Open-loop vs. Closed-loop Credit Scoring
Happy Friday, Fintech Takers!
I hope you’ve had a great week!
Mine was great, until Jason Mikula shared Binance’s latest initiative — crypto accounts for kids (targeting kids as young as six) and a children’s book called the “ABCs of Crypto”.
Yes, this is a real thing, not an early April Fools joke. Entries in the book include “L is for Leverage,” “M is for Memecoins,” and “Y is for YOLO”. Also, in an apparent effort to troll us, they included “K is for KYC” and explained that “Exchanges like Binance ask users to show who they are by sharing their ID documents. This helps keep everyone safe and makes sure only real people can use the platform.”
Speaking as a parent who is trying to teach my kids to always be honest and to avoid taking shortcuts, I would like to tell the folks who created this book: GO FUCK YOURSELVES.
[Takes five calming breaths]
OK, let’s talk about fintech!
- Alex
P.S. — If you haven’t filled out this industry events survey yet, would you do me a huge favor and complete it?
Our team is planning our events calendar for 2026, and I’m hoping to meet up with as many of you as possible!
Open-loop vs. Closed-loop Credit ScoringCopy anchor linkCopied
On May 24th, 2018, President Trump signed the Economic Growth, Regulatory Relief, and Consumer Protection Act into law. That law included a section that directed the Federal Housing Finance Agency (FHFA) to open the credit-score model approval process for loans purchased by Fannie Mae and Freddie Mac, creating a formal path for alternative credit-score models (such as VantageScore) to be considered.
On May 24th, 2018, the price of a single share of FICO’s stock was roughly $180, and the price of a FICO credit score for originating a mortgage was roughly $0.50 to $0.60.
By late 2024, the price of a single share of FICO’s stock hit an all-time high of $2,382.40. At about the same time, FICO announced it would raise the price of the FICO Score for mortgage lending in 2025, from $3.50 to $4.95.
And in July of 2025, FHFA Director Bill Pulte tweeted this:
To put it bluntly, the value of the FICO Score, as a standard measurement of consumers’ creditworthiness, died on May 24th, 2018. However, for a very lucrative seven years, FICO was the only one who knew it.
Now, of course, everyone knows.
FICO has ended its exclusive distribution relationship with the credit bureaus for mortgage credit scores. Equifax has retaliated by pricing the VantageScore way under the FICO Score for the next two years and giving free VantageScores away alongside the FICO Score in 2026 in order to encourage conversions. FICO has revived its dormant cash flow credit score (UltraFICO) in partnership with Plaid (and, glaringly, without any participation from the credit bureaus) and made aggressive moves into generative AI.
As we come to the end of 2025, I think it’s safe to say that the credit scoring market is more fractured and chaotic than it has ever been.
And the question is, what comes next? What forces will shape the credit scoring market in 2026 and beyond?
There are lots of different answers — cash flow underwriting, generative AI, Bill Pulte — but I want to focus on one in today’s essay: the emergence of closed-loop credit scores.
All of FICO’s Moats Are Being Filled InCopy anchor linkCopied
It’s important to point out that the FICO Score’s role as an industry standard isn’t simply the result of the government-granted monopoly that it enjoyed for the last 30 years in the mortgage market. That was a deep competitive moat, but it wasn’t the only one that FICO had.
As I wrote in this essay, the FICO Score’s long-time status as an industry standard was also due to A.) the analytic value of the score in predicting consumer loan defaults, and B.) the educational value of the score in helping consumers understand and improve their creditworthiness.
Both of those value propositions have (quietly) been under attack for the last couple of decades.
Here’s what I wrote about the declining value of the FICO Score, as an analytic tool:
Unlike in 1958, FICO’s expertise in credit scoring is not, today, uniquely valuable.
All of the big banks (which account for the majority of consumer lending volume in the U.S.) have built their own custom credit scoring models, using a combination of bureau data and attributes, external scores like FICO, and proprietary internal data and attributes. These models don’t generally outperform FICO on an industry-wide basis, but they do outperform FICO for their banks’ target customer segments (which is obviously what the banks care about).
Even fintech lenders, which often start out being way too confident in their ability to build an underwriting model that can outperform FICO, eventually accumulate enough performance data and credit risk management expertise to be able to build superior custom models for their own products and portfolios.
The most commonly used version of the FICO Score, for non-mortgage lending decisions, is FICO 8, which was released in 2009. For mortgage lending decisions, FICO 2, 4, and 5 are all still commonly used.
You can understand why lenders (especially large ones) are forgoing upgrades. Why pay more money for a newer version of a score that will be a pain to migrate to and that you don’t really use much anyway?
However, from FICO’s perspective, this is a very bad self-reinforcing trend:
Now, this trend, by itself, wouldn’t necessarily spell the end of the FICO Score as an industry standard. The score, even if it’s a little different from the internal models developed by large banks and fintech lenders, would still be directionally helpful for consumers to understand their creditworthiness in the eyes of lenders and to plan out future financial milestones that require credit, such as buying a car or a house.
However, we also have to consider a different trend, which started in the 2010s and accelerated in the 2020s: fintech companies refused to play nicely with the credit bureaus (and the credit bureaus repeatedly and embarrassingly failed to incentivize fintech companies to play nicely with them).
Kevin Moss and I wrote about this trend in depth, so I won’t belabor the point here. The basic story is that as consumer-facing fintech companies (neobanks, BNPL providers, etc.) grew, they made strategic decisions that undermined two long-standing assumptions in the credit risk space:
- All mainstream consumer lenders furnish repayment data to the credit bureaus.
- The data furnished to the credit bureaus is legitimate and genuinely reflects the financial management and loan repayment decisions made by consumers.
Today, these two assumptions are no longer safe to make.
Fintech lenders — pay-in-4 BNPL providers, most prominently — decided not to give away their hard-earned proprietary data to the credit bureaus and their competitors. And neobanks and other B2C fintech companies discovered that hacking consumers’ credit scores via clever (and occasionally fraudulent) credit builder products was a great way to acquire and retain customers.
This has led to a significant degradation in the quality and coverage of the data that the credit bureaus have, particularly for young, subprime, and credit-invisible consumers. That’s not FICO’s fault, but it is FICO’s problem, as I wrote:
Imagine that you’re a 20-year-old who has never had a credit card or an installment loan before. To date, you’ve mostly used BNPL to finance specific purchases, but you also recently started using the Chime credit builder card in order to build up your credit score faster because you are planning to buy a car next year.
Fast forward one year – you walk into a car dealership, pick out the car you want, and apply for a loan in the finance guy’s office. You are extremely confident because Experian is telling you that your FICO Score is 685. However, your confidence turns to befuddlement and anger when the finance guy informs you that you don’t qualify for any loan because, unbeknownst to you, every lender that sees your credit report is screening out or penalizing you for that Chime credit builder tradeline, and none of the BNPL loans that you paid off were being reported to the bureaus.
This isn’t theoretical. This is happening today.
So, to sum up:
- The FICO Score is not used to make a large and growing portion of consumer credit decisions.
- The FICO Score does not accurately reflect the true creditworthiness of a moderate but quickly growing portion of consumers.
Or, to put it even more succinctly, there is an increasingly large gap in the credit scoring market.
And it is at this point in our story that a new character appears.
Enter: Cash AppCopy anchor linkCopied
Cash App, as you know, is a fintech company and digital bank that serves roughly 58 million consumers.
In 2020, Cash App jumped into lending, introducing Borrow, a short-term lending product. A year later, Block (the parent company that owns both Square and Cash App) acquired Afterpay (a BNPL provider) for a cool $29 billion. In the years since, Block has worked to scale up Borrow and to integrate Afterpay within both the Square and Cash App ecosystems.
Here is where the two offerings stand today:
- Borrow allows eligible customers to take out loans between $20 and $500. The criteria for eligibility and available borrowing limits are based on many factors, but generally require either spending $500 per month on the Cash App debit card or linking an external bank account that shows at least $500 in monthly deposits. Loans come with a flat 5% fee, and the term is four weeks. Customers are given the option to pay in weekly installments, pay the entire balance (or portions of it) at any time, or “pay as you go,” which automatically takes 10% of all incoming Cash App deposits and applies them to the balance due until it’s paid off or the final due date arrives. If customers fail to repay a loan by the final due date, after three days they are assessed a one-time $5 late fee, and after seven days they are charged 1.25% in non-compounding interest every week. If there is money in a delinquent customer’s Cash App account or an external debit card account that the customer has linked, Cash App will automatically use that money to pay off the loan. Borrow loans were originally originated by Cash App’s partner bank, First Electronic Bank. However, Block recently received approval from the FDIC to begin originating and servicing loans directly, using its industrial bank subsidiary, Square Financial Services.
- Afterpay, like most BNPL providers, offers two main products. The first is a pay-in-4 loan, which is used to finance small-dollar purchases at 0% interest over six weeks, with the customer paying 25% upfront and the remaining balance in three two-week installments (plus a fee if you pay late). The other product is an installment loan to help finance larger purchases, with terms of 3, 6, 12, or 24 months (depending on purchase size) and interest between 0% and 35.99%. Block has recently integrated Afterpay within the Cash App ecosystem, making it available pre-purchase with specific merchant partners and post-purchase for eligible transactions made on the Cash App debit card.
Block does not utilize credit bureau data to underwrite these loans.
In fact, the company has written extensively about its belief that the traditional credit system is structurally incapable of supporting Americans’ need for affordable credit:
The reason so many Americans struggle to access affordable credit isn’t a mystery. Today’s dominant approach of creditworthiness — centered on credit reports and traditional credit scores — relies on static, backward-looking data that fails to capture near real-time financial health and fails to serve consumers with thin credit files or none at all.
So, instead, Block has developed its own proprietary approach to consumer credit underwriting, which it refers to as the “Cash App Score”:
Block’s underwriting models are intentionally built to serve customers who have been overlooked by traditional financial institutions. Many of its customers are young, new to credit, or come from communities with less access to traditional banking. Block looks at patterns such as steady paycheck deposits, responsible spending, or consistent payments — indicators that can reflect financial health. For example, Cash App’s first-party data includes near real-time transactional data including peer-to-peer (P2P) payments, prepaid debit card and ACH transactions (including paycheck deposits, bill payments, unemployment payments), retail investing, and more. Because this data reflects how customers are actually earning, spending, and saving in the moment, it serves as a powerful macroeconomic signal for the Cash App customer base. When a consumer or business shows signs of improving financial health, such as building savings or successfully repaying a loan, Block can proactively offer larger credit lines or new opportunities without relying on delayed updates from traditional credit bureaus. This creates a lending experience that adapts to a person’s progress, rather than remaining fixed based on outdated information.
This is essentially the same thing that big banks like JPMorgan Chase and Capital One have been doing for a long time — building a proprietary underwriting model tuned to the specific segments of consumers that the company is serving, rather than using an off-the-shelf model like FICO.
The difference with Cash App, is that its target customer segments tend to either be credit invisible or subprime (70% of borrow customers have a FICO score below 580), so it has built its underwriting model entirely off of a combination of cash flow data (from Cash App and/or external linked accounts) and other proprietary Cash App ecosystem data (P2P payments, investing, etc.)
The company claims that the Cash App Score allows them to approve more customers, safely, than they would be able to using traditional credit bureau data:
And to keep loss rates low as they increase Borrow credit limits and grow their portfolio:
That all sounds great, but it’s fundamentally the same thing that many other banks and fintech lenders have also been doing. Custom, ML-driven underwriting models aren’t new. Cash flow underwriting isn’t new.
However, here’s something that is (as far as I can tell) new:
Today, Cash App announced a pilot program that gives select customers visibility into their Cash App Score, a near real-time measure of a Cash App customer’s financial health.
The score calibrates in near real-time and is surfaced to the customer weekly to reflect current financial health within the Cash App ecosystem. We’ve used Block’s AI tools to help translate our model features into customer-friendly concepts and actionable recommendations, making our state-of-the-art models accessible and digestible for customers.
Select customers can navigate to the Score tile within Cash App to view their current score and understand what drives it. The experience shows specific actions that can help improve a customer’s score, and as customers implement these behaviors, their score adjusts accordingly. The tile also reveals what increased scores can unlock—for example, a customer may see that raising their score by 20 points may grow their Cash App Borrow limit.
Whoa!
Block has basically created its own mini-version of Credit Karma, but built around the Cash App Score rather than a traditional, general-purpose credit score like FICO or Vantage, and is embedding it directly within the app.
Here are some screenshots:
As innovative as this is, it’s also a very intuitive and logical step for Block to take.
The company is not using traditional credit bureau data or credit scores to make decisions. And it’s not furnishing its own repayment data (for either Borrow or Afterpay) to the credit bureaus. So why, when it comes to helping customers understand their creditworthiness and plan for their financial futures, wouldn’t it just surface the score that it actually does use to make credit decisions, and layer on top actionable insights and recommendations (in much the same way that Credit Karma, Experian, and FICO do) for how to improve it?
Once you put it like that, it’s kinda obvious, right?
And it’s getting a really positive response from customers, according to Brian Boates, Risk Lead at Block:
As customers see their score and take action, it can drive positive behaviors and deepen engagement with Cash App.
The FICO Score, in its heyday, was a robust open-loop credit score. It was interoperable across different data providers and different lending products, and it helped lenders, investors, and consumers all make decisions using a common standard.
As we’ve covered (and as I predicted!), the days of the FICO Score as an unassailable industry standard are numbered.
What appears to be emerging now — and Cash App is the first large-scale example that I’ve seen — are closed-loop credit scores, designed to reframe the eternally common consumer question of, “how much credit am I qualified for?” to “how much credit from you am I qualified for?”
This is an extremely interesting shift.
On the one hand, you can understand why companies like Block would want to shift the question to more of a closed-loop framing. Getting users to think about what they can do, narrowly, to be better customers within your ecosystem is obviously very valuable. It increases customer engagement and facilitates ongoing cross-sell opportunities.
Honestly, I’m kinda surprised that no other large bank or fintech lender has done it. I guess the FICO Score Open Access program, which allows lenders that buy the FICO Score to give it, for free, to their customers in their monthly statements and digital banking portals, probably diminished the incentive for someone like a Capital One or American Express to try it.
(Editor’s Note — Very few people know about Open Access, but it’s absolutely one of the smartest things FICO has ever done. It really strengthened the FICO Score’s role as a consumer-facing standard.)
It would not surprise me if we saw other fintech lenders (Chime? Klarna?) and maybe even a large bank follow Block down this path.
However, on the other hand, we should acknowledge that a closed-loop credit scoring system is suboptimal for the end customer, as it limits their ability to shop around and to access affordable credit from other providers.
To their credit, the folks at Block seem to understand this. Towards the end of his presentation at Block’s recent Investor Day, Brian Boates said this:
The Cash App score presents a number of significant opportunities for us to transform how customers understand and access credit. We want to give customers not just more transparency, but also more control over their data and their credit scores, and potentially even be able to take that with them and qualify for credit products that maybe Block doesn't currently offer. The traditional credit ecosystem has many multi-billion dollar players, but we believe it also has a lot of room for improvement, especially when it comes to transparency and control for consumers. So while still very early, the Cash App score shows potential to be a distinctive revenue-generating offering for block, not just a powerful internal tool for first-party products.
Here’s the accompanying visual from his presentation:
Once again, WHOA!
Block appears to be seriously exploring the idea of competing with traditional credit bureaus and credit score providers, and this poses a few questions:
How exactly would this work?
There are a few different ways it could work, in theory.
The most straightforward would be to follow the Credit Karma model: bring in a set of embedded offers from other, external lenders for products that Cash App doesn’t offer, such as auto loans and mortgages. Use the data Cash App has on its customers and the Cash App Score to augment (but not replace!) the credit underwriting processes of the integrated lender partners. Collect a fee for facilitating new loans for those partners.
Another approach would be for Block to sell its Cash App data and score to other lenders, for them to use in their own underwriting processes. Depending on exactly what Block was providing, this would either be structurally similar to FICO (just selling a score) or to the credit bureaus (data + a score). The FICO approach would be preferable, as it wouldn’t require Block to become a consumer reporting agency (CRA) under the Fair Credit Reporting Act. However, it’s not clear to me if Block would or could sell the Cash App Score without Cash App data, as the score appears to need Cash App’s proprietary data to work.
So, the Cash App Score can’t work without Cash App’s proprietary data?
I don’t know, but this is an important question!
From what I can tell, the scoring model mostly runs on a combination of loan repayment data and bank transaction data. Today, that loan repayment data comes from Borrow and Afterpay, and the bank transaction data comes from Cash App. However, there is no reason (in theory) that the score couldn’t work with other short-term loan repayment data and bank transaction data. In fact, we know that it already does work with non-Cash App bank transaction data linked through Plaid.
The question is: can it work solely with external (i.e., non-Cash App) data? Or, like Plaid’s new cash flow underwriting score, LendScore, does the Cash App Score require proprietary ‘network insights’ to work?
Interestingly, if Block doesn’t provide data with its score, other lenders will have a hard time evaluating Cash App customers for loans because Block steadfastly refuses to furnish loan repayment data to the credit bureaus or enable Cash App customers to easily share their data with other companies via open banking (though I hear this second issue may be addressed soon).
What unique value would the Cash App Score provide to other lenders?
This is the most important question, and one that Block will need to grapple with honestly if it’s really serious about exploring this ‘Cash App Score monetization’ idea.
Like all fintech lenders, Block is very proud of its underwriting model and how it has performed, even as Cash App and Afterpay have ramped up their lending volume. That’s understandable.
However, the reality is that Block’s credit performance over the last five years is due, in large part, to the structure of its lending products rather than the intelligence of its underwriting models.
Borrow and pay-in-4 BNPL are small-dollar, short-term lending products. On an individual loan level, the amount of credit risk being taken is minimal. And like what it does on the Square Capital side of the house, Block further stacks the deck in its favor by automating the repayment of Cash App loans, sometimes by pulling directly and programmatically from incoming payment streams.
Block acknowledged this exact point in its investor presentation:
Now, to be clear, it’s very smart of Block to limit its risk through these structural design choices. However, those choices also make me a little more skeptical of the argument that the analytic value of the Cash App Score is uniquely valuable.
If I were a credit risk executive, I would have a hard time trusting the Cash App Score to evaluate risk when originating higher-dollar, longer-term loans because the score itself isn’t trained on data representing those loan types.
Shit, even Block doesn’t put all of its trust in the Cash App Score when originating higher-dollar, longer-term loans! Afterpay still pulls a traditional credit file for its interest-bearing installment loans.
This isn’t to say that the Cash App Score isn’t a good analytic tool or that Block shouldn’t try to build an ecosystem around it. However, what FICO teaches us is that the distance between a closed-loop credit scoring system and an open-loop credit scoring standard is vast.
I doubt Block (or anyone else) will be able to bridge it, but we’ll see!
MORE QUESTIONS TO PONDER TOGETHER
Big news for the endlessly curious (yes, you): I’m collecting your fintech questions on a rolling basis.
What’s keeping you up at night? What great mysteries in financial services beg to be unraveled? Think of it this way, if a stranger is a friend you just haven't met yet, your question is a Fintech Takes conversation waiting to happen.
One that could headline a Friday newsletter or be answered in an upcoming Fintech Office Hours event.
Drop your question here, whenever inspiration strikes!
Thanks for the read! Let me know what you thought by replying back to this email.
— Alex
Join 2,444 other finance and fintech leaders in the Fintech Takes Network
The conversation doesn't have to stop here
Keep learning and connecting in the Fintech Takes Network
EVENTS | FEED | LIBRARY | DIRECTORY
