00:00hi and welcome to the a 16z podcast this
00:03conversation is between a 16 seed
00:05general partner Alex or impel and John
00:07Collison president and co-founder of
00:09stripe and covers all things payments
00:12they talk about building a payments
00:14business how stripes business model has
00:16evolved and how the very business of
00:18payments has changed and not since the
00:201980s the conversation was recorded as
00:23part of our summit event in November
00:242017 I thought I'd open with a funny
00:27story from when I first met your brother
00:29because you and your brother co-founded
00:31stripe together I think this was 2009 or
00:332010 it was a Starbucks in Palo Alto and
00:36I thought I knew a lot about payments
00:38because I been an industry for a while
00:40and I asked Patrick how are you going to
00:42get your customers where are they going
00:43to come from how are you gonna switch
00:45people from Chase Paymentech or from Van
00:47t'v to stripe and I'll never forget what
00:49Patrick told me he said my customers do
00:51not exist yet like that was the answer
00:54and it was remarkably prescient because
00:56if you think about how stripe has
00:57succeeded into this new generation of
00:59companies like lyft like instacart be
01:02pretty much any company that's starting
01:03today is almost a factor using stripe so
01:06is that always the plan from day one we
01:08don't know who the customers are it's
01:09almost like product market future
01:11tailwind fit well first you've to
01:13remember people get a lot of credit for
01:14strategies that you know prove to work
01:16over time that weren't necessarily
01:18intentional I think the lesson we have
01:20all learned from the Internet is one of
01:21the big mistakes you can make is under
01:23estimating how large a market is going
01:25to be there was this pervasive sense vas
01:28online it's a way to accept money on the
01:30inside I mean you probably ran into this
01:31- that it was done you know that was
01:33over and so I think just the fact that
01:36we had first-hand experience with the
01:38problem as software developers we're
01:39like you know trust me it's very much
01:41not solved I think that was maybe the
01:43unique insight that we had which
01:44obviously led to things like realizing
01:47that oh there's all these new startups
01:48that are going to be built and there's a
01:49very different adoption model from
01:50existing companies but how do you try to
01:52model that because that's also hard and
01:54where is it going to go have you gotten
01:55to the point where you can try to
01:56predict future growth or is it just
01:58there's so many different opportunities
01:59at least in your core business of
02:01payment processing have you adapted
02:03since you first started the company one
02:04of the models we've seen that's
02:05interesting is we started purely focused
02:08on startups leaving this you know these
02:10are all going to be created within
02:11startups and now we're starting to see
02:13kind of large companies realizing these
02:15changes as well and so there's almost a
02:17bit of a contest where you know insta
02:19cars could have been started by an
02:21existing grocery chain it wasn't but it
02:23could have been there was no reason that
02:25was necessarily true a priority and so
02:27starting to also as we go up market
02:29getting into helping large companies
02:31capture some of these opportunities as
02:32well it's the replacement cycle if you
02:33think about selling to developers that
02:36goes hand in hand with startups because
02:37it's a start-up kind of the first two
02:39people hopefully are developers and
02:40therefore you're selling to the CFO the
02:42CEO the CMO everybody in that group of
02:44two people and they can just make a
02:45decision to go do that
02:47have you gotten into b2b sales because I
02:49think that's a challenge for many
02:50companies that start off with an API to
02:51start off with developers but then if
02:54you want the larger companies if it is
02:56started by a giant incumbent you want
02:58them to use you is that going to come
03:00from the developer that's working within
03:02the bowels of the organization or is
03:04that going to come from the CFO and did
03:06you have to migrate from kind of API
03:07sold to developers to actually go in
03:10kind of more upmarket or kind of more up
03:11in the organization with real b2b to
03:13sales go to market for developer
03:16oriented or API or into businesses
03:18actually super interesting partly
03:20because you know there's a really well
03:21troubling playbook for SAS right you
03:23know there's so many large companies the
03:25Oracles and the sales forces of the
03:27world such that everything is now really
03:29dialed and you can read all this
03:30expertise as far as we see us there is
03:32not a playbook for building an API
03:34oriented service and so a lot of that is
03:37still emerging but what we notice is
03:38that the sale to startups to new
03:41businesses is very natural because you
03:43are selling in a way developer
03:46productivity and startups really
03:49naturally value that right because
03:50they're running against a clock if they
03:52can't get the product market and find
03:54product market fish within a matter of
03:55you know often months they're just hosed
03:57and so it's really important to them to
04:00be able to get a product out of market
04:01really quickly the size their customer
04:03base really matters the interesting
04:04thing is that all these phenomena should
04:07be relevant to large companies to and
04:09that large companies absolutely compare
04:11about developer productivity especially
04:12as they're building new lines of
04:14business where they are often competing
04:16with these startups and so the adoption
04:18mode we see is that the rip and replace
04:21cycle is always naturally so much harder
04:23whereas if you look at a large company
04:25and say okay they're here right now
04:27where do they need to be in five years
04:29time they need to go from a static model
04:31to a recurring revenue model they need
04:33to go from fully integrated to a
04:34marketplace model they need to change
04:36their geographic mix all these changes
04:38that people that large companies need to
04:39make to their business to actually be
04:41viable that's a really interesting
04:42conversation to have and so you know the
04:44way we look at it with large companies
04:45is you can think the innovators dilemma
04:47in two ways right you know what I did an
04:49existing grocery company nós build insta
04:52cars one reason might be thus the true
04:55innovators dilemma thing of it's
04:57actually bad for them if the instacart
04:59world comes true because you know this
05:01assets that they have which is the
05:03physical retail space that then becomes
05:05commoditized if you're a large grocer
05:07you know you're not in the business of
05:09building awesome digital experiences and
05:10so that tends to either go to market
05:12with the larger companies is where do
05:14you want to get to and especially I've
05:16used a lot in my various talks is that
05:17you know battle between start-up and
05:19incumbent comes down to whether the
05:20startup gets the distribution before the
05:22incumbent gets the innovation and this
05:24was a mistake that I made when I built
05:26trial pay which was we were like a nice
05:27feature on top of payments and I thought
05:30it was a very high gross margin feature
05:31we made a lot of money from it but
05:33actually you're better off being the
05:34plumbing and the pipes so many companies
05:37cannot exist without commerce you enable
05:39commerce you're that operating system it
05:41is kind of low margin it is kind of
05:44boring but you do such a good job of it
05:46that people stick with you you store
05:47their credentials so this was kind of a
05:49take away from me building trial pay is
05:50that if you build the boring layer and I
05:53mean boring in the best way possible
05:53then you have almost unlimited
05:55optionality because you're the operating
05:57system so you can build an accounting
06:00platform you could go compete with
06:01QuickBooks if you wanted to you have so
06:03much optionality what are your
06:05aspirations because you have this
06:07operating system role how do you think
06:08about that yeah it's the classic breadth
06:11and depth right in a way this is about
06:13Oracle did where you start out building
06:15databases and then you just keep asking
06:17your customers what are you using this
06:18database for and you build the business
06:19applications that people are using the
06:21database to build we saw this with out
06:23very notably where Amazon started with
06:25sqs and s3 and EC to the very basic
06:27primitive storage and confused and then
06:29you know moved very quickly up the stack
06:32to you know we're very happy users of
06:34AWS were very happy users of Amazon
06:35redshift redshift you know is a much
06:37more powerful data warehousing tool than
06:40the basics have just
06:41origen things like that and so there is
06:43a somewhat similar dynamic at play with
06:46stripe where you know we process
06:48payments for people that's not the
06:50world's most useful thing like it's
06:51necessary and you know part of wash
06:53helps is that it's necessary at day one
06:55of the business I mean definitionally
06:57what makes you know a business a
06:58business as opposed to a product is that
07:00it accepts money but that's generally
07:02not the limit of what people need to
07:04solve the one caveat I'd put on that is
07:07this is kind of mean in the payments
07:08industry which i think is wrong which is
07:11the core payments is really boring
07:13and you know you as quickly as possible
07:15want to sprint up to other things and
07:16just kind of ignore this payment singer
07:18and we just empirically know for a fact
07:20that's wrong I think part of why stripe
07:22has managed to get to the skill we have
07:24so far is that we haven't been a bastion
07:26about being really interested in
07:27payments there was an old payments world
07:29in the 80s where all that mattered was
07:31you know a merchant acquirer for Visa
07:33MasterCard and American Express and that
07:35was basically a commodity business and
07:37it was really uninteresting if you have
07:39go to Marcus via the Internet today
07:42payments is incredibly strategic for
07:45your product experience and in just your
07:47revenue growth overall so it wishes an
07:49e-commerce company they flipped over
07:51from a checkout they built on mobile to
07:54a stripe elements we spend all this time
07:55optimizing a mobile checkout so it's
07:57good the interest for us to do enough
07:59other people and they saw a 7% uplift
08:02and mobile conversions as a result of
08:04switching to this and so that's the kind
08:06of thing this wasn't possible in the 80s
08:08when payments or bars really boring but
08:10now just the core payment side of things
08:12can actually get pretty interesting and
08:14that's by the way before you go
08:15international you know people assume
08:16things are homogenizing cuz that's just
08:18you know generally being the story of
08:19the internet as more people come online
08:21in more countries you know things are
08:22heterogeneous a more you know now you
08:24have a Chinese markers available to you
08:27but everyone WeChat pay and Ally pay
08:28rather than credit card so it's
08:30important and you talked about kind of
08:31going up the stack from payments not to
08:33ignore that the payments market itself
08:35is actually nos standard or commodity
08:38anymore in the way there was in the 80s
08:40but then yes over time the way we view
08:42the opportunity is basically look what
08:44goes into collecting revenue there's
08:47generally many many full-time people
08:49around the business who are spending
08:51time on fairly unleveraged things like
08:55sales tax for the revenue that we're
08:57collecting or revenue recognition for
08:59accounting purposes or you know managing
09:01a billing system or you know all these
09:03kind of things and so it's interesting
09:05that in the way this you know CRM
09:08emerged as a category there was the time
09:09when there was no CRM and then as like
09:12wait a second you it's really valuable
09:13to have this single customer record that
09:16you invest in Austin keeping hygienic
09:17and then you can kind of do things about
09:18the way we see the world there is no
09:21revenue platform right now and if you
09:23look at all the products that we have
09:24been building better analytics on the
09:25revenue data that you haven't stripe a
09:27source of truth billing system things
09:28like that it's moving in that direction
09:30we're not at all there yet but that's
09:31the world part of it is kind of cost
09:33versus revenue so the PD led a company
09:35to bring down costs are not nearly as
09:37heroic as those that the company that
09:38helped bring up revenue mm-hm
09:40and I remember when I first met Charlie
09:42sharp on the referral he said it's like
09:43how do I get my clients to stop
09:45expletive hating me and my advice to him
09:47as well send them more customers or
09:50increase their conversion do something
09:51for them because if you're seen as the
09:52IRS when you're just taking attacks and
09:55they would have paid me anyway I mean
09:56this might be a fallacious argument but
09:58they would have paid me anyway and all
09:59you're doing is taking a toll I hate the
10:01toll taker I love the person actually
10:03brings in more revenue and I think
10:05that's part of the the argument as well
10:06which is if you're actually maximizing
10:08revenue you're no longer the Tollbooth
10:10and you're not the Tollbooth right like
10:11you're kind of the Tollbooth to the
10:13Tollbooth a Visa and MasterCard and all
10:15the other players I mean the
10:16heterogeneous players you know
10:18throughout the world that actually are
10:19the payments infrastructure you're at
10:21the top of that and there's also
10:22interesting organizational dynamics
10:23where anytime that you have you know one
10:27effect like a you know a positive effect
10:29that can be measured and a negative
10:30effect that can't be measured you're
10:32gonna see all these perverse
10:33organizational incentives because people
10:35track to the metric that can be measures
10:36and so the old schools here for the way
10:38of doing things is I'm just going to
10:40like work on grinding these costs down
10:41to zero and you know that is my metric
10:43and I think we're seeing across a whole
10:45range of industries including ours like
10:47wow this is actually something that we
10:48should pay for a rice you know yes
10:50Amazon is more expensive than racking
10:53our own servers but it would be new
10:54nnessee for us to do that because you
10:56mean the time to ship any product now
10:58there's gonna be 12 to 18 months well
10:59it's interesting like you have some
11:00companies that almost conduct a reverse
11:02RFP process so payment processing who's
11:04gonna be the lowest cost possible or you
11:06know I'm gonna hire a subcontract
11:08from my house who's gonna tile my roof
11:10for the lowest cost possible if you look
11:12at Google you know one of the advantages
11:13and one of the reasons why they are on
11:15their way to becoming a trillion dollar
11:16market cap company potentially is that
11:18all the advertisers compete to pay
11:20Google more money it's actually reversed
11:22so if you can somehow be on the side of
11:24that tribe you said it goes to the point
11:26of revenue maximization versus cost
11:28minimization and I thought I'd ask us I
11:31was checking this morning so these has a
11:33two hundred and fifty five billion
11:35I think MasterCard is about one hundred
11:36and fifty eight MX is seventy five
11:38PayPal's about that you know so you have
11:40if six hundred billion dollars roughly
11:42of market cap for the established guard
11:44of payment companies does that change
11:46how do you think about that because
11:48these and MasterCard don't actually have
11:49any customers of the merchants or of the
11:52consumers so these as customers or the
11:55issuing banks that actually issue those
11:57pieces of plastic that issue those cards
11:58and they have acquiring banks that go
12:01acquire merchants that go accept those
12:03cards and they're a middleman
12:04and they've been a very profitable
12:06middleman for a long time how do you
12:07think that could change in the next five
12:09or ten years right it's kind of two
12:11interesting questions one is what
12:13happens to entertained in credit card
12:14fees and one is what happens to a Visa
12:16MasterCard American Express in
12:17particular and I think Amex maybe leave
12:20them aside they're different because
12:21they were very diversified business with
12:22Europe travel services and things like
12:25these are MasterCard are more
12:26interesting you know there's a value
12:28chain in this stuff where credit card
12:30companies and more importantly the
12:31issuing banks go out and they get cards
12:34in people's hands and they're
12:35incentivized to do so but interchange
12:37flows and interchange flows have been
12:39very stable for a long time
12:41you know the classic 1.6% for debit
12:43cards that came down at Durbin and more
12:45for credit cards depending on the
12:47category but they've been pretty stable
12:48for a long time I think the question is
12:50what is the value that these issuing
12:52banks are providing relative to what
12:53they're charging and the value probably
12:55changes over time right because you know
12:56it's interesting you go to developing
12:58countries and there's lots and lots of
13:00out of home advertising going on for
13:01just like getting credit cards and you
13:03know credit cards are a good thing that
13:04you should have wares of course in the
13:06u.s. everyone has six credit cards right
13:07and it's a much more mature market let's
13:09be clear it's a really valuable service
13:11getting people from kind of the cash
13:12world to you know having a payment
13:15mechanism in their pocket that's you
13:18know is more or less fraud proof for the
13:20customer and they're confident using
13:22everywhere that's a really valuable
13:23service generally and a really valuable
13:25service for businesses in particular I'm
13:27not sure that the rates should be so
13:29steady over time and so what you're
13:30seeing well the interesting thing there
13:32is that it's kind of a double moat that
13:33protects them because not only do they
13:34have a network effect but they also have
13:36this bizarre incentive where if a 2.5%
13:39fee is being charged the merchant well
13:42the consumer is being bribed with maybe
13:431.5 percent of that exactly it's
13:45actually higher fees yield higher
13:47rewards for the consumers and actually
13:49you know in Australia where interchange
13:51went down dramatically it's like 50
13:53basis points rewards went away you
13:55didn't get your miles anymore and then
13:57it was a giant transfer of wealth from
13:58the banks to the merchants but none of
13:59that went back to the consumer well if
14:01you did in the form of prices at some
14:03point but yes sometimes I think that's
14:04obsolete to be what you're seeing is
14:06that in a way I think the prices are too
14:07stable and so it ends up being really
14:10stable at a level for a certain point
14:11and then it gets regulated and so
14:13Australia regulated interchange Europe
14:15famously regulated interchange even the
14:17US regulators debit interchange and just
14:20empirically if you kind of draw the
14:23lines in a curve it seems like that's
14:24the direction but it's very hard for
14:25market forces to contain problems of
14:27concentrated benefit and diffuse harm
14:29yeah and that would be one the diffuse
14:31harm is that you're overpaying and by a
14:33nickel yep or by a penny or a fraction
14:35of a penny and actually that's why the
14:37merchants have benefited more than the
14:39consumers in a place like Australia
14:40because do you care if you're paying 99
14:43cents for a coke or a ninety eight point
14:45seven cents for a coke and that that's
14:47that diffuse harm whereas the
14:49concentrated benefit is that 0.3 cents
14:51to Walmart of Australia times a million
14:54that actually is very very meaningful so
14:56it's hard to disrupt from a market
14:58forces perspective because consumers are
15:00actually worse off when pricing for
15:02credit cards go down mm-hm hi Ron ik Lee
15:05that's your original question way at the
15:07I mean Visa and MasterCard are very the
15:10amount that they actually earned
15:11themselves you know if you do the math
15:12is about 20 or 30 basis points of each
15:14transaction which I think kind of for
15:16them is a fairly stable amount for the
15:19value that they provide globally and I
15:22mean if there's a stat recently this
15:23visa uses 140th of the power of the
15:27Bitcoin network for processing obviously
15:29you know way larger and surprised me
15:31that look yeah yeah or maybe put another
15:33a big coin but a station tomorrow but
15:35uses 40 times the power of the visa
15:37Network despite a kind of a much lower
15:39throughput than righties in network and
15:41so these are kind of very useful
15:42services broadly if I was an equity
15:44investor there's no way I would be short
15:45these companies because you have the
15:46secular growth or just like did you know
15:49they're killing - yeah so a lot of
15:51people talk about data if you talk to
15:52any big company they want to use data
15:54actually that was 15 years ago now they
15:56want to use big data this big data is
15:57bigger than regular data yep
15:59super-sized payments have tons of data
16:02like you could predict economic
16:04macroeconomic micro economic progress if
16:06you have payments data I mean how do you
16:08think about data in payments like what
16:10do you do with it what should be done
16:12with it do you want to do more with it
16:14it's really easy to draw these
16:15theoretical examples and like oh we can
16:17use this signal or whatever and my
16:18favorite thing to do now is to like push
16:21people on well do you actually use that
16:22signal it's like well no no you know we
16:24did an experiment or whatever but like
16:26the really common behavior you observe
16:28is this really interesting speculative
16:31uses of data and a really narrow boring
16:33actual set of uses of data so say in
16:35payments we've been talking about using
16:37new fraud signals for literally decades
16:40and if you look at how the fraud systems
16:42that actually back your credit card work
16:45what happens is you know so you have a
16:47chaser Bank for America or whatever
16:49credit card and there's this head of
16:51neural networks that were trained and
16:53you're like oh cool neural networks that
16:55sounds really awesome and modern but not
16:57from kind of the current wave of neural
16:58networks but when they were originally
17:00first you know building product eyes
17:01back in the 1960s and that's basically
17:03when these models were trained would
17:05like handful of features like you know
17:07time of day and who this merchant is an
17:08amount and behavior and stuff like that
17:10and the system's more or less haven't
17:13been updated since and that's why you
17:14know the experience of using your credit
17:16card out and about is so frustrating
17:18it's like you know that I'm traveling I
17:20have your app installed on my phone I
17:22use this card at an ATM you know
17:23whatever it is and so when it gets kind
17:25of clobbered and frozen by the bank
17:27that's why that experience is so
17:28frustrating and so the way we think
17:30about data is not only what is the data
17:32available to people but how you make it
17:34useful we have much more interesting
17:36data that we actually make available to
17:37our engineers and so when we're looking
17:39at blocking fraud we're doing stuff like
17:41you know looking ash the you know we
17:44have JavaScript running on the end
17:45customers web browser and so we can tell
17:47a lot more about the veracity and is
17:49this customer who they claim to be than
17:50your traditional bank that spire who's
17:52trying to make a fraud decision then the
17:54flip side of it is making data useful
17:57actually on behalf of our customers and
18:00so you're putting all this data into
18:02stripe and the traditional thing that
18:03you get from a payments company is like
18:05a paper statement that gets put in a
18:06drawer at the end of the month and
18:08that's the extent to which it gets used
18:09and so now we give them a full sequel
18:12console across all of the data they have
18:14in stripe that's often faster than their
18:17own data warehouse or you know ETL
18:18pipelines because we can keep it up you
18:20know we are building a data warehouse
18:21for all our customers across the
18:22hundreds of thousands of people that you
18:24stripe and then you know we can build
18:25products on top of that that are more
18:27useful so it's both making the data that
18:29theoretically exists actually useful and
18:32used both internally for things like our
18:34fraud management purposes but then also
18:36for our customers given if they're
18:38putting what is maybe the most
18:39interesting data set of the business
18:41their revenue data in stripe what do you
18:42think the biggest opportunity adjacent
18:44to payments is like what is what are
18:46your compliments with an e that you
18:47would like people to develop or where do
18:49you think there are opportunities I
18:50think your colleague Chris Dixon who
18:52wrote a blog post about the fast startup
18:54the idea that people are taking a
18:56distribution layer or technology layer
18:58and being willing to own more of the
19:00stack behind us and it feels to me just
19:03looking around me that we are very early
19:05in that transition because you know it's
19:08always this age-old question coming
19:11combination the new world faster than a
19:13new startup can come along and get you
19:15know won't get distribution but to get
19:17good at what it is that that incumbent
19:18previously did and so can a challenger
19:21Bank come along and you know get
19:23regulated as a bank and you know bring
19:25back to market faster than existing
19:28banks can make their app good can you
19:29know a new startup come along and enter
19:31the healthcare space faster than
19:33existing healthcare companies can get
19:34their act together and that battle is
19:35still playing out across so many
19:37different industries as often pretty
19:39hard to tell a priori and so that's the
19:42space that we end up watching pretty
19:43all right great well John thank you very
19:45much thank you so much