Enjoying the podcast? Don’t miss out on future episodes! Please hit that subscribe button on Apple, Spotify, or your favorite podcast platform to stay updated with our latest content. Thank you for your support!
Welcome to a new occasional series we are doing on the podcast called Fintech Revealed. This is where we take a deep dive into one specific topic with a couple of industry experts.
In our first episode we are focused on one of my favorite topics: cash flow underwriting. I invited two of the leading credit experts, Kevin Moss and Alex Johnson, to provide a well-rounded discussion on this topic highlighting both the benefits and challenges of cash flow underwriting. We covered so much territory in this conversation, so whether you are curious about cash flow underwriting or are a seasoned veteran, I am confident you will learn a great deal.
This episode is sponsored by Prism Data, the modern cash flow underwriting solution.
In this podcast you will learn:
- What cash flow underwriting is exactly.
- Why cash flow underwriting is not captured in the FICO score.
- How Kevin was using transaction data at Wells Fargo in the early 2000s.
- What has happened that has accelerated the use of cash flow data in underwriting.
- Why banks haven’t moved more quickly to implement cash flow underwriting.
- How subprime and prime consumers have been segregated in the current credit system.
- How cash flow data is different from alternative data.
- Why cash flow underwriting gets us back to the traditional way we used to be approved for credit.
- The real mission of cash flow underwriting.
- How well cash flow data predicts default risk and where it has the most value.
- What tokenization for connecting bank accounts will look like for borrowers.
- How streamlining the authorization process will supercharge cash flow underwriting.
- The types of lenders and products that are using cash flow underwriting today.
- The most promising use case for cash flow underwriting.
- How it can help the mortgage lending process.
- Kevin’s view on how cash flow underwriting will replace debt-to-income.
- Why regulators such as the CFPB like cash flow underwriting so much.
- How cash flow data creates a lending market with more fairness.
Read a transcription of our conversation below.
FINTECH REVEALED EPISODE NO. 1 – CASH FLOW UNDERWRITING
Peter Renton: Welcome to a brand new series we are doing on the podcast called Fintech Revealed. This is where we go deep on one topic with a couple of subject matter experts. We are kicking off the first episode in the series with one of my favorite topics, cash flow underwriting. Joining me on the show today is Kevin Moss, formerly with Wells Fargo and SoFi, and one of the most experienced credit executives in the country. Also joining us today is Alex Johnson, the founder of Fintech Takes, another credit expert and someone who has written extensively about cash flow underwriting over the last year or two.
But before we kick it off, here is a word from our sponsor, Prism Data.
The team at Prism Data has developed their cash flow underwriting solutions over nearly a decade and billions in credit originations, resulting in a mature, reliable compliant system with the most predictive power available in the market. It starts with the cash score, Prism’s simple three-digit credit score for automated cash flow underwriting. And if you want to go deeper, Prism offers thousands of data insights and attributes to power even the most sophisticated risk models and strategies. Learn more at prismdata.com.
Welcome to the show, Kevin and Alex.
AJ: Thank you.
KM: Thank you, Peter. Good to see you.
PR: Good to see you guys. So for the very small minority of the audience that don’t know who you guys are, why don’t you do a quick intro? Kevin, we’ll start with you.
KM: Happy to be here. Kevin Moss, former Chief Risk Officer for SoFi, 31-year banker, Chief Risk Officer at Wells Fargo in Consumer Lending. I do a lot of advising in three principal areas with startups and some banks, fraud, banking, and lending.
PR: Okay. Alex?
AJ: I don’t like to introduce myself after Kevin introduces himself because it’s embarrassing, but Alex Johnson, I write a newsletter called Fintech Takes that analyzes the intersection of finance, technology, and public policy, and I’m a lending nerd in my heart of hearts. So excited for this conversation.
PR: Yes. And you also have a wonderful podcast. So let’s kick it off. We’re talking about cash flow underwriting. Maybe we start off with how cash flow underwriting is different from traditional underwriting. Kevin, why don’t you take that one?
KM: First, you have to define what cash flow underwriting is. And I’m going to keep it a little bit narrow. People have checking accounts. Most people, in fact, 96 % of Americans have a checking account. Inside a checking account, you have debit transactions, ATM transactions, ACH transactions, checks, P2P between two people, merchant credits. And so through the use of deposit aggregation with companies like Plaid, Finicity, and MX, the availability of deposit transaction data has now become commercially available on demand, with people entering their credentials. And you typically can get up to two years of history. On average, it’s about a year that you get. And you get the raw transaction history associated with that. So cash flow underwriting is about taking that raw transaction history from the deposit account or other accounts and creating useful attributes or features that could be in models that summarizes the information value in that transaction history. So cash flow underwriting, which people have affectionately called it, you know, which has some type of implication around thinking about your inflows and your outflows and how much cash flow you have left over. Really, what it is, in practice, is the creation of attributes that have information value in predicting the probability of default and then using things like machine learning and other statistical methods to build models and strategies around how to underwrite credit in the same way that the FICO Score or Vantage Score does.
PR: Right.
KM: And it does a very, very good job, particularly in situations where somebody might have less or no credit history, which is, unfortunately, and we’ll talk about this later, quite a few reasons why there’s quite a big population of people that traditional credit underwriting doesn’t serve as well as cash flow underwriting could.
PR: Right. So, Alex, I was curious; I was reading your newsletter a while back. You had a big thing on cash flow underwriting and you were talking about the CEO of FICO from decades ago. He was talking about cash flow underwriting, and you commented that you thought they would not have been all that into it, but he liked it. You used to work at FICO, explain to me why cash flow underwriting is not part of the FICO score.
AM: So Larry Rosenberger, who was the CEO of FICO in the nineties, is on record as saying we would have used the data if we’d had access to it. And so I think to Kevin’s point, you know, using data on someone’s cash flow and how much money they have and how much money they spend on a recurring basis. That’s a very useful and very logical underwriting signal. It speaks very directly to ability to pay. It speaks really directly to the stability of income, all these factors that are really important for underwriting. The reason that it’s not captured in the FICO score simply because when the credit bureaus were first formed and then digitized in the 1960s, 1970s, and 1980s, and they consolidated into the three credit bureaus we have today, they were built specifically to share specific attributes on repayment behavior. The bureau file we know today was built to capture a very narrow, specific set of data. Originally, the retailers actually had more so than banks and other lenders. And then over time, banks and other lenders became more the big contributors to the credit bureaus. But the only reason that we don’t have this data in the FICO score is that the FICO score was developed before we had machine readable access to it. So there’s no real valid reason why. As Larry Rosenberger said, they would have used it if they could have, because it is a very valuable and useful signal that’s deeply complimentary to the data that the credit bureaus do have. And I know we’re going to talk about this, but we’ve seen even FICO try to adapt to this shift in cash flow underwriting that’s been driven to Kevin’s point by open banking and the ability for consumers to programmatically share access to their bank account data.
KM: And Peter, you know, one thing I want to say is as a banker for 31 years, I actually was using deposit transaction attributes as early as 2001 – 2002 for on-us deposit accounts. But open banking really opens up the opportunity for off-us where you may not have the deposit account to get the same type of data that you would have on your own customers.
PR: I just want to dig into that Kevin, since you brought it up. You worked at Wells Fargo, right?
KM: Correct.
PR: You’re saying you used the information in Wells Fargo customers checking accounts to underwrite. Is that what you’re saying?
KM: Yeah. I mean, without getting into the specifics of it. I used to think about what was a retail bank’s secret sauce. And it was that most of the time, the first relationship that someone established with a bank was opening a checking account.
PR: Right.
KM: And the fact that you had both sides of a consumer’s balance sheet inside that. Banks today, I would guess close to 70% of their checking accounts have some kind of recurring or direct deposit going into it as the primary checking account where their income goes, I could see both sides of a consumer’s balance sheet and make a really good assessment by just a very small number of attributes that summarizes a lot of good value that came out of that. Today, with Prism and Nova and Truv and a whole bunch of others, there are literally tens of thousands of attributes that have been built to summarize what the transaction history represents. But even back then, with a very simple set of, I had about a hundred attributes, I could increase the separation from a modeling standpoint of goods and bads by 10 to 15 % just by adding in some very basic deposit transaction history.
PR: So why aren’t we seeing that more? You don’t hear about banks using that data today. Were you unique in that, or are banks using some of those, just a small number of attributes?
KM: I think I was probably ahead of most other institutions in recognizing the value that deposit transaction data provided. I’d be surprised if more retail banks are not using some of that deposit data today in credit underwriting.
PR: Right.
KM: Now, there are a lot of issues around this that we’ll probably talk about as we get more into this, which you have to think about. Certainly, when you get into transaction data, sometimes, using transaction history is very hard for a consumer to understand why you took adverse action on them. So, you have to be very careful in the construction of the data that you use so that it’s understood and explainable. And also, you have to be careful about the fair lending implications. But having done this for more than 20 years, all of those things, if they’re thoughtfully considered, deposit transaction data can be a very valuable source of underwriting capability.
PR: So then Alex, turning to you then, what had to happen for cash flow underwriting to become more mainstream? And I wouldn’t say it is completely mainstream yet, but it’s certainly much more than it was four or five years ago. But what had to happen for that to really come into the consciousness of chief credit officers and risk officers?
AJ: Yeah, there’s been a couple of things. One is obviously the access to the data. And I think that’s by far the most important, right? And so the progress we’ve made in the U.S. to date around open banking and enabling consumers to share their data, that’s been huge, right? Until then, it was only someone like Kevin at Wells Fargo who had a lot of honest deposit data that you could use meaningfully at scale to help improve your underwriting advantage. But if you weren’t an incumbent, you didn’t have to use this data. I think, getting to the question you were asking before, a big reason a lot of banks didn’t is that it’d be nice to get a marginal lift and maybe approve a few more people or maybe be able to price some a little bit better. But the reality is banks were in a privileged position of being like the sea lions that got into the dam, where there were so many fish that they didn’t really have to worry too much about being efficient. And they could just do what they did. It took fintech to put some pressure on them, right? And so what happened was we had the development of open banking, which was market-led; it wasn’t driven by regulation. And one of the earliest sets of use cases around open banking in the US was cash flow underwriting. You had fintech companies, like Petal, which is a good example, that would underwrite borrowers, specifically those with limited or no credit history, using cash flow data as a way to do that. And so you had some very early companies demonstrate the validity of this as a way to attack the market, as a way to potentially peel off customers who otherwise would have just tried to get on the treadmill of working with banks and going through that whole process of getting credit when you don’t have credit and that sort of catch-22 that we’re all familiar with. They were peeling these customers off, saying, you don’t have a credit score, but that doesn’t mean you haven’t demonstrated financial behaviors that you could share with us that could be used to underwrite you. I’d say after open banking, fintech companies coming in and demonstrating the validity of this model at a small scale was really, really important. And that was, roughly, circa 2015 to 2020. Around that same time, you started getting regulatory interest in cash flow underwriting. Cash flow underwriting is one of those topics that’s just awesome for regulators. Most of the time, when regulators weigh in on a new trend or topic, they’ll not be totally on board with it. Because you like innovation as a regulator, but you don’t want to give full-throated support to innovation that could hurt consumers. Cash flow underwriting is unique in that regulators see it as a way that’s just going to be positive for the most part for consumers. If we can consider data beyond what the credit bureaus have, we can bring more people into the credit system. If there are people who have damaged credit profiles who are being pushed to more predatory forms of credit outside the mainstream, cash flow underwriting can bring them back into the market. It can lead to better pricing even for full-file consumers and a more efficient credit system. So regulators tend to love almost everything about cash flow underwriting. And there was a very concerted effort on the part of the CFPB, who just recently finalized their rule around open banking, the OCC who had a project called Project Reach that was focused on expanding access through cash flow underwriting. Regulators did a lot of outreach to the industry to encourage not just fintech companies but incumbent large banks to experiment with cash flow underwriting to improve access. And I would say it was the combination of infrastructure improvements, fintech competition, and regulatory ‘approval’ in soft quotation marks that allowed cash flow underwriting to emerge to the point it is now.
KM: I have some points of view, which some people may not like about the why part of this. So, first off, why didn’t banks move more quickly? Because deposit systems and credit systems operated in very different worlds. And it wasn’t easy, even in my prior banking life, to set up a process where you could get access to a file and have attributes calculated correctly and then available for credit underwriting and ultimately built into any kind of credit scores, for example. So, the separation of the way banks operate, deposits operate like in the retail bank, the community banking side, very different than where the credit card business typically operates. And getting those two together is, from a technology standpoint in a bank, because many banks are not on modern tech platforms, not the easiest thing. My point of view on the reason why regulators like this is there’s been so much talk over the years about how the traditional credit system leaves so many protected classes behind. And simply, these are roughly the numbers. We said 96 % of the people have a checking account. When it comes to becoming a banked customer, the checking account is the easiest one to qualify for. If you’re not a fraud, you can get a checking account. It’s just that simple. Ninety-six percent have some type of transaction history, let’s say, associated with them, or the majority of those. In the credit system, Alex brought up the subprime; the subprime operate in a different world, and there are separate credit files for the subprime customers, and the prime world typically doesn’t interact very much with the subprime credit files. And so, to some extent, I think when it comes to fair lending and fair access, I think the checking account, because of its deeper penetration, offers a broader outreach and opportunity for people who are in the full spectrum credit world to get access and consideration by lower priced, more prime or near-prime lenders. Whereas in the traditional credit system, and Alex and I have written about this, there are specialty files, and those files create separation, in my view. If you were designing a credit bureau system today, you would never design it based on having a separate file because this is payday lenders or buy here, pay here. You’d figure out a way. The current credit system hasn’t accommodated all of the different types of credit products out there, including fintech product innovation. Therefore, this segmentation has been created, which acts, in my view, as a barrier to entry into the prime credit spectrum. So, I think cash flow underwriting creates greater outreach and opportunity for those segments that travel in the world that are separate and segmented from the traditional credit system.
PR: Yeah, that’s a good point there, Kevin. That’s really interesting. Before we go any further though, I want to clarify something because you often see the term alternative data thrown around, and cash flow data is also thrown around. How is cash flow data different from alternative data?
AJ: In a broad sense one of the problems with alternative data is that alternative is just anything that’s outside the mainstream. Someone I talked to once said, “Alternative data is like regular data, but they’re rebelling from their parents, and they dyed their hair a crazy color, right? That’s alternative data.” In a traditional sense, the way we’ve defined alternative data is kind of going to Kevin’s point. It’s repayment data similar to what’s captured in the Bureau, but for whatever historical oddities were not captured by the Bureau, right? And so alternative data at various times has been considered telco data, utility data, rent data, and payday loan data. As Kevin illustrated before, the Bureaus have had an interesting push-and-pull relationship with alternative data over time because they obviously want to capture as much of this data as they possibly can. Over the years, a lot of that data has been brought into the core credit file. The credit bureaus have a lot more rent and utility data than they used to, but a lot of it also tends to get segmented off into separate files that have separate sets of attributes built on top of them and that aren’t integrated into the core credit file. So, there are some interesting infrastructure challenges to incorporating alternative data. Cash flow data, by contrast to Kevin’s point, gets almost back to the more original approach that we used to take in the old days to underwriting, where we really would underwrite you as a bank based on your checking account and your income and your employment and your general financial situation. So, in a lot of ways, it’s more core and more traditional to underwriting, even than a lot of forms of repayment data, and kind of gets us back to that core area. Again, the difference is that, unlike alternative forms of data, the cool thing about cash flow data is it’s largely being shared through consumer permission, right? And so I think the other aspect to remember to all of this is that it really does relate to how lenders think about using this and how it impacts the credit bureaus. We’ve never really had a large-scale source of insight used for underwriting that consumers have directly and granularly controlled, right? And so that is a pretty big distinction between this and other forms of alternative data where even if it’s like your rent or utility data, if someone is furnishing it, you as the consumer checked a box somewhere one time, and then it’s being aggregated and collected without you being aware of that happening, and someone is storing it, whether it’s a traditional credit bureau or one of these alternative credit bureaus that have popped up over the years. Consumer permission data, which is the backbone of cash flow underwriting, is totally different because it is something that you, as a consumer, are explicitly granting permission to use every time you’re applying for credit. In fact, under the new rule that the CFPB just finalized, that permissioning process is becoming even more granular, transparent, and restricted. There’s even less ability for lenders or other third parties to access that data to use it in other secondary ways. So it’s a big shift, not just in terms of the type of data, as we’ve been talking about, but also in terms of the mechanism by which that data is being shared.
KM: And, just to add a couple of things to what Alex said. Alex talked about how credit bureaus evolved. Well, in the deposit world, ChexSystems and Early Warning were the credit, the CRA entities for deposits, and they were negative bureaus. They existed where if someone had a foreclosure of a hundred bucks or more and hard principal balance, you’d report them to ChexSystems or Early Warning and they’d stay out there, I think, for five years. And then somebody would be blackballed from getting a checking account at another institution that happened to use those services. And frankly, neobanks evolved to originate and support those customers who got pushed out of the banking system because of Early Warning and ChexSystem, right? Because they advertise no credit checks. But the real value, in my opinion, of cash flow data is the positive information, just like the real value in credit bureau data is also the positive trade line history. Open banking has enabled people to access the full transaction, not just the negative stuff out there. Certainly, negative stuff helps find risky people. But the exciting thing about cash flow data is how it can create access to credit for fair price credit to a whole bunch of people now who might not have the same opportunities without it. The thin files, the immigrants, the people who are cashed in the cash-based society that operate in a different world than most of us do. To me, that’s the opportunity; the mission, if there is one associated with this, is to create greater access to fair price credit for a whole bunch of people who’ve had challenges through the traditional credit system.
[Sponsor message] PR: Are you looking for a way to say yes to more customers without taking on additional risk? Do you want to lend more profitably without sacrificing growth? How about both? Cash flow underwriting is a game changer in consumer lending and Prism data is the modern cash flow underwriting solution that makes it possible. It starts with the Cash Score, Prism’s simple three-digit credit score for automated cash flow underwriting. And if you want to go deeper, Prism offers thousands of data insights and attributes to power even the most sophisticated risk models and strategies. An increasing number of lenders, from top 10 credit card issuers to fast-growing fintechs, have adopted Prism’s solution to bring the value of cutting-edge cash flow underwriting to their businesses. Using Prism, clients have been able to achieve double-digit percentage improvement in their approval rates without taking on additional risk. With open banking, it’s easier than ever for consumers to share bank account data. Prism is the tool you need to put that data to work.
PR: The reason this is becoming popular is because it is predictive, right? So, how well does cash flow data predict default risk?
KM: Very well. And it depends on what your use case is. So, I’ll give you a whole bunch of different use cases. For someone who has an 800 plus FICO score, and a long credit history, it might make a marginal difference, but it’s not going to change your credit decision. For someone who has one trade line, who’s been open for six months, you can get a VantageScore. But the value of that score, in my experience, is very limited. And the reason is because it’s based on very, very limited information about that customer. Now, if I have a deposit account and a couple of years’ history on that, I can see what their income is, what’s coming in, and how their expenses are going out. At the end of the cycle, there’s still some cash left over every month. I don’t see return items or overdrafts. I see responsible use of that account. That tells me a lot more about that customer than I can ever learn from the credit bureau. So, to me, the value creation that this data has is much more beneficial as the value of the credit bureau data diminishes. And that could be subprime credit, where I could use this data to really pick out…what you find, and as you get into deeper subprime, the FICO scores and Vantage Scores, they don’t do a great job because most of those people look alike. But I can take two deposit transaction histories and find a lot more information value out of someone who is managing to pay their bills through their deposit account, their utilities, their rent, their mortgage, whatever it is, than I can through only the credit side of the balance sheet, where some of that stuff is reported and some of that stuff might be spun off into another credit file somewhere where I don’t even see it. I can see Buy Now Pay Later. I can see things in the deposit account that the credit bureau has no representation of, whether it’s because it’s in a separate file or it’s not reported at all. I can see rent to own. I can see payday. I can see that stuff. The people who have deep experience in building attributes out of the deposit transaction history can isolate those as separate attributes that can be used to really assess the risk of that customer. The bottom line is that if you can get the right attribute set coming out of that transaction history, you can have insights into people, much deeper insights about what they’re doing and what products they have than you can see in a credit file, particularly one that has very, very limited history on that customer. I’m extremely bullish on the opportunity for fairness, on the opportunity for credit access, and penetration into underserved market segments. Alex brought up a great point earlier: lenders are discouraged because it’s permission data and it creates friction in the funnel. Who are people using it for today? They’re looking at it for second looks, and for people they declined. They’re looking at it as an opportunity to offer better terms if they permit them to look at their deposit account. They’re looking at it for thin files and no credits and using it to underwrite subprime. But those segments are not the majority. They’re relying more on the traditional credit bureau data for the majority of their applicants. Now, once we get to the point where the opportunity to use an aggregation solution becomes tokenized, what that means is if I’m SoFi, and my customer has a checking account at Chase and wants to originate a loan with SoFi and SoFi wants to see the Chase checking history, the customer will be able to authorize the one-time use without putting in their banking credentials at SoFi. Right now, something like 50 % of the people, on average, opt out because they’re uncomfortable sharing banking credentials. I think once tokenization happens, we’ll see a much greater adoption of this by customers and also by lenders. And then, at that point, I think we may see deposit transaction data as a real competitor with the credit bureaus.
PR: Interesting.
KM: I think it’s a few years away, but five years from now, I think it’s very possible that deposit transaction data could be on par with credit bureau data.
PR: Right. Alex, what would that tokenization look like? What would that look like for the borrower’s journey?
AJ: So, as Kevin was saying, the concern on the part of lenders is that we just spent 20 years figuring out exactly how to do online account opening in a way that introduces the minimum amount of friction. And now you’re telling us we have to smash all our work with a baseball bat and start over? They don’t want to do it. For all of the great things about open banking, it’s not the most reliable infrastructure in the world today. Let’s just be honest about that. We’ve made a lot of progress, but if you talk to someone who builds on this as critical infrastructure, they will tell you it breaks all the time. Screen scraping is really unreliable. This aggregator just got in a fight with this big bank, and suddenly, that bank turned off their access to this data aggregator. This stuff happens all the time. It breaks, it doesn’t have perfect coverage. There’s a lot of concern that by putting that consumer permissioning step at the front of the borrower’s experience, the friction and the abandonment rate won’t be worth the increased ability to underwrite accurately, whether it’s subprime or full file customers. As Kevin was saying, what we’re hoping for, and I think the rule from the CFPB that’s now been finalized, should be a step in the right direction, is more seamless and reliable infrastructure to build on. I would say there are two things that will help improve us there: one is just the move from screen scraping and a sort of reluctant data provider community, meaning big banks, to one that’s more uniform, right? Under the new rule from the CFPB, everyone will have to stand up a developer interface, aka an API. That API will have to have uptime and performance standards that you can’t monkey around with. And so, the basic access to transactional deposit account data should become a lot more reliable. Coverage should be more even. Connectivity should be more solid. So, there won’t be as many of those concerns. And then the second layer of concern is even if that consumer permissioning step is more reliable and it doesn’t break as much; there’s still the natural friction of you asking me to plug in my username and my password. Do I remember it for this bank account? There are all kinds of things that are naturally friction-filled as a part of that experience. Consumers have gotten a lot more trusting in doing this over the last 10 to 20 years, but still some consumers are a little nervous, justifiably, about handing over their banking credentials, especially when they see scams and other things happening at a greater rate these days. There’s just a natural reluctance. And I will say we’ve been softly critical of the credit bureaus in a lot of ways. One thing that’s really great about the credit bureaus and the credit file is it’s super low friction to get access to. You plug in your name, last four of your social, and your date of birth, and you are away, and you don’t have to do anything else. Open banking is a little more friction-filled to interact with. So as Kevin is describing, the other big benefit is going to be streamlining that authorization process so that instead of having to plug in your username and password, and this is how it works in the UK, by the way, if you’re using open banking for a payment use case, the experience for the consumer is: Do you want to connect to your bank in order to streamline all of this. Yes, I would. Okay, great. We’re going to send a tokenized request to your bank. They will give you a notification that will pop up on your phone. Your mobile banking app on your phone pops up and says, hey, so and so is trying to get access to a thing. Is this you? You go, yes. You use your biometrics on your phone to authenticate it. So you don’t even have to log into your mobile banking app. It sends the tokenized approval back to the requester of the data, and away you go. You don’t have to type anything, and the whole process takes about a second and a half. If we get to that experience, we will have streamlined a lot of the friction. At that point, I agree with Kevin; lenders will put open banking at the top of their workflow. And we’ll be asking for that as a step at the very beginning because, absent anything else, you would want this data. As a consumer, the more data I can easily share that increases the odds of being approved or the more likely I am to get a better price, the happier I will be. There’s really no objection on the part of the consumer as long as we make it easy.
KM: Exactly.
PR: Let’s talk about the lenders today. What types of lenders use cash flow underwriting for second looks, shall we say? And what types of products are they using it for?
KM: I would say cash flow underwriting is deeply penetrated into deep subprime lending. I’ve worked with one or two of several well-known deep subprime lenders who rely on it exclusively. They won’t even pull a credit report. They rely only on the cash flow. The reason is they will use the deposit transaction history to estimate people’s incomes and establish their ability to pay. Then, they will use the inflows and outflows and create attribute summaries based on their models and policies. And someone might be borrowing a few hundred bucks, but it works very well. It works better than the subprime credit data does for them. So I would say that’s one area that it’s deeply penetrated. The second area is in the BNPL space; people are using it and playing with it. Again, it has a lot of potential because BNPL is like a non-prime, near-prime customer. I think in the long run, there’ll be broad adoption in BNPL, especially when tokenization takes place. And quite honestly, I think the place that is like in the top of the first inning in baseball terms, it’s ironic that we call it cash flow underwriting. The place that has one of the most promising opportunities is in measuring the ability to pay, which has not taken off broadly. People think about this as replacement attributes and a replacement for their credit score, but actually like debt to income, other than for a mortgage transaction, because in a mortgage, most of the time, both earners in the household will apply, you’ll get both incomes, you’ll dedupe the debts. But in a lot of consumer transactions, only one person applies, but you might have joint debts, for example, your mortgage. In a lot of cases, debt to income is wildly inaccurate when it comes to consumer credit underwriting; I think cash flow underwriting has the potential to completely replace the way consumer credit thinks about ability to pay. But that has not been the major focus for cash flow underwriting yet. So I think it’s to come, and I think it’s going to be a huge compliment to the probability of default modeling and attribute sets that are being created and are now being more widely used. But I think ability to pay is going to be one of the most interesting and helpful use cases because there are a lot of people who look like they have high debt to income, and it’s only because they’re being fully burdened on joint debts where the secondary income in the household, other than credit card, which is the only place you can use household income under the regs, they can’t get the benefit of the rest of the income in the household. I’m hugely bullish on ability to pay and the use of this data.
PR: Yep. Alex, who are you seeing as early adopters here?
AJ: It’s similar to Kevin. I think that where you see the adoption is where if you think of underwriting as a triangulation of ability to pay, willingness to pay, and an assessment of collateral, you see this getting early traction where it is weighted more on the ability to pay side. BNPL is a useful example because, if we’re talking about pay-in-4 BNPL, the product’s structure is designed to minimize any risk of unknowns around willingness to pay. If I’m loaning you $250 to buy a couple of pairs of tennis shoes, and it’s a six-week term where you’re going to pay 25% upfront as the first payment and then the remaining in three increments on two-week payment cycles, I’m not really concerned about if you have a 10-year long willingness to pay, managing all these different types of obligations. That doesn’t matter. In that type of product construct, I’m just narrowly interested in your ability to pay over the next six-week period. I’m looking very narrowly at the debt that you have at this moment and your total household income, and building off of a point Kevin made; this can be really, really important, especially for people who have lumpy income or uneven income.
KM: Like gig workers.
AJ: Yeah. That stuff is huge. I was looking at the FDIC’s report that they do every year on underbanked and unbanked households in the US; they just released the data for 2023. They asked about BNPL for the first time ever, and BNPL usage is actually much more common in households with lumpy, uneven, or unpredictable income. The reason for that is the product structure doesn’t require us to have a really good, stable, long-term income. As long as I can assess over the next six weeks, I will be in good shape to loan you this money. Where you’re seeing cash flow underwriting really take off is where the product construct is well suited to naturally be aligned with ability to pay. On the flip side, mortgage lending is on the opposite end of the spectrum. Mortgage lending involves a bunch of different components that go into it. There is absolutely going to be a desire for a long time on the part of mortgage lenders to look at your FICO score. They will also look at your long-term history of managing debt and long-term stability. However, in that context, I think what you’re going to see, and we’re already hearing some early examples of this, is income verification in mortgage lending. Regulators are comfortable with this, and that’s where consumer permission data and providing access to your cash flow data can be really useful. And if you’ve ever tried to get a mortgage having a non-W2 job.
KM: Or if you’re self-employed.
AJ: Right. It’s a nightmare. You’re trying to supply all these 1099s, and you’re giving them to your mortgage lender, and they’re trying to construct a picture of your income. The manual way of doing that in mortgage lending means you’ll probably not get a conforming mortgage. It will be hard for the lender to resell it to Fannie and Freddie. It’s a very, very difficult process. Cash flow underwriting can’t do everything in mortgage lending, but it can potentially solve a specific step in the mortgage lending process that will be useful for a very specific set of people in the population. Where you’ll see cash flow underwriting add value depends a lot on whether it is a product construct where cash flow underwriting can be the primary way you evaluate a consumer, or if it is a product construct where cash flow underwriting can add a lot of value for a specific step in the underwriting process.
KM: The GSEs, which control roughly half the market, plus FHA and VA, which are true government loans, how they go, that’s how the mortgage market goes. It’s going to be a little while. I know there is interest in cash flow data there, and they’re looking at it. But in the same way that it took how many years for Vantage to actually get adopted.
AJ: A lot of money and a lot of lobbying dollars.
KM: Right. And the version of FICO that people were using was from the mid-90s. The mortgage market moves slowly. It’s big. It’s our biggest credit product. Obviously, buying a home is a goal of many people and their families. It will take a while, but I believe 10 years from now, that cash flow will replace debt to income. I think that cash flow, if it’s done in the right way, does a better job of representing. You have a fuller set of debts and expenses and you can really measure what someone’s inflows and outflows are much more cleanly than just using people’s debts. I firmly believe that the most beneficial thing that will come out of cash flow data will be a better way of representing people’s ability to pay.
PR: Alex, I think you’ve had two conversations with the head of the CFPB in the last month, if I’m not mistaken.
AJ: I have.
PR: The CFPB seems to love cash flow underwriting. In fact, I was actually at the Philadelphia Fed when Director Chopra came up on stage on the day they announced the open banking rules. I think the very first use case that he talked about was cash flow underwriting. Why do you think the CFPB likes it so much?
AJ: Yeah, good question. My sense of that, having spoken to the director a couple of times and closely following all the rule-making that they’ve been doing around this, is that there are two things. I think that the CFPB is naturally suspicious of the credit bureaus and the orientation we have in the US around data privacy. One of the things that Director Chopra talks a lot about in relation to open banking is that we don’t have a national data privacy law in the US as they do in other countries. In the absence of that, they try, in all of the individual rules that they build that relate to consumers’ data, to build in a lot of control for consumers, to protect their privacy, to limit the scope of how their data is used, to be in control of how their data is used. One reason that they like cash flow underwriting is, unlike the credit bureaus and the broader data broker universe, which the CFPB is doing some separate rulemaking on to rein that universe of data brokers in and pull them in under the Fair Credit Reporting Act, they don’t like private companies hoovering up Americans’ data and using it for all kinds of purposes. While the credit bureaus are a very well-regulated corner of the data broker market, they still functionally are a data broker. There is more value extraction in many ways than value addition. One benefit that the CFPB sees in cash flow underwriting is it’s a way to put some power back in the hands of consumers, remove a little bit of power and control away from data brokers like the credit bureaus, balance out the scales, and create more data privacy. That’s one perspective. And then the other perspective, and I got to interview Director Chopra on stage at Money2020; he specifically wanted to talk about competition for the FICO score. The other aspect of this is it’s Director Chopra’s CFPB, and who knows, the CFPB tends to change a lot, from administration to administration, so we’ll see. At least under Chopra, the CFPB has had a very anti-monopolistic bent. They are very focused on any concentration or market power that is potentially being abused. Director Chopra talks a lot in his public comments about the FICO score, the fact that there’s no real, meaningful competition to the FICO score, the fact that FICO has been on an aggressive campaign of raising the price of the FICO score every six to nine months. I think those are examples to the CFPB of a lack of competition that is hurting the market and hurting consumers, because, at the end of the day, if mortgage lenders are paying more for the FICO score, they are passing those costs along to you in the form of higher origination fees. And that’s not a very transparent part of the market. That’s not a place where consumers shop around. It’s just a place where more comes out of our pockets. While Director Chopra is realistic about the fact that the FICO score provides a lot of utility as a common standard, and that’s not something that you can replace overnight, I think he’s optimistic that with cash flow underwriting over time, you will see a proliferation of lots of different types of scores and lots of different types of attributes that will weaken the central hold that the FICO score and the credit bureaus hold on the market right now.
KM: The only thing I would add to what Alex said is, it’s been a while, but if you read the work that FinRegLab did, there’s some belief by the regulators that the data is fairer from a responsible and fair lending standpoint also. It ties back to what I said earlier, which is a much deeper penetration of transaction data, with 96 % having a checking account, and you don’t have nearly 96 % having a deep enough credit file.
AJ: Yeah. The contrast between the CFPB’s approach to AI in lending versus cash flow underwriting tells you a lot about the fairness point in Kevin’s argument. Both theoretically have the potential to increase access to credit, which is great, but one is a black box that scares the hell out of the CFPB. The other is looking at your bank account data and underwriting you based on your ability to pay. One is clearly one that the CFPB feels much more comfortable with from a fairness perspective, and one is scary to the CFPB and others.
KM: Yeah. And plus, I think, in general, the history of CFPB being the primary regulator of the credit bureaus, there’s been a lot written about the accuracy of the credit bureau data, et cetera. And maybe even there’s a belief that the deposit transaction history may be a better, more accurate representation of what’s happening with a consumer as well.
PR: Well, that’s a good place to leave it. Kevin, really appreciate your time today. Alex, thank you so much. It’s always a pleasure talking with you guys and I really learned a lot in this conversation. Thanks a lot.
AJ: Thank you.
KM: Thank you for the chance, Peter. It’s always great getting together with my friend Alex and yourself.
PR: This episode has been brought to you by Prism Data, the modern cash flow underwriting solution, and the team behind the market-leading cash score. If you’re interested in cash flow underwriting, now is a great time to reach out about a backtest or proof of concept. Prism has just released the newest version of their cash score, Cash Score V4, which is raising the bar for cash flow underwriting solutions. The new score was trained on a large consortium data set representative of the whole US population, just like a FICO or VantageScore, and includes Prism’s cutting-edge data categorization and analytical insights. Lenders adopting Cash Score V4 can increase the number of loans they approve by an estimated 28 % without increasing risk, as compared to the prior version of the Cash Score. Prism’s latest release also includes tools for income verification, ability to pay assessment and more using open banking data. Check out prismdata.com today to learn more.