A couple of months ago the Bank Policy Institute held its annual Fintech Ideas Festival in San Francisco, gathering leaders from the largest banks and fastest-rising startups to share and debate the issues facing the industry. Our CEO Douglas Merrill (left) and Plaid CEO Zach Perret (center) were on stage at the event with CNBC tech reporter Ari Levy to talk about how Zest and Plaid represent a new kind of fintech company -- one that doesn’t compete directly with banks and fintechs but partners with them to increase profits and improve customer experience. Here’s the transcript of their conversation, titled “Zack and Douglas’ Excellent Adventure.”
Ari Levy: You guys are here for Zach and Douglas's excellent adventure. And you're not Keanu Reeves and Ted Winter in a bad high school heavy metal band.
Douglas Merrill: I am wearing Iron Maiden tattoos though.
Ari: So there you go. The Bill and Ted analogy is not going to work completely, but you are two guys with long hair trying to modernize consumer finance, so there is some of that narrative thread. Let's start with how you got there: Douglas, you were CIO of Google. You spent some time in the music industry before starting an online lender. And Zach, you were in consulting. As you're thinking about starting a company, why finance?
Douglas: I've been lucky in my career to be hanging out at places that were solving really interesting problems. And for me, interesting means mathematically really hard, and it matters if you get it right. And I've been sort of wandering around after I left Google doing some random stuff, including running a record label. And I realized that there's sort of an interesting problem that had occurred since about 1950: Underwriting hasn't changed at all.
These two guys, (Bill) Fair and (Earl) Isaac, popped out of academia and realized they could use what — at the time — was the most advanced mathematics available, logistic regression, plus the data from these credit bureaus and produce a generic score. Arguably, that's one of the top five innovations of the 20th century, although nobody seems to praise them for it. They should be praised for it, but things shouldn't be the same since 1950. Nothing else is.
Could we apply the math that we built at Google to the problem of credit underwriting and see value? And it turns out, you can and it does.
Computers have grown, right? Computation is free and infinite. Storage is free and infinite. Bandwidth is free and infinite. Why shouldn't you do something more interesting? And so I wondered: Could we apply the math that we built at Google to the problem of credit underwriting and see value? And it turns out, you can and it does.
Ari: Zach, what's your story? How'd you end up in the backend of the banking industry?
Zach Perret: My story is a little different. It's a bit more accidental. Early in my career, I spent a ton of time in software engineering, writing code, and then landed at Bain, the consulting firm, and got exposed to working with financial institutions. And there's this big divide between the work that we were getting paid to do at Bain and my history in software. And you would just see time and time again these problems that came up with financial institutions, with banks that should be solved by software, and yet weren't as much.
Along the way, I started spending a lot of time thinking about the impact that finance has on consumers' lives. If you ask a consumer, what's the most stressful thing in your life? Time and time again, they say money is the most stressful thing in my life. It's the thing that stops me from doing what I want to do. It's the thing that stops me from having the life I want to have, or spending time in the way that I want to spend it. It's the limiting factor.
When I ended up leaving Bain, my co-founder and I set out to build a set of products that enable consumers to live better, simpler, easier financial lives. Allow them to control the way they spend their money, the way that their money is used, so on and so forth. Consumers at the time, and this was seven or eight years ago, they had a lot of distrust of the existing financial products that are out there.
Our original push was to create our own version of a consumer financial product. How do you monitor your money better? How do you track your spending? How do you build your budget? We found that it was impossible to build that because there was no infrastructure. In order to do anything for our consumer, you either had to be a bank and have access to all the rails, or you had to try and build your own custom version of infrastructure, and then maybe you've created this one given product. And so, we decided to build a platform.
Ari: There were other consumer finance startups out there who were building on top of existing infrastructure and getting customers, getting traction. Right?
Zach: But how many? I mean there was PayPal. There's Mint.
Ari: When are we talking? What year was this?
Zach: Seven, eight years ago.
Zach: I mean there were certainly others. Venmo had been launched a little bit before then and having trouble getting user adoption, especially as the ACH onboarding was pretty difficult, and they were using credit and debit cards. Tons of other products were out there in the nascent phases but weren't able to truly hit scale. And, I would say, we frankly got a bit lucky to be pivoting into building this infrastructure. But the thread that unites our company is we want to make money easier for everyone. You want to make money easier for consumers, and a great way to do it is to build a platform that people can use all over the place — banks building better bank products, non-banks building better fintech products, both.
Ari: I think what we've seen — and you'll correct me if I'm wrong, and you'll probably think I'm wrong to some degree — in fintech startups the last couple of years, at least the more successful ones, is they grow in a particular market, largely through a millennial or a younger audience, possibly through a lending product, money management, stock trading or payments, and then they start to branch out into banking services. And then, all of a sudden, they want to collect deposits. And then they want to be a bank, but they're not going to get an FDIC license. You sort of get this image of you can do one thing and do it well but eventually, you have to expand, and that becomes harder. Are we going to see startups turn into these full-service banks?
Douglas: I guess my answer would be: Good Lord, I hope not. Go back to Frederick Winslow Taylor in the early 1900s. The theory of the corporation about transaction costs and using corporate structures to minimize transaction costs. Because transaction costs used to be quite high, these big multiline corporations and multiline banks came together because it's just too hard to make it work otherwise.
Transaction costs today are, to first-order, zero, right? Why wouldn't you let companies specialize in what they're good at? I'm not going to become a bank. I'm not good at taking deposits. I don't know how to do that. I'm not good at checking accounts. I don't know how to do that. I'm really good at selling underwriting tools, and the banks in the room should get to do what they want to do, and I'll do what I do to create value.
I'm not good at checking accounts. I don't know how to do that. I'm really good at selling underwriting tools, and the banks in the room should get to do what they want to do, and I'll do what I do to create value.
Zach: We work with about 2,500-3,000 financial products out there, and we see tons of different fintech products built by everyone in the ecosystem. And over time I’ve developed this theory that, if you look at what a bank was five, 10 years ago, it was a monolith, or is a monolith. Most of the big banks, they have every product for every person. You can walk into a bank and say, “Hey, can I have a checking account to hold my $25? Can I have a mortgage to buy a house? Can you help me plan for retirement? Can you give me a business loan? Can you give me really everything?”
The shift that we've seen over the past five years or so has been going to these more microservices: There's a specific product that will let you get a $2,500 loan; there's a specific product that'll let you get a mortgage; there's a specific product that'll let you pay a friend, make an investment, buy a bitcoin or whatever it is. And when you're in the monolith, all of the demographics went to the monolith because that's where you went. Now, when you have a bunch of different microservices, the demographics are going differently. They're distributing differently.
What we're starting to see is a lot of these point services, a lot of these microservices are starting to rebundle, but they're rebundling around a specific need for a specific demographic. It's for people of a certain income level, certain preference or certain investment philosophy. The rebundling is occurring in these smaller areas, so we do see a lot of it. I don't necessarily know that any of the desire of those applications is to build a full stack bank. I think it's really just to serve their demographic well and giving them the products that they need to do so.
Ari: Do you envision some of these companies becoming deposit-taking institutions along the way, whether it's to become a full-service bank or not?
Zach: I certainly imagine that they will. Whether or not it's their own deposits, whether or not they're doing the rent a charter thing, whether or not they're partnering with another financial institution to enable that. A speaker earlier today talked about every company becoming a fintech company. And Uber, for example, they want to give their drivers a bank account. Uber's not going to be the one to give them a bank account because Uber's not a bank. They're going to partner with someone to give them the bank accounts, so it's more about distributing the functionality than it is about who holds the deposits.
Ari: Douglas, you've been doing Zest for about a decade now, right? I remember talking to you early on about the efforts to do underwriting based on data that banks don't use. Whether it's social data, other data out there that we all have, but has not made its way into credit scoring today. And I'm going to completely butcher what you do for a living, but you take all of this data, you create models from it, and you do underwriting based on those models. And the models get smarter over time, and you get better underwriting. Where has been the biggest hurdle to get from where you wanted to be a decade ago to where we are now?
Douglas: You did mangle what I did for a living.
Ari: My intention wasn't to mangle it, having only been speaking to you for a decade.
Douglas: Zest sells tools to banks to help them get machine learning models into production. Every bank in the world has some number of people sitting somewhere doing ML modeling, and they're probably doing a pretty good job of it. Modeling is a little bit of a solved problem. But the problem is, they can't get from that lab into production. And what my tools do is make it possible by using clever AI to take a model someone just finished building, pass it through model validation and your regulators, and get it into production. We use a bunch of AI to manage a bunch of AI.
The biggest barrier… and I think you're maybe looking for me to complain about the regulators, which has not been my experience. The regulators have been quite accepting and clearly trying to fulfill their mission of serving the country as an honor, as opposed to as a job. I think ultimately the biggest barrier has been the internal structures of our clients.
I was a CIO at Google for a very long time. One of the things I tried to do is figure out what governance should look like for technology inside a pure technology company. And ultimately, what I tried to do is decentralize everything I could, except the core things that were shared across the entire company, so a couple of different data stores, different search algorithms. And as we walk into companies that still have traditional CIOs, it's sometimes hard for them to figure out, well should they do this work or not? Because they don't own the value. They just own the cost. And so, I think that's the biggest barrier we've had.
Ari: So getting customers to understand the value of what it is that you're providing, and how that can help their underwriting or how their lending process?
Douglas: As part of the tool implementation, you get a model that says, here's how much you could save if you took out all the losses. Most people don't. Most people take it all to approval rate, and our customers get between 15% and 30% increases in approval rates with no increase in losses. They get this massive win, and they already marketed to all those people, so those are basically free accounts. But you know what the economic return is going to be. The business folks are like, “Hey, I want that. Thank you very much.” You show up at the CIO, and the CIO is like, “OK, to implement this, I'm going to have to go touch my loan origination service. I'm going to have to go touch a bunch of data. If I mess this up, I'm going to lose my job, and if I get it perfectly, that person's going to get all the benefits. Why should I do this again?” Which is not a totally unreasonable question.
Ari: What about the ability of machines to underwrite? Where are you on that now? And where do you think we're going?
Douglas: It’s easy to teach a machine to do something dumb. You probably shouldn't do that. For example, with one of our early clients — just for fun we were trying — we said what if you let the machine do it all? Let the machine select the variables. Let the machine build the models. Let the machine do everything. And we got this model, right? That is what you get. And that model predicted that you should approve every loan coming from Arkansas. I'm from Arkansas.
Ari: As we both are.
Douglas: Arkansas is a lovely place full of very nice people, great vacation spot. Not a place you should give all loans to. And it just turned out that there was exactly one loan we made in Arkansas, and that person paid back, therefore, 100% of loans to Arkansas paid back, so you should make more loans to Arkansas. You can get a model to learn something dumb. You probably shouldn't. Part of what we spend a lot of time trying to do is make models that aren't dumb.
Ari: Zach, talk about banks as friend or foe for how some of your product can go to market now.
Zach: I think it's an interesting question. You can even almost abstract to all of fintech. We are, again, enablers of fintech. We build the infrastructural platform that allows the consumer to interact with their money, interact with their data in the place that they want to. Many of our biggest customers are banks building what I think is pretty cutting-edge technology for their customers. We, of course, work with a lot of nonbanks as well, so it's par for the course that we spend a lot of time with everyone.
But I think the interesting thing that we started to see more of are partnerships between fintech companies and financial institutions. When you think about where a consumer spends their time and who has the largest consumer base, it's the banks. The biggest banks have 30-, 40-, maybe even 50-million consumers, and that is an amazing distribution point for any financial product.
We are seeing this increasing pace of partnership between fintech and financial institutions, and I think that's really exciting because ultimately, the consumer is winning. The consumer is getting more access to more products.
What we've seen time and time again over the past couple of years are examples of the banks themselves wanting to go and license or partner with a fintech company, and distribute that product into their user base, and do something like a profit share or revenue share or value share. Because the consumers are saying, “Hey, I want that fintech product,” and the bank is saying, “Well, I could build that or I could partner with that or maybe I'll buy it, and I can get it to market faster.” We are seeing this increasing pace of partnership between fintech and financial institutions, and I think that's really exciting because ultimately, the consumer is winning. The consumer is getting more access to more products.
Ari: How much resistance have you faced from banks in trying to deploy?
Zach: I think in the early days we faced a lot. We sad, “Hey, we'd like to help create a more tech-savvy ecosystem. We'd like to go shake things up. We'd like to change things.” And in an industry where many institutions make a lot of profit from things being the way that they are, it was a little bit controversial. What we saw over time is just repeating the consumer-centric story, saying ultimately, everything goes back to adding value to the consumer's life, and helping the consumer take control of their money, make the decisions they need to make, get better products, tools, and services. It's incontrovertible. No bank would ever say, “I don't want to help the customer.” Certainly, I would hope they wouldn't. No product that I've seen would say that either. Repeating that story over time opened doors for us.
Ari: Venture capitalists have invested roughly $131 billion last year, a record number by far. Fintech has seen its share of that. Softbank has been writing significant checks, billion-dollar-plus checks, into the industry. What does the influx of capital mean for both your companies, your competitors, the competitive ecosystem? Any insights there?
Douglas: We haven't taken very much money. We've been privileged to run on our operating value for quite some time.
Ari: When was the last time you raised?
Douglas: Three or four years ago. And we're lucky to have amazing partners on our board. I like my board meetings, which is unusual. I think what's interesting is that a lot of money has been cast at not great companies. A Financial Times article two weeks ago cited a survey that found that 40% of European AI companies had zero AI in their infrastructure.
Marc Andreessen famously said that software is eating the world. A friend of mine from Google named Sergei Bezrukov said it's BS that is eating the world. And he went off and did a survey and found more like 80% of AI companies have no AI in them. I'm looking forward to the hype cycle being over. I'm looking forward to there being less money put into fintech. I was in the first AI bubble 25 years ago, and when money started flowing in, productivity and quality drop. But when the money went away, both increased. I'm sort of looking forward to that.
Zach: We just raised a quarter of a billion dollars.
Douglas: A lot.
Zach: And hopefully, we are not one of these companies that you speak of.
Ari: You don't claim to have AI.
Zach: Yeah, we don't claim to have any AI. We're plumbers, right? Financial plumbers. We raised a fair amount of capital, and I think though, on the backend, we've been quite efficient with the use of capital. Relative to revenue and to most metrics, the team is still pretty small and lean. And we like that. I think it keeps you healthy.
I think just speaking a little bit more broadly about the ecosystem, having a bunch of capital means there's a bunch of different bets being placed. What we've seen is that the traditional way that you place bets: you launch a product, then you go try to acquire customers, and you're spending a lot of money on advertising to acquire all those customers. The traditional path of getting big for a fintech company, that's getting hard. It's getting really hard. Customer acquisition costs through paid advertising used to be, when Instagram first launched their advertising platform, it was amazing. You could go out and buy all these viewers you wanted to for basically nothing. Turns out Instagram's now very expensive, and all of the other paid platforms are very expensive.
What we're seeing is that the fintech companies that are able to acquire through referrals, through organic, are growing so well, sustainably, and for a long time. And those that are just paying are not. And, going one layer deeper, what that tells you is that the companies that are able to focus on acquiring within networks, creating networks of people, again, this demographic push, if you can get a demographic right, you can get people to tell their friends about it. That's a really big deal. And people that are building the best products are continuing to win.
I would say, in a capital-rich or a capital-poor environment, we might see more bad bets. We might see more bad choices in a capital-rich environment. But the companies that win are still the same. It's still about product quality. It's all about serving the consumer in the best way you can.
Ari: So we can honor the spirit of the session, let's imagine that you guys are Bill and Ted, and you have access to Rufus's time machine. Rather than traveling back in time to collect historical figures so you can save the future with your history exam, you want to save consumer banking. What's one thing that if you could go back in time and change to save consumer banking, what would that be?
Douglas: Wow. If I could go back in time, and I wouldn't say "Excellent" very many times, but I would change how we build loan origination service systems on scarily bespoke platforms and have them on more of a shared instance. I don't care which one, to be clear. It would make it easier for companies like Plaid. It would make it easier for companies like us. It would make it easier to unlock the value out of that bank.
Zach: We heard Jamie Dimon say recently that JPMorgan is a fintech company. They are a tech company. And that represents a shift in mindset. Whether or not it's true, I'm not commenting on. But the fact that banks are now thinking of themselves as tech companies is the biggest thing. Going back 10, 15 years, saying banks should be thinking of themselves as technology companies, that would be huge. When you look at the platform cycle, the shift to the Internet, the shift to mobile, financial institutions are three, five, seven, 10 years behind — in general — the rest of technology. Going to the banks and saying, “You are a technology company. Please try to keep up.” We're starting to see those cycles shorten, so banks are now getting much closer to keeping up with the platform shifts but getting that message out more broadly a little bit longer ago.
Ari: Excellent. Thanks, guys.