a16z Podcast - a16z Podcast: The Evolution of Payments

Episode Date: February 11, 2018

The battle between every startup and incumbent comes down to whether the startup gets distribution before the incumbent gets innovation, oft observes a16z general partner Alex Rampell. But how does this play out when most of the players, big and small, think the innovation has already happened in a particular space? What if there are unmanifested and untapped opportunities in a space? This episode of the a16z Podcast explores these questions through the case study of Stripe. Based on a conversation that took place with Rampell and Stripe co-founder John Collison at our most recent Summit event, the episode covers how the classic battle between startups and incumbents has played out in the payments space; how the broader payments processing landscape has evolved over the past four decades; and what might happen to the established market cap of the "old guard". Stripe is an interesting case study since the company, which was founded in 2010, entered the payments processing scene when the (pervasive) sense was that payments were "done"... and yet at the same time, its co-founder Patrick Collison believed their customers "did not exist yet". So what happened? And how does go-to-market change as a startup evolves, and its mix of customers too changes?

Hi and welcome to the A16Z podcast. This conversation is between A16Z general partner Alex Rampel and John Collison, president and co-founder of Stripe and covers all things payments. They talk about building a payments business, how Stripe's business model has evolved and how the very business of payments has changed and not since the 1980s. The conversation was recorded as part of our summit event in November 2017. I thought I'd open with a funny story from when I first met your brother, because you and your brother co-founded Stripe together. I think this was 2009 or 2010. It was a Starbucks in Palo Alto. And I thought I knew a lot about payments because I've been in the industry for a while. And I asked Patrick, how are you going to get your customers? Where are they going to come from? How are you going to switch people from Chase Payment Tech or from Vantive to Stripe? And I'll never forget what Patrick told me. He said, my customers do not exist yet. Like, that was the answer. And it was remarkably prescient. Because if you think about how Stripe has succeeded into this new generation of like Lyft, like Instacart.

Pretty much any company that's starting today is almost de facto using Stripe. So is that always the plan from day one? We don't know who the customers are. It's almost like product, market, future, tailwind fit? Well, first, you've to remember people get a lot of credit for strategies that, you know, prove to work over time that were necessarily intentional.

I think the lesson we have all learned from the Internet is one of the big mistakes you can make is underestimating how large a market is going to be. There was this pervasive sense that online, it's a way to accept money on the internet. I mean, you probably ran into this too, that it was done, you know, that it was over. And so I think just the fact that we had firsthand experience with the problem as software developers were like, no, no, no, trust me, it's very much not solved.

I think that was maybe the unique insight that we had, which obviously led to things like realizing that, oh, there's all these new startups that are going to be built and those are very different adoption model from existing companies. But how do you try to model that? Because that's also hard and where is it going to go. Have you gotten to the point where you can try to predict future growth? or is it just there are so many different opportunities, at least in your core business of payment processing,

have you adapted since you first started a company? One of the models we've seen that's interesting is we started purely focused on startups, believing that these were all going to be created within startups, and now we're starting to see kind of large companies realizing these changes as well. And so there's almost a bit of a contest where, you know, Instacart could have been started by an existing grocery chain. It wasn't, but it could have been. There was no reason that that was necessarily true a priority.

And so starting to also, as we go up market, getting into helping large companies capture some of these opportunities as well, it's the replacement cycle. If you think about selling to developers, that goes hand in hand with startups, because at a startup, kind of the first two people hopefully are developers. And therefore, you're selling to the CFO, the CEO, the CMO, everybody in that group of two people, and they can just make a decision to go do that. Have you gotten into B2B sales? Because I think that's a challenge for many companies that start off with an API, they start up with developers. But then if you want the larger companies, if it is started by a giant incumbent, you want them to use you. Is that going to come from the developer that's working within the bowels of the organization? Or is that going to come from the CFO?

And did you have to migrate from kind of API sold to developers to actually going kind of more up market or kind of more up in the organization with real B2B to sales? I think the go to market for developer-oriented or API-oriented businesses is actually super interesting, partly because, you know, there's a really well-traveled playbook for SaaS. right? There's so many large companies, the oracles and the sales forces of the world, such that everything is now really dialed and you can read all this expertise. As far as we see it, there is not a playbook for building an API-oriented service and so a lot of that is still emerging. But what we notice is that the sale to startups to new businesses is very natural

because you are selling in a way developer productivity. And startups really naturally value that, right, because they are running against a clock. If they can't get their product to market and find product market fit within a matter of, often months, they're just host. And so it's really important to them to be able to get a product out of market really quickly.

The size of their customer base really matters. The interesting thing is that all these phenomena should be relevant to large companies, too, and that large companies absolutely care about developer productivity, especially as they're building new lines of business where they are often competing with these startups.

And so the adoption mode we see is that, The rip and replace cycle is always naturally so much harder, whereas if you look at a large company and say, okay, they're here right now, where do they need to be in five years' time, they need to go from a static model to a recurring revenue model, they need to go from fully integrated to a marketplace model, they need to change their geographic mix. All these changes that people, that large companies need to make to their business to actually be viable, that's a really interesting conversation to have. And so, you know, the way we look at it with large companies is you can think of the innovator's

dilemma in two ways, right? you know, why did an existing grocery company not build Instacart? One reason might be that the true innovator's dilemma thing of it's actually bad for them if the Instacart world comes true because, you know, this asset that they have, which is the physical retail space, that then becomes commoditized. If you're a large grocer, you know, you're not in the business of building awesome digital experiences. And so that tends to be the go-to-market with the larger companies is where do you want to get to...

An expression that I've used a lot in my various talks is that, you know, battle between startup and incumbent comes down to whether the startup gets the distribution before the incumbent gets the innovation. And this was a mistake that I made when I built trial pay, which was we were like a nice feature on top of payments. And I thought it was a very high gross margin feature. We made a lot of money from it. But actually, you're better off being the plumbing and the pipes. So many companies cannot exist without commerce. You enable commerce. You're that operating system. It is kind of low margin. It is kind of boring. But you do such a good job of it that people stick with you. You store their credentials.

So this was kind of the takeaway for me building trial pay is that if you build the boring layer, and I mean boring in the best way possible, then you have almost unlimited optionality because you're the operating system. So you can build an accounting platform. You could go compete with QuickBooks if you wanted to. You have so much optionality.

What are your aspirations? Because you have this operating system role. How do you think about that? Yeah, it's a classic breath and depth play, right? In a way, this is what Oracle did, where you start out building databases and then just keep asking your customers, is what are you using this database for?

And you build the business applications that people were using the database to build. We saw this with Cloud very notably, where Amazon started with SQS and S3 and EC2, the very basic primitive storage and compute, and then, you know, moved very quickly up the stack to, you know, we're very happy users of AWS. We're very happy users of Amazon Redshift.

Redshift, you know, is a much more powerful data warehousing tool than the basics of just storage and things like that. And so there is a somewhat similar dynamic at Playwood Stripe where we process payments for people. That's not the world's most useful thing. Like, it's necessary, and part of what helps is that it's necessary at day one of the business. I mean, definitionally, what makes, you know, a business as opposed to a product is that it accepts money. But that's generally not the limit of what people need to solve.

The one caveat I'd put on that is this is kind of meme in the payments industry, which I think is wrong, which is the core payments is really boring, and, you know, you as quickly as possible, want to sprint up to other things and just kind of ignore this payment singer. And we just empirically know for a fact that's wrong. I think part of why Stripe has managed to get to the skill we have so far is that we haven't been a bashed about being really interested in payments. There was an old payments world in the 80s where all that mattered was, you know, a merchant-acquirer for Visa MasterCard and American Express.

And that was basically a commodity business and it was really uninteresting. If you have go-to-market via the internet today, payments is incredibly strategic for your product experience and just your revenue growth overall. So Wish is an e-commerce company. They flipped over from a checkout they had built on mobile to Stripe Elements. We spent all this time optimizing a mobile checkout just because it's the interest for us to do and not other people. And they saw a 7% uplift on mobile conversions as a result of switching to this. And so that's the kind of thing that wasn't possible in the 80s when payments was really boring, that now just the core payment side of things can actually get pretty interesting. And that's, by the way, before you go

international. You know, people assume things are homogenizing because that's just, you know, generally being the story of the internet. As more people come online and more countries, as you know, things are heterogenizing where, you know, now you have a Chinese market available to you, but everyone's on wheat chat pay and alley pay rather than credit card. So it's important as you talk about kind of going up the stack from payments, not to ignore that the payments market itself is actually not standard or commodity anymore in the way there was in the 80s. But then, yes, over time, the way we view the opportunity is basically look at what goes into collecting revenue. There's generally many, many full-time people around the

business who are spending time on fairly unleveraged things like managing sales tax for the revenue that we're collecting or revenue recognition for accounting purposes or managing a billing system or all these kind of things. And so it's interesting that in the way that CRM emerged as a category, there was the time when there was no CRM. And then it was like, wait a second, it's really valuable to have this single customer record that you invest in not and keeping hygienic, and then you can kind of do things with that. The way we see the world, there is no revenue platform right now. And if you look at all the products that we have been building, better analytics on the revenue data that you have in Stripe, a source of truth billing system,

things like that, it's moving in that direction. We're not at all there yet, but that's a role. Part of it is kind of cost versus revenue. So the people at a company to bring down costs are not nearly as heroic as those of the company that helped bring up revenue. And I remember when I first met Charlie Sharp, I'll never forget what he said. It's like, how do I get my clients to stop expletive hating me? And my advice to him is, well, send them more customers or increase their conversion, do something for them. Because if you're seen as the IRS, where you're just taking a tax, and they would have paid me anyway. I mean, this might be a fallacious argument, but they would have paid me anyway. And all you're doing is taking a toll. I hate the toll taker. I love the person

that actually brings in more revenue. And I think that's part of the argument as well, which is if you're actually maximizing revenue, you're no longer the toll booth. And you're not the toll booth, right? Like you're kind of the toll booth to the toll booth of Visa and MasterCard and all the other players, I mean, the heterogeneous players, you know, throughout the world that actually are the payments infrastructure. You're at the top of that. And there's also interesting organizational dynamics where any time that you have, you know, one effect, like a, you know, a positive effect that can be measured and a negative effect that can't be measured, you're going to see all these perverse organizational incentives because people track the metric that can be

measured. And so the old school CFO way of doing things is I'm just going to like work on grinding these costs down to zero, and that is my metric. And I think we're seeing across a whole range of industries, including ours, like, wow, this is actually something that we should pay for, right? You know, yes, Amazon is more expensive than racking our own servers, but it would be lunacy for us to do that because it would mean the time to ship any product now is going to be 12 to 18 months. Well, it's interesting. Like, you have some companies that almost conduct a reverse RFP process. So payment processing, who's going to be the lowest cost possible? Or, you know, I'm going to hire a subcontractor from my house. Who's going to tile my roof?

for the lowest cost possible. If you look at Google, one of the advantages and one of the reasons why they are on their way to becoming a trillion-dollar market cap company potentially is that all the advertisers compete to pay Google more money. It's actually reversed. So if you can somehow be on the side of that trade, because it goes to the point of revenue maximization versus cost minimization. I thought I'd ask us. I was checking this morning. So Visa has a $255 billion market cap. I think MasterCard is about 158. Amex is 75. PayPal's about that. You know, So you have $600 billion, roughly, of market cap for the established guard of payment companies. Does that change?

How do you think about that? Because Visa and MasterCard don't actually have any customers of the merchants or of the consumers. So Visa's customers are the issuing banks that actually issue those pieces of plastic that issue those cards. And they have acquiring banks that go acquire merchants that go accept those cards. And they're a middleman. And they've been a very profitable middleman for a long time. How do you think that could change in the next five or ten years? Right. It's kind of two interesting questions. One is what happens to interchange in credit card fees and one is what happens to Visa MasterCard and American Express in particular. And I think Amex may leave them aside. They're different because they're a very diversified business with travel services and things like that. Visa and MasterCard are more interesting. You know, there's a value chain in this stuff where credit card companies and more importantly the issuing banks go out and they get cards in people's hands and they're incentivized to do so via interchange flows. And interchange flows have been very stable.

for a long time, you know, the classic 1.6% for debit cards that came down with Durban and more for credit cards, depending on the category. But they've been pretty stable for a long time. I think the question is, what is the value that these issuing banks are providing relative to what they're charging? And the value probably changes over time, right? Because, you know, it's interesting. You go to developing countries and there's lots and lots of out-of-home advertising going on for just like, get a credit card. And, you know, credit cards are a good thing that you should have. Whereas, of course, in the U.S., everyone has six credit cards and it's a much more mature market. Let's be clear, it's a really valuable service, getting people from kind of the cash

world to, you know, having a payment mechanism in their pocket that, you know, is more or less fraud-proof for the customer and that they're confident using it everywhere. That's a really valuable service generally and a really valuable service for businesses in particular. I'm not sure that the rates should be so steady over time. And so what you're seeing... Well, the interesting thing there is that it's kind of a double moat that protects them because not only did they have a network effect, but they also have this bizarre incentive where if a 2.5% fee is being charged to the merchant, well, the consumer is being bribed with maybe 1.5% of that. Exactly.

So actually, higher fees yield higher rewards for the consumers. And actually, you know, in Australia, where interchange went down dramatically, it's like 50 basis points, rewards went away. You didn't get your miles anymore. And then it was a giant transfer of wealth from the banks to the merchants, but none of that went back to the consumer. Well, it presumably did in the form of prices at some point. But yeah.

Sometimes. I think that's absolutely what you're. seeing is that in a way, I think the prices are too stable, and so it ends up being really stable at a level for a certain point, and then it gets regulated. And so Australia regulated interchange, Europe famously regulated interchange, even the U.S. regulated debit interchange, and just empirically, if you kind of draw the lines on a curve, it seems like that's the direction. Well, because it's very hard for market forces to contain problems of concentrated benefit and diffuse harm. Yep. And that would be one. The diffuse harm is that you're overpaying

buy a nickel or by a penny or a fraction of a penny. And actually, that's why the merchants have benefited more than the consumers in a place like Australia, because do you care if you're paying 99 cents for a Coke or a 98.7 cents for a Coke? And that's that diffuse harm. Whereas the concentrated benefit is that 0.3 cents to Walmart of Australia times a million, that actually is very, very meaningful. So it's hard to disrupt from a market forces perspective because consumers are actually worse off when pricing for credit cards go down, ironically. But to your original question way at the beginning, I mean, Visa and MasterCard are very,

the amount that they actually earn themselves, you know, if you do the math is about 20 or 30 basis points of each transaction, which I think kind of for them is a fairly stable amount for the value that they provide globally. And, I mean, there's a stat recently that Visa uses 1.40th of the power of the Bitcoin network for processing, obviously, you know, a way large. amount of money. I'm surprised it's that low. Yeah, yeah, yeah. Or maybe put another way, Bitcoin is kind of inefficient enough that it uses 40 times the power of the Visa Network, despite kind of a much

lower throughput than the Visa Network. And so these are kind of very useful services broadly. If I was an equity investor, there's no way I would be short these companies because you have the secular growth of just like digital payments. They're killing cash. Yeah. So a lot of people talk about data. If you talk to any big company, they want to use data. Actually, that was 15 years ago. Now they want to use big data. This big data is bigger than regular data. Yep. Super-sized.

Payments have tons of data. Like, you could predict economic, macroeconomic, microeconomic progress if you have payments data. I mean, how do you think about data in payments? Like, what do you do with it? What should be done with it? Do you want to do more with it? It's really easy to draw these theoretical examples and like, oh, we can use this

signal or whatever. And my favorite thing to do now is to like push people on, well, do you actually use that signal? It's like, well, no, no, you know, we did an experiment or whatever. But the really common behavior you observe is this really interesting speculative uses of data and a really narrow, boring actual set of uses of data. So, say, in payments, we've been talking about using new fraud signals for literally decades. And if you look at how the fraud systems that actually back your credit card work today, what happens is, you know, so you have a chaser, Bank for America, or whatever credit card. And there's a set of neural networks that were trained.

And you're like, oh, cool, neural networks. that sounds really awesome and modern, but not from kind of the current wave of neural networks, but when they were originally first built-in product eyes back in the 1960s. And that's basically when these models were trained with a handful of features like time of day and who this merchant is and amount and behavior and stuff like that. And the systems more or less haven't been updated since.

And that's why, you know, the experience of using your credit card out and about is so frustrating. It's like you know that I'm traveling. I have your app installed on my phone. I use this card at an ATM, you know, whatever it is. And so when it gets kind of clobbered and frozen by the bank, that's why that experience is so frustrating. And so the way we think about data is not only what is the data available to people, but how you make it useful. We have much more interesting data that we actually make available to our engineers.

And so when we're looking at blocking fraud, we're doing stuff like, you know, looking at the, you know, we have JavaScript running on the end customer's web browser. And so we can tell a lot more about the veracity and is this customer who they claim to be than your traditional bank that's prior who is trying to make a fraud decision. Then the flip side of it is making data useful actually on behalf of our customers. And so you're putting all this data into Stripe. And the traditional thing that you get from a payments company is like a paper statement that gets put in a drawer at the end of the month. And that's the extent to which it gets used. And so now we give them a full SQL console across all of the data they have in Stripe.

That's often faster than their own data warehouse or, you know, ETL pipelines because we can keep it up to, you know, we are building a data warehouse for all our customers across the hundreds of thousands of people that use Stripe. and then we can build products on top of that that are more useful. So it's both making the data that theoretically exists actually useful and used, both internally for things like our fraud management purposes, but then also for our customers, given that they're putting what is maybe the most interesting data set of the business,

the revenue data in Stripe. What do you think the biggest opportunity adjacent to payments is? What are your compliments with an E that you would like people to develop? Or where do you think there are opportunities? I think it was your colleague Chris Dixon who wrote a blog post about the fast startup, the idea that people are taking a distribution layer or technology layer and being willing to own more of the stack behind it. And it feels to me, just looking around me, that we are very early in that transition because, you know, it's always this age-old question.

Can incumbents get good at technology in the new world faster than a new startup can come along and get, you know, one, get distribution, but two, get good at what it is that incumbent previously did? And so can a challenger bank come along and, you know, get regular as a bank and, you know, bring that to market faster than existing banks can make their app good. Can, you know, a new startup come along and enter the healthcare space faster than existing healthcare companies can get their act together? And that battle is still playing out across so many different industries. And it's often pretty hard to tell at priori. And so that's the space that we end up watching pretty closely. All right. Great. Well, John, thank you very much.

Thank you so much.