Facing Credit: The Credit Score After FICO

0:00 Hello, and welcome back to the Fintech Takes podcast. I have another episode of our credit and credit risk focused series Facing Credit. Um this is an episode I've wanted to record for a while. As I've written about in the newsletter, as I've talked about on the podcast, um I think the credit scoring market in the US has changed more in the last year than it had changed in the prior 30 years. And a lot of that has to do with the sort of degradation of FICO's monopoly on the credit scoring space, but a lot of it has to do with a new innovations that have come into the market, cash flow data, cash flow underwriting, uh the development of proprietary in-house models like Block's Cash App score, new regulatory uh initiatives, uh regulators taking a step back and deregulating in certain spaces around things like fair lending.

0:54 So much innovation, so much uh sort of regulatory uncertainty. And that has left us in a place where it's really uncertain. What credit score should you use if you're a lender? Should you build your own credit score? What data sources should you look at? If you're a consumer, do you know how your creditworthiness is being determined? And do you have an ability to understand how you can improve it? These are the questions I wanted to talk about, and there was no one better to talk about them with than Rich Franks. Uh Rich is a fintech advisor and consultant with 20-plus years of history in credit risk, working on both the bank side and the fintech side. Rich has seen everything, he knows everything, and he's been thinking deeply about a lot of these same questions. So it was a delight to have Rich come on. We talked about uh the sort of emergence of cash flow data, the blooming of a thousand cash flow scores, and what that market will eventually look like when it consolidates down. We talked about the sort of end of FICO's de facto monopoly over the space and what FICO and the credit bureaus are going to do next. We of course talked about Block's Cash App score, which I am personally obsessed with, and even got a little AI talk in at the very end. So, uh great episode, great discussion, very grateful to Rich for coming on. So, without further ado, here is another episode of Facing Credit.

2:31 Okay, uh Rich, thanks so much for coming on. Glad to be here, Alex. I'm delighted to have you. This is uh a conversation I've wanted to have for a while. You've been on the podcast before, but um for our credit-focused discussions that we have uh in this little show that we do in the podcast feed, I have so many questions uh that have been building up, I think over the last gosh, like 12 to 18 months. Um specifically around the credit scoring space. Um and I guess the way I would frame it to you, Rich, and I would love your response to this is I think credit scores as an industry uh have changed more in the last year to 18 months than they have in the last 35, 40 years before that. And uh that's kind of a crazy thing to say, but you know, if you think about kind of where we've been historically, and I I used to work at FICO, so I I am intimately acquainted with this history.

3:29 The credit scoring space has been, I think, largely defined by one score that everyone uses, that has sort of a de facto monopoly on uh every part of the ecosystem, right? Not just the score that lenders buy, that's an input into their decisioning process, but also the score that they buy for account management and servicing use cases, also the score that they buy in order to resell their loans on the secondary market. Also, the score that they market to consumers that teaches consumers how to think about their creditworthiness and make decisions about their financial future, the score that regulators understand and sort of index their understanding of credit risk and the performance of the market to, everything is this one number. And um you know, it's funny I I wrote a piece a few years ago in which I sort of predicted the eventual end of the FICO score. Which is not to say the death of the FICO score, but just the end of its sort of reign as the score, the one score to rule them all.

4:26 And when I published it, uh everyone thought I was like out of my mind. And um fair enough because FICO had been on this incredible run for many, many years, decades, uh of just dominance. And I think, you know, people especially in fintech had seen a lot of like "We're the new FICO score. We're building this alternative score. We're doing this thing with machine learning." And then all of it would just sort of crash into this like wall that just could not be overcome.

4:52 And so it was a little bit of a uh uh controversial take at the time, but what I was looking at was the sort of slow accumulation of a bunch of different trends. And I think 2025 and early 2026 is where we've seen those trends sort of finally really manifest themselves, right? So, obviously we saw um everything happen with the FHFA and uh Bill Pulte uh from a mortgage score perspective. Um In some senses we knew that this was coming. And one of the things that I had uh written about uh was this idea that FICO was very sort of quietly raising the price of the FICO score um ever since uh the Credit Score Competition Act was passed and we knew that eventually Vantage would have a crack at the mortgage market as well.

5:41 And so the FICO score's price goes up, Bill Pulte kind of pulls the pin on the grenade, suddenly now we have competition in the uh mortgage credit scoring market, which is one of the largest and most lucrative parts of the overall credit scoring market. So, we had that happen. Obviously, you've been very, very close to everything happening in the world of cash flow data and cash flow underwriting. So, we've seen this like I mean, Rich, we used to talk about alternative data as like alternative data like social media data for big data credit scores and things like that. This is, in some ways, very different than that. This is like bread and butter cash flow data that any lender would see as valuable and want to use, and now it's programmatically available. So, it's getting built into scores.

6:24 We've even seen some really interesting stuff recently with like specific lenders building their own internal scores, which is a thing that's been happening for a long time, but now maybe those scores will be made available to other lenders in the marketplace. So, I just want to pause and let you reflect on that last year, 18 months. Like, what have you noticed, and how are lenders maybe reacting differently than they have in the past to all this innovation?

6:51 Yeah. Well, first of all, Alex, this is really fun because I think that you know this topic at least as well or better than than I do. So, it's going to be a really fun conversation. You know, I'll start back from my my own perspective. You know, I started my career the last week of 2004. And my first project as a data scientist and and you know, a big credit card company was to go through a TransUnion assessment and look at all the commercial scores that were available and figure out what the best option was.

7:19 And then also, beyond that, to go, you know, build some some custom models off of it, right? Okay, so, I do this exercise, you know, it's you know, a few months of work, whatever. And I come back and I say, we should be using Score X. Maybe you know, I you've probably heard of Score X. A lot of people probably haven't. Yeah, yeah. Describe it for the audience who doesn't know. Yeah. Yeah, no, but Score X was it was a it was a FICO competitor made by this this team within Experian.

7:44 And so it was kind of like a you know, kind of an Experian only version of what Vantage is today. It was just an alternative approach, but it was amazing. It did so well. Um, and and I came in, you know, 22 years old and said like we're going to change this you know, number X credit card company. We're going to change our score into Score X and everyone laughed at me and said, you know, we can't seriously not use FICO.

8:08 Um, and that was I think, you know, to your point. Like that was the world for a long time. It was like we can build our own scores. We can use unique data, unique functional forms, unique whatever, but at the end of the day FICO needs to be part of the thing. And to replace FICO with something like, you know, at that time it was Score X. Now it would be, you know, Vantage or many other things.

8:28 It was just unheard of. Um, and so and and I think that that was that's been the world for many many many years even as other things have happened. So I think there's been one you know, one big shift that's happened is a combination of machine learning plus computing power means that now if I'm building a credit score for my own business, I can build, you know, a 200 actually I could build a 2000 variable machine learning model.

8:56 Probably shouldn't. Some people are. Um, but but you know, I could build a 2000 variable machine learning model relatively easily tuned to my portfolio, tuned to my population, tuned to my product and update it every 12 months I want to, right? And so I think that's that's the first thing that happened. That happened, it unlocked people's ability to really improve internal risk scoring. Yeah. But still we added FICO, right? Maybe FICO's in that model. Maybe maybe FICO was a matrix at the end of it, whatever. But I you know, I think that that just it's just kind of we just accepted this is the way the world works.

9:28 Um, so I think that was that was the first big shift. I think another shift, now I'm going to go a little bit out of order here, but >> Please, yeah. I think to your point cash flow underwriting it's like this new, really exciting, really orthogonal dimension. And if you look at everyone who builds scores for cash-flow underwriting, they all kind of say the same thing, which is that you get um by adding a cash-flow score to an underwriting decision that's based on bureau data, you get about a 30% lift, give or take, on top of the original decision, right? Which is absolutely incredible. Now, there's >> That's huge. That's huge, right? Just to like put it in context, like you spent most of your career looking for like It's like um It's like Moneyball, right?

10:05 Where it's like, "Can we find someone who gets on base just like, you know, 50 basis points more?" Yes, oh my gosh, this is amazing. Like we found this like You don't just find like Babe Ruth just sitting on the bench just like waiting to to come off and like play, right? Like that's 30% increase. It's It's totally magnitude higher than anything we've seen. Yeah. That is Yeah, that I mean, that would be like if you were a data scientist who unlocked that on your own, you know, you're just, you know, mining your company's data and you find 30%. Like that would be a career-defining moment for you. Totally.

10:35 That's a That's a Yeah, I mean, this is huge. Um and uh yeah, and so I think that all of a sudden everyone's looking at that differently and saying like, "Okay, wait a minute. Um how do we incorporate this into a decision?" I have some strong opinions here, but you know, there's, you know, you can do um a standalone score that you buy, you can do a standalone score that you build off of attributes that you buy, you can build your own attributes, you can you can buy a composite score that includes cash-flow data and bureau data together, you know? So so many different ways to do it, but I think that's, you know, that that chaos of coalescing around the right answer in cash-flow underwriting is is kind of opening people's minds a little bit, too, to hey, there's more here. Yeah. Um I think then the You know, you talked about alternative data, you know, I think like the social media data, maybe outside the US, uh has gotten some traction, maybe a little bit less here, but um but certainly there's a lot of dimensions beyond the three traditional credit bureaus that lenders are saying, "How do I incorporate this into my into my models?" And so that that's unlocking people's minds. And then finally, to your point, Bill Pulte comes in and says, "Actually, we don't love the FICO monopoly."

11:38 Which to your point has gotten way more expensive over the last it's actually ironic. I think like 2018 was when the legislation came that said we can do other stuff. And that's also interestingly the moment that you know, FICO's prices started going up dramatically. >> It's such a such a coincidence. Yeah, that those things happened at the same time. And it's it's funny because I think that you know, FICO is incredibly smart, right? I mean like no one no one maintains a business like this for decades without really having a great sense of like the market, the dynamics, whatever. And I think the story that was under appreciated from 2018 to now was that FICO kind of knew the game was up and just didn't tell anyone, right? And so they're like, "Oh, we're going to raise the score. The These are just price increases that we haven't passed on for decades." And like the question that anyone paying attention should have asked was, "Well, why didn't you pass on those price increases for decades and why are you doing it now so aggressively?" And of course the answer is because they knew eventually this was going to happen. Now, the Bill Pulte I I use the analogy of pulling the pin on the grenade like he went early and he went aggressively and very loudly and I think that created more chaos than it otherwise would have. But you know, it brings up a really interesting point which is part of this is a data science question, right? Like um does this model outperform this other model? Should we build our own attributes? Should we buy attributes off the shelf? Should we use those attributes to build our own score?

13:02 We're going to buy the FICO score anyway and it does it how does it fit into our underwriting model? All that kind of stuff. So part of it's a data science question, but as you illustrated with your awesome story um about Score X, data science is uh not the thing that ends up deciding these things, right? Like the predictive lift uh even quite frankly, even with like a 30% increase with using cash flow data just by itself you'd think, "Oh, that we're going to do this. Obviously everyone would do this.

13:27 There's a lot of other things that go into this and I think that the other thing that's kind of broken over the last what uh 7 to 8 years has been that the hold that FICO and the bureaus and just the way we've always done this has kind of held on lenders has sort of been broken, right? So, I'll give a couple examples. One is when Bill Pulte pulled the pin on the grenade, it set off like a massive fight between FICO and the credit bureaus. One that I'm close enough still to FICO that I could kind of feel the reverberations.

14:01 Like I'm I'm I'm in Bozeman, Montana. FICO's global headquarters is actually Bozeman, Montana. Like I could like physically like feel a disturbance in the force. And that was very much this question of Well, okay, are we going to go to the bureaus to get the FICO score? Oh, no, FICO is now selling the FICO score directly to mortgage lenders. You don't have to get it through the bureaus. Like that is a huge change just mechanically to the distribution of Where do I get the score? Do I have to get the score from the same place I get the data? No, they're different and now the pricing between the two is different. So, like that just breaks mechanically a thing that has existed and been unquestioned for years and years and years. I don't think people necessarily appreciate you could never buy the FICO score from FICO before, right? It always came through the bureaus and then FICO would like audit the bureaus in terms of how much money they were getting paid and like sometimes they'd get into fights about it. Like it it is a huge change that you can now get the score directly from FICO and the pricing is different. Another thing I think mechanically that broke and Rich, I want to get into cash flow cuz let's let's just go there since uh this is such an interesting area. You know, the cash flow thing the the flaw or the the thing that's been hel- holding lenders back for years on cash flow is mechanically it's kind of difficult, right? And you and I have talked about this before. You know, credit bureau data for all of its flaws and all of its limitations, it's incredibly robust and easy to get, right? Like, you know, the consumer fills out an application, they give you the last four of their social, last name, address, boom. I know if I pull it that if there's a file, I'm going to find it. I know that the performance and latency is going to be incredibly strong. I know that um there's very little friction I'm imposing on the end customer. There's no adverse selection cuz the bureaus collect all the data and the you know, bureaus are all the same today. There's no zip code preference routing tables anymore.

15:54 That has really, I think, stopped cash flow from being able to make as much headway as it otherwise would cuz comparatively, cash flow, I have to get the consumer to granularly permission this data. Where do I put that step in the underwriting flow? This is weird. I don't want to break this like digital lending funnel that I've spent decades refining and building and perfecting. And ultimately, uh I don't even know if like there's going to be really good quality data. I don't know if the latency and performance is going to be good. Oh gosh, the aggregators are in a fight with PNC and this customer uses PNC. I can't access the data. Like, open banking and cash flow data mechanically has been really difficult to get into the underwriting process, even though as we've been talking about the analytical lift is there. And that's the other thing I've seen change a lot over the last couple of years in particular is open banking infrastructure's gotten better. Consumers have gotten more used to it. Uh the idea of, you know, being able to like work this in as a step, whether it's second look or for particular segments or populations or what have you, that's much less of an anathema to lenders than it used to be even like five years ago. So, talk a little bit about like mechanically what's changed and what opportunities that's created for lenders.

17:10 Yeah. So, I think that you're right. The the big thing, in my opinion, since the beginning that that held cash flow underwriting back has been friction. Yeah. And the idea that you know, if you look at the you know, the um different aggregators, they all quote like a conversion rate and you know, let's say the the best number you'll hear quoted is somewhere in the 70 to 80% range and that's incredible, right? To think that four out of five people, five out you know, four out of five people are going to complete the uh the process of linking their bank accounts.

17:39 Um yet if you take a step that has an 80% conversion rate and put that at the top of your funnel, >> you're going to lose a fifth of your business. That's crazy. No one's going to do that, right? >> Nope. And so I think that's uh you know, that's been the problem and I I keep telling people, you know, my my my version of this is um like it doesn't need to be like asymptotically 100%. It needs to be 100% conversion rate, right? You you need to eliminate the concept of drop-off and until you do that, there's going to be holdouts.

18:09 Um and so I think that there are some industry players that are now focused on eliminating the concept of drop-off through different um you know, different structures. We'll see how those go, but I'm I'm excited and hopeful about it. Mhm. Um but I think that's the first thing. I think the reason that uh you know, the reason that you see cash flow getting its um you know, kind of finding its way in Fintech PL first and then let's say Fintech other products second is that um in Fintech PL specifically, there's a lot of verification tasks, right?

18:40 Employment, income, etc. Bank account disbursements, uh details where if you're a lender, you can say, "Well, wait a minute. Um that's a lot of friction for an underwriting step. But if I think of this as a step that does three different verification activities and underwriting at the same time, all of a sudden this is a step maybe I can afford to have somewhere in my funnel." So I think that's you know, that's uh that's probably getting people excited there.

19:07 >> all the tasks, right? So it's like, "Hey, we're going to we're going to have to do all these things individually. There's kind of friction built into each one of these. Let's just bundle all of them and that makes that like friction risk more tolerable because at least we're doing all the things with this data rather than again like 30% lift. You'd think that would be enough, but like people are very protective of drop-off rate friction experience. And it makes sense, right? Because like from a lender's perspective, A, they, you know, they just want to book more profitable loans, right? And so like I can either put more in the top of my funnel and have fewer kind of leak out or I can really like strengthen the underwriting and do a better job evaluating the ones I'm doing, but like either way I have to end up with more in my bucket at the end of the day. And so like if I'm trading one for the other, you know, the data scientist might be so excited about this thing, but if I'm the line of business leader in charge of the P&L for this product, you know, whether it's marketing that's filling my bucket or it's better underwriting that's filling my bucket, it kind of half of one six and six of one half dozen of the other. The other thing is there's an adverse selection thing that you're always concerned about, right? Like experienced lenders are always like peering around a corner and going, "Okay, if I introduce this step that adds friction, and I think about this way with acquisition, too, but like let's just use underwriting. If I add this step that adds friction, who's opting into that friction? Who's opting out of that friction and how's that impacting my performance?" So like there are things that don't necessarily sell it, anyway.

20:37 Yeah. No, I think that's the exactly right. And you know, this is a a thing I've I've come to realize after, you know, more than 20 years of doing consumer credit risk. I've realized that the people who own the consumer experience is probably do more for credit risk than I do as a as a data scientist because you're right. I mean, one one little step, um, you know, when you do verification tests, for example, you can very clearly see it in the data. You lose your absolute best customers and your absolute worst customers, right? The best ones say, "I'm not going to do this." The worst ones know that they're not going to pass it. So, um, absolutely. I think in in you know, you mentioned the the you know, sort of the GM versus data science thing. I think the like the other the other framing of that point is that um you know, we think of of lenders as you know, these homogeneous units of of you know, profit maximization engines. But in reality, it's a bunch of different teams with different perspectives and points of view and goals and stuff like that and I I it is um you know, I think a lot of credit teams are not accustomed to having to go make a case to a product and engineering team and you know, a UX team and and stuff like that and like it's just hard for them to to to do that. Now, every time one of their peers does it and they're one of the only ones left standing who haven't done it, it becomes a little bit easier. But I think Well, you you go to these conferences, too, right? Where it's like uh you can kind of pick out the person who's not at least piloting cash flow because like it's almost like all of their peers have surrounded them and are like peer pressuring them to do it. It is kind of funny to watch cuz like the the data science credit risk community in consumer lending is it's a smaller universe than you would think once you get to know all the folks and they all talk to each other all the time. So, I think you're right, but it kind of speaks to the point about um you know, FICO just dominating the space for so long like data science teams and credit risk teams weren't used to making the business case because there was no business case to make. It was just we're going to use FICO and rich 22-year-old rich like don't come to us with your clever idea for using a different thing. It's not going to happen. So, like that muscle just hasn't been built up and that's kind of seemingly what's getting built right now. Yeah, I agree with that. Um I think you know, the other thing that's that's um that's uh maybe you know, on we're we're kind of talking about two trends. There's this cash flow underwriting trend. There's this what's going on with FICO trends that are converging right now. The other thing is that I think FICO has not had a great answer to cash flow underwriting, right? I think that um my view is well, I don't think it has to be my view. Like I I don't think there's a lot of people who have used UltraFICO, for example.

23:06 >> I I was there when it was launched in 2018, and I I can tell you from personal experience that um there there it was a great story to tell, right? And that this has been the history with FICO's alternative scoring products is they sort of addressed like narrative storytelling problems that the company has. And so like 2018, there's Project Reach at the OCC, this is starting to become a bit more of a thing, open banking has matured. And so here comes UltraFICO, and at the time, I and this is an interesting note comparing where it was then to where it is now, UltraFICO was FICO, Experian, and Finicity. Those were the three companies that were doing it. Uh it was solely being used as like a fallback second-look type of implementation. So like >> Mhm. look at this now, and I I'm guessing what you've observed is that like that stance on how they're doing it and who they're doing it with has changed a bit.

24:00 Yeah, I mean, I think I mean certainly there you know there there partnerships have changed, but I think that the the core point that um I I think like I spoke to a guy who was in um government relations at FICO. And about UltraFICO, and he said to me uh I I he said, "The customer of UltraFICO was not the bank's customers, it was my customers." Right. >> And I think that's exactly right. It was to your point, you know, OCC Project Reach, stuff like that, the legislature, the monopoly, etc. I think that was a lot of the of the story around it. Um Mhm. But I think the thing is that in my view, like okay, it's kind of you know kind of fun to pick on the you know pick on the the the big guy, but um I think the truth is that the FICO score or sorry, that the UltraFICO score the the the architectural issue for me was that it was a composite score.

24:45 Right. >> Which is to say, if you know, I think that if I'm doing something where, you know, let's say most people will have bureau data, very few people but some will have cash flow data, there'll will a Venn diagram where there's a tiny bit of overlap of probably, you know, there's some overlap, some not overlap, whatever. Um, in that world, the idea of just one score scares me as an underwriter. Right? What do I What do I do with that?

25:07 How do I trust that? Etc. versus saying I'm going to go get um, a bureau score and then I'm going to go buy the best cash flow score out there and I'm going to figure out how to use those things together, how to chain them into a decision. So, I think there was a little bit of an architectural, um, issue there as well. But, I think because um, you know, okay, so cash flow has become a thing, people can't go to FICO to get the cash flow thing because they don't like the architecture, maybe I don't like the architecture, whatever. Um, and so they start looking at what else is out there.

25:35 And then all of a sudden you've got all of these, um, you know, fintechs and startups and aggregators and data companies, whatever, that have their own cash flow underwriting analytics. And, you know, the first order impact is, okay, let's all look at those and figure out what else what we should use. But, the second order impact of that is maybe, um, I should be thinking this way about every dimension in my underwriting. Yeah, I mean, that's So, let's talk about those exact, uh, scores because this to me is I think like, and it's funny, I I was having this conversation the other day about stable coins where it's kind of the same thing. Like, we are in the era of a thousand flowers blooming in the desert. Oh my gosh, look at this. This is amazing. We have so many stable coins. A similar thing is happening in the credit scoring space, right? Just running down the list, I mean, we have UltraFICO, which has we've been talking about has been sort of re-launched by FICO. Uh, interestingly, not partnered with Experian this time around. So, uh, read into that what you will about the drama that's been happening with Bill Pulte over the last year. Um, you have the Experian Credit Plus Cash Flow score. Um, you have Prism Data's, uh, Cash score. You have, uh, Plaid's Lend score. You have Nova Credit's Nova score.

26:51 Uh, you have Pave. You have I know one I just found was called clout score, which is like specifically cash flow scoring based on uh like creators and digital economy like, you know, the number of hours you've worked on Uber, that kind of thing. So, there are just so many of these scores that are proliferating right now. My general read of it, and Rich, you would know better, is that they're all taking slightly different approaches, right? Some are composite scores combining credit data and cash flow data. Some are sort of stand-alone uh general purpose cash flow scores.

27:24 Some are cash flow scores built for specific products, uh which is has an analog historically, right? FICO has a auto score and a bank card score and a mortgage score, so you can do that as well. Um some are doing just the score, some are doing scores plus their own library of attributes that they can make available to more sophisticated lenders. So, like, I guess the question I have is that's a lot. Um that is certainly that is certainly more than um we've had I again, this is why like this market has changed so much in the last year relative to last 30 years combined.

28:01 I mean, that's that's way more than we've had over the last 30, 40 years. And I guess I'm curious like, what do you think about the competitive dynamics between all the scores? It seems unlikely to me that we're eventually going to need all of these different scores, so I would imagine some consolidation is probably coming to the market. Um you know, obviously, large lenders are going to be I think more interested in like attributes or building their own scores versus smaller lenders that more sort of want something off-the-shelf. What's your read of kind of the competitive dynamics given all these options that now exist? Yeah. Uh wow, you're absolutely right. I think that we're kind of in the stage right now. It's like you want to you know, you want to organize your closet, you have to make a mess by pulling everything out of it before you you know, you you put everything back and organize it. We're kind of in the you know, we pulled everything out of the closet phase right now. Um and we got to now figure out how to put it back in. Uh yeah. I think you're right that if you asked each of them who you're focused on, you know, I think Pave would say we're, you know, more down market EWA and adjacencies. I think Prism would say we're kind of, you know, PL fintech. Um Novo would would claim to be going after the bank audience. Um and then uh then Experian, I think, like, you know, I don't know audience-wise who they would claim to be going after, but I think that they would say um they're kind of like the you know, the big brand stamp of approval product. Yeah, yeah, yeah. Yeah. And so, I think like, okay, right now maybe there's a a case for each of these in the market, but um or sorry, they're making a case for themselves in the market. I think in reality, um there has to be, you know, two or three winners, right?

29:35 That's usually how these things go. You end up with two or three winners. And so, um yeah, so like my guess is that whichever of these companies can go after a, um you know, kind of a broad enough audience with good enough performance, you know, recognizable brand, recognizable brands using them, stuff like that, like, you know, a couple of them win. And probably a couple of them win attached to credit bureaus, either because it's a credit bureau who won or because a credit bureau buys one of the winners. Like, I just think that's that's probably the most likely outcome there. Um I think that, you know, the the um you mentioned scores versus attributes.

30:12 Like, we talk a lot about the scores, but I think the reality, as you said, is that most lenders will build their own scores off of the attributes. Like, if you think of bureau data as the analogy, most people buy attributes and then build their own custom models, and then maybe they use a commercial score on top of the thing. I think that's probably where we're going. And so, I think that the um the cash flow underwriting solutions that are invested in attributes, I think that's like, to me, like, you know, you're telling the story via the score, but the attributes are what's going to make you successful. Uh Um, and and that is a ground game. You know, like that is a um, categorization.

30:50 Yeah. >> Uh, you know, looking at like hey, this the name of this company looks like it's a coffee shop, but it's actually a restaurant or, you know, vice versa. Like that is that is just really, really hard work. And um, and I think that the like in the end, the people who are telling a great story, but not doing the really hard work are going to be exposed. Um, and we're going to realize like this is this is just a um, it's a really, really messy problem. But I think in the long run, most lenders are not using cash flow scores off the shelf. And most lenders are not doing their own categorization and attributization. I think we follow the bureau model. Yeah, and then to your point like there's two or three players left. So it's exactly the bureau model.

31:31 Yeah, right, right, right. It consolidates down. And and the question is like and the bureau thing is interesting because the bureaus uh, for a long time, and I'll just use Plaid as the example because they were sort of famous for this. Like they were like, "No, no, no, we're not a CRA. Like we don't we're just dumb pipes that transmit the thing." I remember having arguments with them where they're like, "Yeah, it's just like the digital equivalent of taking a shoebox full of bank statements in and like blah blah blah." And I'm like, "Come on. Like that's No, like A, no, and B, like if you want to, you know, get any value beyond being the dumb pipes, you're going to have to do some assembling and evaluating, which Right. that you're a CRA and blah blah blah. And of course we've seen that Plaid and a whole bunch of other companies in the space have embraced the designation of being a CRA.

32:14 And so it is interesting because in addition to having all these new providers of um, you know, scores and analytics, we have quite a few new credit bureaus, which is the I guess kind of the under-discussed innovation that's happened in the market recently. Like we have all these new bureaus, yay. And you know, it's interesting because to your point, I couldn't agree more about the ground game. Like if you've been in the trenches of especially working with like large lenders where all the volume is, attributes are the business, right? And like you can, you know, sell a score, but a lot of times the score is more like this is the thing we give to people when they want to buy our loans on the secondary market, or this is the thing we show regulators. Like this is like the surface level thing, but the actual meat and potatoes of the business is okay, what attributes do you have, what new ones are you introducing, how are you like maintaining attributes is really hard, right? Like there's just a ton of like dirty unfun work that the credit bureaus and FICO do at the the level below the score that kind of blends together with an expertise in the data. And I think you're right, like we're going to have a new set of companies, and some of them might get absorbed by the credit bureaus. I I noted with interest that Experian work in this space from a analytics perspective is like separated from their role as a CRA, which I find kind of funny. So, like they they just want to be like Experian wants to be the FICO of cash flow data, which I think is kind of funny given that they're already a credit bureau, but I guess the other thing I'm curious about Rich is um cash flow data orthogonal to traditional credit data, super duper valuable. The other thing that I think is maybe under discussed a bit with cash flow data is it can provide a level of insight into consumers' financial lives that you just can't get through credit data, right?

33:57 And there's a huge amount of analytical value in that, a huge huge amount. And this gets down to the level of individual attributes and like what store were you shopping at and like just like little tiny things that had a lot of weight from an analytics perspective. They also, I think, would make regulators uncomfortable in some cases. Um and you were kind of talking about the fact that like credit risk teams sort of let their muscles get a little flabby in terms of making a business case. Guess what? Regulators got a little flabby over the last 30 years of being like, "Oh, fair lending. Well, if you're using the credit bureau data where there's not a lot of sensitive, you know, data that's going to get you in trouble, and you're using traditional scores and attributes, like you're probably fine. We won't worry about it too much." Regulators are going to have to get engaged in this conversation, I think, in a pretty meaningful way because I'll just use one example that I've heard. There's a I'm sure hundreds of others.

34:50 If someone is paying their wireless bill every month, that's great. Hey, they're paying their wireless bill. We can see that in their cash flow data, good responsible performance. Uh ironically, utility data that has made its way into traditional credit bureau files as well. So, there's a bit of crossover there. Somewhat, somewhat. Um but now with cash flow data, I can distinguish well, which provider do you work with? And like, for example, do you are you a Verizon customer or are you a Cricket Wireless customer? And that distinction might seem like a distinction without a difference, but analytically, you might be able to assign a higher level of credit risk to the Cricket Wireless customer as opposed to the Verizon customer, even if it's the same network that they're using.

35:32 And I I mention that because it just my intuition would be you might run into some disparate impact issues with that as an example or with any of the other variables you could look at from a cash flow data perspective. So, where are we in terms of like the vetting of all of this data and attributes and analytics from a compliance perspective, and how do you see that playing out? Yeah, I mean, this is so interesting because there's there's a few different dimensions here.

35:57 I mean, one dimension is like, are we you know, are we violating fair lending obligations with any of this data, right? That's the most like simple dimension, right? But then there's the then there's the stuff where it's like, well, technically it's it's okay, the model tasks, and we can do our business justification, but Mhm. should we? You know, are we doing the right thing? And then there's the stuff that's that's not even really bank account data that's showing up in some of these cash flow models where, you know, I question does the consumer reasonably expect that by participating in activity X, they are impacting their um their credit outcome. So, I think on the first one, the the fair lending piece of it, um you know, a lot of I've seen a lot of institutions shifting from input testing to output testing on the fair lending side. Which mirrors, by the way, the shift in regulators' attitude around fair lending, at least under this administration. Yeah. Exactly, but I think that, you know, in a world where, to your point, we've got an entire new category of data that we've never used before, yeah. We're going to have to go back and do some input testing. Um and I'll tell you, you know, I a perfect example of that is uh at one point, I I worked for a lender, um and the town that I worked in, there was a very high-end grocery store, and there was a very um low-end grocery store.

37:08 >> Yeah. Yeah. And um we tested transaction data, and of course, it showed exactly what we expected, that the people who shopped at the very high-end one had way better credit than people who shopped at the very low-end one, right? >> Yeah. I can also tell you, if I walked into those two stores any time of day, I can tell you your fair lending problem, just by looking around, you know? >> Totally. Totally. That That scares me, right? That And so, there's got to be like some, you know, I think um a thing that has driven a lot of my career choices in my my um time in consumer lending is is going after um working for companies that have a do-the-right-thing type of mentality, right? And everybody has a mission or vision statement that says it, but do you actually Yeah. When you have to make the trade-off and and give up $50 million to do the right >> Totally. thing, are you going to do it, right? And I think I've been fortunate to, for the most part, end up with um with companies that that do. Um but I think that attitude and that approach, you know, the obligation on the data scientist, on the the local team, to say, "Is this the right thing?" is way higher now than it has been in 20 years.

38:12 And I'll say the um you know, the the the last piece of this, so let's say what, you know, we talk about Cricket Wireless versus uh Verizon, we talk about that fancy grocery store versus the, you know, the the the other one. Whatever. I think that even beyond this, like if I look at um like I'll just I just give you example, like the um the the Plaid network insights that they talk about, right? Which is if I use if if I get scored by Plaid's cash flow score, um included in that might be um you know, how many 401ks I connected via Plaid or how many investment accounts I connected via Plaid.

38:48 I don't think that the consumer connecting their Chase account to their Fidelity account through Plaid. By the way, they don't might even not know that they're going through Plaid. Sure. Um reasonably expects that this may impact their credit outcomes. Yeah. >> Right? And so I think there's there's this thing where it's like, okay, yeah, like on on page 42 of the terms and conditions you have the right to do this and your fair lending testing is is green and like all of this stuff, but just take a step back and tell me like, is this okay, you know? Mhm. And of course, like obviously, people much smarter than me have have looked at those products and said like, yes, this is okay. And so >> Mhm. um I think the the point I'm just trying to make though is that we, you know, um there is like this this, you know, ethical moral dimension to credit that I think regulation attempts to resolve, Yeah. but doesn't. Sometimes it goes overboard, sometimes it doesn't go far enough, but like separate from that, like we just have to as an industry look at it and say like, do I feel like I'm doing the right thing or not? Does yeah, does this feel right?

39:50 >> I I think you're right. And again, like this is why it's so important to have these conversations is that like we just haven't been really confronted with these problems for 30, 40 years, right? Like I I can tell you for a fact having read a lot of the history books that talk about this, they were talking about this in the 70s and 80s, right? They were having lots of conversations. That's where most of our uh laws on this stuff come from, right?

40:11 This 1970s were just this proliferation of like, hey, you have all this data, what do we consider to be fair and not fair? And it it shapes people's behaviors and actions, I think, in a very profound way. That brings me to the last thing I wanted to ask you about, Rich, which is um there's this consumer-facing element to all of this, right? And um it's so so so important because I think, again, for for its flaws, for its limitations, credit bureau plus one single score that everyone understands is a pretty clear road map to consumers, right? And um I I think you, I, a whole bunch of other people have that like FICO pie chart with the factors that go into the FICO score like seared into my brain. Like I can close my eyes and I can see it. And um did it make total sense? No. Are inquiries in particular a really weird thing to have be have be so impactful on your ability to get credit?

41:04 Yes. Are there plenty of examples where your debt-to-income ratio will be weird for totally legitimate reason, but it'll harm your credit standing anyway? Yes. Like not in any way perfect. Um but it was one thing. And I even remember when um uh I think it was uh Credit Karma was working with um Vantage Score, right? Cuz it was uh a cheaper alternative to the FICO score uh for the purpose of monitoring. So, of course, we'll use this. It's close. It's the same. And there was at times like flare-ups online where consumers were going into lenders and being like, "What the hell? I thought I had a 720 FICO score." And then they're like, "No, no, you have a 720 Vantage Score." And that's slightly different. And like, you know, people would realize just on the margins like, "Oh, this thing I thought is not actually reflective of how lenders view me or think about me."

41:55 Okay. We take that on a small level, and now we go forward to there's how many scores? There are how many different places I as a consumer can go to uh view like like a copy of my report or like to exercise my sort of rights under FCRA, like it is just a broad and very varied world out there. Someone could be using, you know, Plaid network insights and suddenly the explanation I'm getting for why I was declined is that um you know, I connected too many uh investment accounts to my bank account or whatever.

42:29 Like there's just a a a broad and more diverse scope of impacts to consumers and in particular, one I wanted to ask you about Rich is uh Blocks Cash App score. Um and I feel like you and I maybe are the only two people who care about this, but whatever, it's my podcast, so we'll talk about it. Um I could not be more obsessed with the Cash App score, which is an internal model built by Block uh machine learning model that takes in all kinds of data from the Block ecosystem. So Cash App customers use of P2P payments, uh their repayment patterns for the loans that they have, their direct deposit patterns. Um I even heard like they're weaving in obviously they weave in all of the Afterpay stuff cuz that's another sort of lending input to their score. I've even heard there's like data coming from the Square side in terms of like Square sellers and buyers. So it's this big amalgamated score of everything Block knows uh compressed into like what we think your creditworthiness is. Going back to your point, it's the thing we now have the capability to build and so sophisticated lenders are.

43:34 And the thing that Block is doing that I've never ever seen before is they are uh going to be packaging up and A making that score visible to consumers. So if I'm a Cash App customer, I can now see my Cash App score. I can get uh information on what is causing the score to be where it is, what I could do differently to make my score go up, how can I be a more creditworthy customer to Block and potentially to other lenders because the other thing Block is of course doing is uh they're going to be selling their score to other lenders to evaluate uh Cash App customers for loans that uh Block doesn't offer today. Things like uh I think auto lending is a good example of one they've talked about, credit cards.

44:21 I have 10 million questions about it. We probably don't have time to get into all of them, but I want to get your take A on like what you think about the Block Cash App score just as a concept, and B like what you think it's representative of in terms of this fracturing of the consumer experience where depending on who you work with on the banking side or depending on which lender you go to, they might be evaluating your credit in wildly different ways. Yeah. So, okay.

44:47 So, let's let's go back to your your you started with Credit Karma. Let's go back there. Um with consumer display. So, they display um Vantage 4 or or Vantage 3. I can't I don't remember which one it was honestly, but um but uh you know, you're right. So, so let's say that Credit Karma had unlimited pockets and could, you know, fully serve the customer, right? So, Yeah. now they're going to display Vantage 3 and 4 and FICO 2 4 5 8 9 10 T bank card auto, right?

45:12 And then tell the consumer, you're going to be evaluated based on one of these when you go to seek credit or perhaps none of these because you might go to a company that just uses their own internal scores. Right. >> right? Exactly. Here's Here's what we can do. This is the most accurate picture, yeah. Yeah. And so, I don't know that um that, you know, like selling serving the right score is the answer because there is no right score.

45:33 >> is the right score, yeah. >> Credit Karma said um with Light box was, "Hey, let's let's just size up this whole thing. Instead of saying like here's, you know, the score that you might or might not be used to get approved, let's just tell you whether you are approved or not. That's much easier whether you're likely to be approved or not." Um okay. So, that was that was the Credit Karma uh learning. And then you can say, "Great. Um this this Cash App thing is exactly the right uh you know, the right learning there, which is we don't need to go buy a commercial score. Let's take the best that we know about the consumer and say based on everything we know about you, you seem to be pretty good or pretty bad or really good or really bad or whatever. That's that's amazing. Um and uh um and we're going to use all this stuff across the ecosystem. And by the way, it's probably the thing we're going to use if we underwrite you for something.

46:14 So, you know, have a really good sense of what your life looks like inside our ecosystem, which is amazing. And by the way, um benefit to Block and to um or to Cash App and to the consumer is none of this stuff, for the most part, gets reported to the credit bureaus right now. Nope. And so, we will commercialize this. So, that if you apply to a mainstream um bank, you get credit for all the good things that you've done on Cash App. Yeah.

46:41 >> I think that's amazing. I love this answer. Now, the the issue is it's like, okay, the um you know, in a in a single single game game theory, we got the right answer, right? Multi-game. >> Double it. Yeah. >> Now, you got a question, right? Okay, so so, fine, they do this cuz they have products that don't show up on on Metro 2, right? Sure. >> What if, like, I don't know, PNC decided to do this.

47:02 And they say, "We're not going to report to the credit bureau. We're going to go to the PNC score." >> Yeah. >> of America, if you want to underwrite a PNC customer, you got to buy the PNC score. By the way, we won't sell it to you because we don't want you to take our like you just imagine the chaos, all right? So, I don't know. I love the short-term answer. I really question what the extrapolation of this is.

47:21 Yeah, it's funny, right? Because you're right. And this goes back to the whole like moral dimension to credit and credit risk in a way, too, where it's like, you know, and I I think about this a lot, especially with fintech, right? We'll pick on BNPL as an example. BNPL providers don't want to share their data with the credit bureaus. Uh now, Affirm is but Affirm is more of a point-of-sale lender that does like installment lending for the most part.

47:45 So, like, yeah, great, but that also doesn't really count. But like, if you do like true pay-in-4, this is very different than the data that's captured in the credit bureaus. There was a legitimate argument a couple of years ago that the credit bureaus are not like from an infrastructure perspective ready to take this data in. Uh if the data gets taken in, we don't know how the scores will, you know, address it or evaluate it, blah blah blah blah blah.

48:06 You fast forward to today, while some of those concerns persist, the bureaus and FICO have been doing a lot of work to try to say like we are ready to take this data in if you are willing to give it to us. And lo and behold, the BNPL providers were like, "No, we still don't really think you're ready. We're, you know, we're nervous about this, whatever." I have always sort of thought, and I I can't really blame, this goes back to your single player game theory thing, I don't really blame any one company for thinking this, but I think the actual answer is and I'll just use Klarna as the example, we go into every market that we compete in. The first thing we do in that market is we give everyone a loan for a couple hundred bucks.

48:43 The people who pay us back, we learn something about them and their willingness to pay, particularly because a lot of these folks are young, maybe they haven't had a lot of exposure to credit before, and you see this in the data that Klarna publishes, their loss rates go up when they go into a new market, and then they go down. And the reason is because they have these short-term loans that they churn through to basically discover goods and bads in that population for their product. That is a huge expense from a go-to-market and a risk perspective to expand into a new market. And I think if you were to get Klarna in a moment of public honesty, uh they would never actually say this, but I think if they were to be totally honest, they would say, "We don't really feel like sharing that data with everyone else. We kind of think that that's our proprietary data, that's our proprietary advantage. And like there's really no advantage to us. Yeah, there's some loan stacking and some things that we would like to be able to catch if we shared data, but for the most part, like we can sort out goods and bads on our own, and we don't really want to share this data with our competitors or to buy it back in the future when it's been combined with other companies' data."

49:43 That to me is much more, and this is why the Block thing is so interesting, that to me is almost more of a 1970 versus 2026 thing, right? Where it's like in the 1970s the credit bureaus went to lenders and were like, "Hey, we're digitizing, we're consolidating. Just here's the Metro 2 format. You plug in machine to machine. We're saving you a ton of time actually because you don't have to like, you know, hand deliver us all of these credit records and like go through all the work that you were doing before. It's just machine to machine. It's much easier, blah blah blah." And lenders at the time, merchants, retailers, and and banks were like, "Great. Awesome. You're saving us a ton of time. Who needs this data? Why is data valuable?" Fast forward to 2026, every company, particularly these large sophisticated fintech companies like Block or Klarna, they know their data's really valuable.

50:30 So, I I do think your point about this like game theory question and what's right for an individual company what versus what's right for the market, I I I don't know that this is the last time we're going to see this and I I am really wondering, like a thing I'd love to know and I just don't know is is there someone at Chase who's watching the Cash App thing and are just like, "Huh, you know, like this is interesting because we have a huge part of the ecosystem. We have a ton of credit card customers. We have a ton of deposit data. Like, what if we didn't share?

51:02 Like, what would that look like? It's It's interesting to pull on that thread." Yeah, absolutely. I mean, I think, you know, okay, so I'm I tend to believe that the market usually gets to the right answer with enough time, but might do some damage along the way, right? So, if that happened, like I think we get to a place where you you end up with out an oligopoly. You end up with like this many-to-many relationship and you end up with, you know, the market finds market fair prices and and stuff gets shared and consumers, you know, come out um perfectly fine, but I think the process of getting in there would be really messy in that world. And I think consumers might, you know, suffer along the way. So, I yeah, I don't know. Um it's going to be fun. Let's Let's do this again in a year and and reflect on what happened.

51:46 I That's exactly what we need to do. Yeah, I mean, I think we didn't even talk about like AI as the overlay on top of all of this, but it's interesting because you know, I think like AI, depending on how you squint and look at it, could be something that either like requires the market to come back to a more consolidated point of view, right? So, you could say like, "Hey, you know, like we don't want to burn tokens having AI chase all of this stuff and try to like weave together a picture from disparate sources. One of the things you're seeing, I think, as firms get more rigorous about like architecture for AI and for LLMs is really structured data that LLMs can just go through really fast and get great answers out of more tabular data, that kind of stuff. But, on the other hand, you could also see a version of AI where you know, hey, it's fine that it's a many-to-many relationship. All as a consumer just have my AI agent go across all these different places, assemble a picture for me, and like I don't know, is it is it crazy to think that in the future our AIs might act as the sort of like ambassador or gofer for lenders we're working with, where it's like, "We'll go collect all the data, we'll bring it to you in a bespoke machine-readable format, and lender, you take what we give you and run with it." Like, you could see that playing out, too. So, it's to your point, the messy middle, I don't know how this plays itself out.

53:11 Yeah, I don't know. Um, that's interesting. I have not thought of this angle where it's it's sort of consumer-owned, consumer-initiated, but uh, but you know, when we talk about um, deposit data, like that's kind of the genesis of the conversation. It's like, the consumer owns this data, so why not, right? Yeah, yeah, exactly. No, I think that's I think that's right. Um, all right, Rich, I will let you go. Thank you so much for coming on. Uh, truly, this was a conversation I'd wanted to have for a while. So, really appreciate it, and uh, we'll have you back again soon.

53:37 It was a lot of fun. Thanks so much, Alex. >> I