{ad_logo_primary}

Quick favor: Our records indicate that you aren’t opening this email. But records can be wrong. Please click here if you’d like to remain subscribed to Fintech Takes. 

Happy Monday, Fintech Takers!

And happy (belated) Mother’s Day to all the fintech moms out there! 

I’m extraordinarily lucky to be surrounded by incredible moms, including my own and my wife. I can’t come close to repaying them for all they’ve done, but I’ll keep trying, on Mother’s Day and every other day.

Let’s get into it.

- Alex 

P.S. Next week, I am hosting a webinar with Narmi on the way forward for community banks in the age of fintech. I’d love to have you join us if you’re available. Register here!

Was this email forwarded to you?

{ad_content_primary}

Czechoslovakian Army Entering Vladivostok, Siberia, in 1918 by George Benjamin Luks.

3 FINTECH NEWS STORIES

Dave reported better-than-expected Q1 2025 earnings:

Dave, currently valued at $1.44 billion by market capitalization, reported its first-quarter 2025 earnings, surpassing analyst expectations with a significant jump in both earnings and revenue. The company reported earnings per share (EPS) of $2.48, substantially higher than the forecasted $0.75. Revenue for the quarter reached $108 million, exceeding the anticipated $92.63 million. Following the announcement, Dave’s stock surged by 28.63%, reflecting strong investor confidence. 

What was responsible for this unexpected leap? Here’s one reason shared by Dave CEO Jason Wilk:

I want to provide an update on our recent transition to the new fee structure for extra cash. As a reminder, on February 19, we fully transitioned to a new fee structure consisting of a flat five percent fee on all extra cash transactions with a $5 minimum and a $15 cap, removing optional tips as well as additional transfer fees to Dave checking.

Consistent with the testing we performed at the end of last year and into early this year, results have been better than expected. With this change, we’ve unlocked enhanced member lifetime value through improvements in conversion, retention and monetization among new and existing members. Approximately 60% of total originations were on the new fee model in Q1. So we will receive the full benefit of the change in Q2 onwards.

Dave finally stopped asking its customers for tips, which I have been BEGGING fintech companies to do for years now, and lo and behold! Dave made more money!

Who could have predicted this?!?

Here’s what I wrote last year:

Fintech apps asking for tips, like they’re struggling art school students picking up extra waitressing shifts in order to scrape by, is complete and utter bullshit. It makes me embarrassed to work in this industry.

Pick whichever side of the argument you want. There’s no good defense for it.

If you argue that it’s a bad deal for consumers because most of them will consistently choose to pay a tip and will, in many cases, pay more than a company might choose to charge as a mandatory fee, well, obviously, that’s bad. Like, predatory lending bad.

If you argue that it’s actually not that bad because most customers don’t choose to tip consistently, and the ones that do don’t pay very much, well, that’s bad too! How do you build a sustainable business when you don’t know how much money you’re going to generate?

Tipping is fine when it's an optional way to show gratitude to individual employees working in the services industry. When it’s a central component in a financial services company’s business model, it’s the dumbest fucking thing ever. Either you will rip your customers off (which Dave was trying to do with bullshit like this) or your customers will rip you off (by choosing not to tip, even if they would have been willing to pay a mandatory fee). 

Recently, it appears that Dave’s customers had been ripping it off. And by moving to a new fee structure, in which Dave clearly articulates the rate and the maximum and minimum amounts that a customer will pay, the company is generating more revenue with no adverse effects. Here’s Kyle Baumann, Dave’s CFO, responding to a question about the relationship between the new fee structure and credit performance:

To answer the question about the relationship between credit performance and our new pricing model, we just haven’t seen any changes there in terms of like something like adverse selection or anything like that that might occur. It’s been all positive. I think as we’ve mentioned, our conversion rates are up for new customers and existing customers have adapted to the new pricing structure as well and we’re seeing no negative impact.

It’s all been upside from a customer orientation standpoint and I think the upside on the size per origination as well as the average revenue per origination speak to the business benefits.       

I’m usually not one to say I told you so, but …

Jonathan McKernan, who had been tapped by President Trump to become the next full-time Director of the CFPB, has now been nominated to a senior post at the Treasury Department:

President Donald Trump's pick to lead the Consumer Financial Protection Bureau, Jonathan McKernan, has instead been nominated to a senior role in the Treasury Department, the Treasury said in a statement on Friday.

McKernan, who Trump tapped in February to head the consumer watchdog and was awaiting Senate confirmation, will now be nominated for the role of undersecretary of domestic finance at the U.S. Department of the Treasury.

The move means McKernan will no longer be under consideration to lead the CFPB and a new nominee will need to be named, a White House official said.

In keeping with this administration’s modus operandi, this announcement caught everyone by surprise, and no one seems to have any idea what it means or what’s coming next.

Fun!

The simplest reading of this news is that McKernan realized what a bad job running the CFPB over the next four years would be, and he jumped to a life raft. This is literally what Democratic Senators warned him about during his confirmation hearings, with Elizabeth Warren saying, “It kind of feels like you’ve been lined up to be the No. 1 horse at the glue factory.” And Jack Reed telling him, “I have this sinking feeling that you’re departing Liverpool on the Titanic. Good luck.”

Well, good news! McKernan avoided boarding the Titanic, like he was Sven Gundersen losing to Jack Dawson in poker. 

It’s not great for those of us who are on the boat. Russell Vought and Mark Calabria still seem intent on gutting the bureau and undoing every single thing that Rohit Chopra accomplished over the last four years, with open banking as the latest target.

I think the best we can hope for is that Scott Bessent takes a much more hands-on role in managing the week-to-week goings on at the bank regulatory agencies, and he keeps the trains (mostly) running on time, with the help of McKernan. This would be unusual for a Treasury Secretary, but Bessent seems to be leaning that way based on his public comments and what I’ve been hearing behind the scenes (make sure to listen to this week’s Fintech Takes podcast for more on this!)     

Among Stripe’s many recent announcements, this one stood out to me:

Stripe’s payments foundation model has been trained on tens of billions of transactions, Emily Glassberg Sands, Stripe’s head of information, said. So it “captures hundreds of subtle signals about each payment” that other models would miss, she said.

Model development has always been a mix of art and science.

The science (various forms of statistical analysis and machine learning) can be leveraged to identify patterns in large datasets that can aid in predicting specific events, like fraud or credit default. The art is in pointing those scientific tools at the right areas, which was, traditionally, the province of humans with ample domain-specific experience. These folks based their model design decisions on a combination of data and their own hunches and intuition.

Stripe’s announcement that it has built a foundation model specific to payments represents, perhaps, a move beyond this traditional paradigm.

Here is a quote from a tweet by Gautam Kedia, Applied ML Leader at Stripe. I apologize for the length, but I think it’s utterly fascinating:

For years, Stripe has been using machine learning models trained on discrete features (BIN, zip, payment method, etc.)  to improve our products for users. And these feature-by-feature efforts have worked well: +15% conversion, -30% fraud.

But these models have limitations. We have to select (and therefore constrain) the features considered by the model. And each model requires task-specific training: for authorization, for fraud, for disputes, and so on.

Given the learning power of generalized transformer architectures, we wondered whether an LLM-style approach could work here. It wasn’t obvious that it would — payments is like language in some ways (structural patterns similar to syntax and semantics, temporally sequential) and extremely unlike language in others (fewer distinct ‘tokens’, contextual sparsity, fewer organizing principles akin to grammatical rules).

So we built a payments foundation model—a self-supervised network that learns dense, general-purpose vectors for every transaction, much like a language model embeds words. Trained on tens of billions of transactions, it distills each charge’s key signals into a single, versatile embedding.

You can think of the result as a vast distribution of payments in a high-dimensional vector space. The location of each embedding captures rich data, including how different elements relate to each other. Payments that share similarities naturally cluster together: transactions from the same card issuer are positioned closer together, those from the same bank even closer, and those sharing the same email address are nearly identical.

These rich embeddings make it significantly easier to spot nuanced, adversarial patterns of transactions; and to build more accurate classifiers based on both the features of an individual payment and its relationship to other payments in the sequence.

Take card-testing. Over the past couple of years traditional ML approaches (engineering new features, labeling emerging attack patterns, rapidly retraining our models) have reduced card testing for users on Stripe by 80%. But the most sophisticated card testers hide novel attack patterns in the volumes of the largest companies, so they’re hard to spot with these methods.

We built a classifier that ingests sequences of embeddings from the foundation model, and predicts if the traffic slice is under an attack. It leverages transformer architecture to detect subtle patterns across transaction sequences. And it does this all in real time so we can block attacks before they hit businesses.

This approach improved our detection rate for card-testing attacks on large users from 59% to 97% overnight.

This has an instant impact for our large users. But the real power of the foundation model is that these same embeddings can be applied across other tasks, like disputes or authorizations.    

If I’m understanding correctly, this is an early example of how foundation models, built on a transformer architecture, can be trained to replicate the domain-specific experience and intuition of the people who have traditionally built predictive models in financial services.

Put simply, foundation models can be artists. 

{ad_content_secondary}

2 READING RECOMMENDATIONS

A really good breakdown of the news that the CFPB may try to vacate the open banking rule, written by one of the primary architects of the rule.

If you work at a fintech company that relies on open banking (i.e., most fintech companies), you should be concerned about what’s happening here.

Lost amidst all the grandstanding and political posturing that’s happening around the GENIUS Act right now is the fact that the legislation itself has some pretty glaring problems, one of which is the insolvency provisions (i.e., the rules governing what happens if the stablecoin issuer or one of its reserve custodians goes bankrupt). This post from Adam breaks it all down. 

1 QUESTION FROM THE FINTECH TAKES NETWORK

There are a TON of interesting questions being asked in the Fintech Takes Network. I’ll share one question, sourced from the Network, each week. However, if you’d like to join the conversation, please apply to join the Fintech Takes Network. 

How will tariffs impact demand for loans and borrower performance?

This question is inspired by this WSJ article on the impact of tariffs on car buying. Consumers are rushing to buy new cars before prices go up, but lenders are (rightly) terrified of extending loans to desperate car buyers with a potential (tariff-driven) recession on the horizon. 

If you have any thoughts on this question, reply to this email or DM me in the Fintech Takes Network!

HIRING EXPERTS

Through the Fintech Takes network, I meet a lot of smart and experienced fintech operators. Many of them are looking for consulting or project-based work where their very particular set of skills can be put to the best and most efficient use.

Think of them as the fintech version of Liam Neeson in Taken.

I will be featuring anonymized descriptions of these folks in this section of the newsletter. If you are interested in chatting with them, simply reply to this email and let me know!

Thanks for the read! Let me know what you thought by replying back to this email.

— Alex  

Get your brand in front of 40,400+ fintech and banking executives.

Welcome to Fintech 101

I designed this course to give you a foundational understanding of fintech so you excel in today’s job market and thrive in the industry.

Enroll here.

Workweek Media Inc. 
1023 Springdale Road, STE 9E
Austin, TX 78721

Want to ruin my day? Unsubscribe.