wouldn’t be available or necessarily available to them. However, there is also a natural death percentage of businesses. There is a natural turnover of people that leave their employer going somewhere else. So it’s not bulletproof either. And the moment interest rates went up, cost of funding went up. You see who’s swimming naked when the tide goes out. So these trend upward trend expose most of these players. Not all in fairness to them, to the poor underwriting processes or even anti-fraud processes that they had in place or maybe not sophisticated enough. Yeah.

Jacob: Wow. Very cool. It certainly seems like there were wonderful ideas when they’ve come on and they definitely have had their share of issues, but it would seem just like traditional lending that having that data set, increasing the accuracy and everything is going to make them viable options for the long term. Let’s move to maybe the hottest topic of the year at this point, which is AI. It’s inevitably brought up on every episode of this podcast we do at this point because it’s everywhere. Everyone’s talking about it, and even if it’s been around a lot longer in the world of tech and fintech, especially now, everyone, no matter where you go, what industry you’re in, they’re talking about it. So how are AI and machine learning technologies being used by Credolab If they are currently in, where do you see use of those types of machines growing in the future for the company?

Michele: So we do not use AI, we do use machine learning and our entire data modeling pipeline. So the way we build scores, the way we assess credit. Worthiness or propensity or intent is all rooted in machine learning algorithms. We don’t use AI because we want to explain the outcome of our calculations. And with AI, when you build a neural network, for instance, you may have a higher predictive power. However, you will not know how to explain it. And when you come to a highly regulated market like the US, that’s a key requirement. So for us, we can explain the outcome, at least statistically. I made the example earlier that if you take too many selfies, you are probably going to be rejected. And that may be explained not just statistically because of our technology, but also behaviorally. You know, you tend to be more narcissist, perhaps less good as a repair than somebody who doesn’t take selfies. So we made a conscious decision, not to use AI. However, I played with ChatGBT a couple of weeks ago and I asked Chatgpt, what do you think about Credolab?

Jacob: I saw this. Yeah, this was very interesting.

Michele: And it was a funny kind of experiment. And surprisingly enough, Chatgpt identified the four main pushbacks that we receive about data privacy. Is it really true that you don’t process personal data? How do you explain the score and the regulatory compliance? So it was surprising for me to see how accurate this thing actually was.

Jacob: Yeah. Well, I can also say as someone who for yourself and your company and every company that I speak with, I now also in my research, one of the first things I’ll do is ask a chatgpt or other type of software of Can you build me a quick profile of this company, this person? And it’s really helpful and is really a lot quicker path to right now in its early days. Still double checking all of that like okay when were you pulling that from and so everything accurate but it is super helpful related to AI in a way another big topic in the fintech world and that I believe I’ve seen you write and or speak about in a little bit of my research was everything as a service or SaaS as it’s attempting to be said. I think I just personally stick with saying the full thing, everything. As a service. Do you see everything as a service as the new normal for most companies to just be outsourcing as much as possible? And then the second part of that and where kind of AI comes into it is I believe I saw you discuss or write about at one point the kind of tipping point for companies that would be best to outsource, but at scale they could start to be the size where building internally becomes more accessible to them and maybe the better option. Do you see things like AI or machine learning and tools of that nature, making it easy and cheaper to build internal tools that might kind of slide and tip that equation of when a company hits the size and scale, where instead of going to a credit lab, instead of going to all of these different players, all of these different services, now, they could be building things internally a little easier and kind of tip that scale, so to speak.

Michele: It’s the old question of buy versus build. Why would you buy a solution when you can build it in-house and vice versa? Why would you build it in-house when you can go perhaps faster if you buy it from outside? So every business is different and the things that can be outsourced are those that are not core to the business itself. So and that’s how, for instance, to stick to the credit industry. I don’t think any lender should go about decisions without credit bureau data. If it is available. So can they build an internal score? Yes, they should. Most of them have. Can they make it better by ingesting other scores from other providers? That’s also highly recommendable. So at that point, the only question becomes how much should I pay for every incremental data set, every incremental score that brings marginal improvement in the way I do business. So in the same way for other activities, I is welcomed whenever it can improve the way you do business today to avoid manual tasks or to improve the accuracy of some calculations. But in general there is a singularity point where it will take over the world and we’ll be dead. But before we get to that, I think we have a long way still. And you had also one other part of the question, right?

Jacob: Well, just that, you know, Does it change for you at all the kind of equation of how large a company would need to be? Because to switch from that, we should be buying versus we should be building or, you know, I totally take your point that regardless of even if you are building, you should also be doing a little buying if there’s additional things to make things better. But just the idea that normally a company would have to get to a pretty massive scale to say, okay, we’ve got these 12 tools we’re purchasing from all different spectrums to run most of the parts of our business really lean. But now building certain things, maybe not your credit scoring system, but other things might be come much easier or much cheaper or less time or manual intensive to where that equation just shifts a little bit. The size, the scope, the scale of a company would need to be before it starts to think more about the build versus the buy.

Michele: Now, I think size will definitely play a role. If you are a small company of ten people, there is only as much as you can do and usually you want those resources to focus on running the business, on getting things done for the core business. The question then of what to buy versus to develop in-house. I think there is a lot of influence that comes from the venture capital world. So we have received some pushback, for instance, from Buy Now Pay Later players where in a time where money was free and VCs were lining up to give money without even almost making any due diligence of the business. And so they said, No, we’re good, right? So they didn’t need to invest on making risk better, more accurate risk assessment. And VCs were pushing them to deploy money elsewhere. So I think the question is also how does a CEO run his own business, her own business with that business at heart, rather than pleasing investors that have different agendas and they may not even be close to you for the entire life of your company. So I think it’s a bigger discussion there.

Jacob: Yeah, and certainly times have changed very quickly in that money is not so free flowing or readily available these days and definitely causing a lot of changes for a lot of different folks. Any other market or industry trends you’re keeping an eye on in 2023 and beyond that is kind of top of mind for you and Credolab to kind of stay on the forefront of here in the near future.

Michele: Oh, there is one that is particularly dear to Credolab and is the Cookieless world. So cookies are going away. And as you mentioned earlier, people are increasingly more particular about who they give access to their data to. So besides Facebook, Instagram, TikTok, that seems to be a given. We all accept that they will mine our data and give nothing in exchange. Whenever we do subscribe to a different service, download a new app, we tend to be a lot more peculiar about, okay, yes, I trust this brand or not, but in general, cookies are going away. So from a marketer, a marketing manager, for instance, who’s pushing ads, the inability of measuring who is seeing your ads or how many times the same person is seeing one particular banner, it will affect the return on marketing investment. It will affect the return on ad spend. So we have one component of our technology actually is able to identify returning devices without processing personal data. So at least at a device level, I believe there is a way for us to play a role in this cookieless world with full privacy protection and in full permissioned mode from the user. That’s something that I’m quite keen to observe. The other trend that we see is decentralization. Even perhaps if we had spoken in December, I would have said crypto lending could have been a big trend for this year, I don’t think anymore, you know, with FTX and I don’t think that’s that’s going to be anytime soon again.

Jacob: Yeah, well, I don’t know that it’s gone. Certainly the moment came in past, at least for the time being, for it. It’s fascinating. On the Cookieless front for disclosure, my married to a woman who runs digital ad does digital advertising and digital marketing and everything. And so thinking about the changes in that definitely hit home very directly for me. I wonder from a growth perspective for Credolab is the marketing side of things where you see the biggest growth opportunity maybe over like the next decade in how your services can expand in the different places you can be going into.

Michele: We calculated the TAM, the total addressable market of marketing, the marketing space, and it’s. Double, then the combined time of risk and fraud. So risk still is our core product. That’s our core expertise. And, you know, every time you innovate, you need to go by adjacencies. So we looked at what to do next from risk fraud was the natural next step. And between fraud and marketing, also there is an adjacency there because of ad fraud, because of bots, because of a number of actors that are gaming the system. And guess what? We can use the same data, same technology to address those pain points as well.

Jacob: Yeah, amazing. Well, Michele, this has been an absolute pleasure. I can’t thank you enough for all the time and knowledge you’ve shared. Before we go, for anyone listening who would be interested in learning more about Credolab or who’d like to follow along you, where would be the best place for people to go to do that?

Michele: So there is credolab.com. I’m not a very social person from a social media standpoint, so my LinkedIn profile is there and perhaps if you want to connect to Tucci@credolab.com.

Jacob: Wonderful, awesome. Well we will link all of that below. It’s been fantastic. I can’t thank you enough for joining us on PayPod and hope to talk to you again soon.

Michele: Likewise. Take care.

Jacob: If you enjoyed this episode and want to hear more, head on over to Soarpay.com/podcast to subscribe on your podcast listening platform of choice. That’s s o a r p a y dot com slash podcast.