Financial startups are getting an edge by growing up

No industry is immune to technological change and disruption, and for many years billions of dollars of investment capital have been poured into financial technology, or fintech. The thesis is simple: finance is a massive industry which moves incredibly slowly. Just about everybody, whether they’re inside the industry or outside it, sees enormous inefficiencies and room for improvement. And they know that if they wait for the billion-dollar banks to do the innovating, they’ll be waiting a very, very long time.

At some point in the past couple of years, however, an important shift has taken place. In Beatles terms, we’ve started moving from “you say you want a revolution” to “you tell me that it’s evolution.”

The shift was captured quite nicely in Nathaniel Popper’s bitcoin book Digital Gold: while bitcoin’s past is a story of dreamers and idealists, its future lies in the hands of billionaires and massive financial-services companies.

Similarly, Bernard Lunn, in his fintech predictions for 2016, includes this: “The Great Convergence between Banks and Fintechs commences, as both get judged on the same metrics by consumers, regulators and investors.”

Most importantly, the fintech startups themselves are increasingly judging themselves on the same metrics that are used by much more established companies. They’re looking at old-fashioned things like their balance sheets and their income statements, rather than baroque indicators designed to make themselves look good to angel investors with dollar signs in their eyes.

“You can’t build a bank or a payment system or a lender with a semipermanent ‘we never need to make any money’ Silicon Valley view of the world,” says Max Levchin, the PayPal co-founder who now runs one of the hottest fintech startups, Affirm. “You have to build a profitable business, and you have to be very careful about not pushing the envelope beyond what is responsible and intelligent.”

For the kind of people who think technology is all about innovation and cool new things, then, fintech is about to get even more boring. For a while it showed occasional flashes of inspiration: bitcoin, of course, but also things like crowdsourcing, which had never really been possible before.

The fun part, however, is now over. Fintech has matured to the point at which all the value lies in iteration and execution, rather than in some kind of revolutionary new product or idea.

To be honest, ‘twas ever thus. Look at the way in which peer-to-peer lender Lending Club took an early lead over Prosper, by discarding feel-good notions about “the wisdom of crowds” and working instead on building a smart and powerful underwriting engine which could predict just how likely any potential borrower was to repay his or her loan. Even back in the early days of PayPal, the real value creation and technological wizardry was much more centered on building sophisticated fraud detection algorithms than it was on the nuts and bolts of sending money from A to B.

To see how these things play out in practice, look at two companies founded at roughly the same time, Simple and Stripe. Both of them, initially, wanted to be banks, but both of them realized that was a little too ambitious. So they went in different directions: Simple became a beautiful mobile-native banking interface, while Stripe became an elegant way for anybody to be able to collect payments online.

Of the two, it’s Stripe which became the celebrated unicorn; Simple ended up selling to a Spanish bank for a relatively modest $117 million. But both of them are still growing fast, with very little in the way of product development visible to the naked eye.

What both companies have discovered is that you get real customer growth just by obsessively honing and optimizing the processes that take place under the hood. Your customers, old and new alike, are unlikely to be conscious of the changes you’ve made, but the changes still make a big difference. Just ask Amazon, which worked out long ago that people buy measurably more stuff when you accelerate page-load times, even by imperceptible amounts.

Simple, for instance, says that it is currently growing at 10% per month, with its highest-ever customer satisfaction ratings. How is it driving that growth? Mainly, by paying attention to the feedback loop. Simple has developed a reputation for friendliness and responsiveness, which means that its customers feel comfortable talking to its employees about what they want and need. Those requests can represent big or little pain points, and then Simple can get to work ensuring that whatever it is those people wanted to do is doable, easily and intuitively within the app.

Simple is working on big projects too, of course, like moving their bank from Bancorp to BBVA, and (hopefully, eventually) offering joint accounts. At some point, there might even be loans. Still, it’s clear that smaller tweaks can be surprisingly effective on their own, even without larger-scale changes.

The ultimate feedback loop, however, is the one you find at lenders. They have an algorithm, they use it to determine whom to lend to and at what interest rate, and then they discover, in the months and years after the loan is made, who paid back the loan and who didn’t. Every payment a lender receives, whether it’s principal or interest, is (or should be treated as) a valuable datapoint; every default is an even more valuable datapoint. You can see which rates attract the most customers, you can see which rates generate the most profit; you can see which deals go wrong and which go wonderfully right. And so by rights your system should be able to get better and better every day. (This is one of the reasons why Warren Buffett is such a huge fan of Wells Fargo and American Express: he knows just how hard it is for a new competitor to come in and compete with a financial-services company with a century of institutional knowledge and experience.)

Historically, this process has been constrained by the way companies fund themselves. Lending Club, for instance, grew at an amazingly constant rate in its early years, and quite deliberately didn’t grow any faster than that. (It had various customer-acquisition dials it could turn up any time it wanted to, and didn’t.) The reason? It was acquiring lenders along with borrowers: Lending Club just sits in the middle between the two. If it grew too fast, then a very large proportion of lenders would be exposed to a single vintage of loans, and if that single vintage went bad for whatever reason, it would pose an existential risk to the whole company.

Younger lenders, however, like SoFi and Affirm, don’t have that problem. That’s because they have become vastly more sophisticated (and also much more secretive) in how they fund themselves. If you’re a consumer lender, then your assets are never going to be particularly complicated: they’re just going to be a portfolio of consumer loans. The liability side of your balance sheet, on the other hand, can be real rocket science. These companies might well be issuing hundreds of millions of dollars in equity, but that’s not all they’re doing. They have a highly complex mix of equity, preferred stock, secured debt, unsecured debt, and even, in some cases, more exotic instruments.

For instance, while Lending Club sits in the middle of a two-sided market (the borrowers on one side, and the lenders on the other), Affirm is in the middle of a three-sided market, involving borrowers, lenders, and merchants. Or even four-sided, if you include Cross River Bank, which originates all the purchase loans. Investors, using a combination of debt and equity and various hybrid instruments, can advance money to merchants as well as borrowers; meanwhile, merchants can subsidize loans in order to get more order flow. (If you buy a Casper mattress interest-free using Affirm, for instance, Casper will essentially make your interest payments for you, out of its profit on the mattress.)

Who are Affirm’s customers? Some are hedge funds, some are venture capitalists, some are merchants, some are individuals buying stuff. With the exception of malign actors who try to defraud the company, Affirm has to keep them all happy. At the same time, it must keep a close eye on all those moving parts so that its core business is profitable, and also make sure that a whole suite of regulators stay happy with what it’s doing. It’s a deeply complex computational problem.

“There’s an infinite number of variables that you can track and fine tune all the time,” says Levchin. His strategy is to out-compute the other participants in the market: “at some point you get to the level of quality or sophistication where no competitors can hope to get close to you because you’re just 30% better on every dimension,” he says.

Something that looks incredibly simple from the outside—for instance, a payment plan that allows you to pay for your purchases over time, and is easy to sign up for from your phone—can turn out to be insanely complex behind the scenes. That’s especially true when online merchants are willing to pay surprisingly steep fees, if doing so substantially increases the rate at which they convert shoppers to buyers.

We live in a world where millions of Americans have essentially no liquidity, and where a whole generation has grown up rationally suspicious of credit cards. Companies like Simple and Affirm and Klarna, which also makes it incredibly easy to buy things using your phone, are building a world where those Americans can transact painlessly on a day-to-day basis, can purchase big-ticket items without incurring nasty revolving debt, and can trust some of these new-generation companies to do the right thing, even in an industry which loves to prey on the weak.

But getting there, at scale, is tough. And it requires a lot of money, so that these companies can run thousands of concurrent real-time experiments.

The problems such fintech companies are trying to solve aren’t the type that can be tackled by a few hyperactive coders in a garage. Rather, they require dozens of different skillsets, not to mention the ability to manage them all. In that sense, the startups are becoming much more like the banks they’re seeking to disrupt. That’s Lunn’s Great Convergence. No one believes the banks are going to solve these problems. The trillion-dollar question is, can the fintech companies do something important and socially useful before they, like the banks, become bogged down in regulation and bureaucracy.