a16z Podcast: Fintech for Startups and Incumbents

0:00:02 – Hi, this is Frank Chen.
0:00:04 Welcome to the A16Z podcast.
0:00:07 Today’s episode is titled Three Ways Startups Are Coming
0:00:11 for Established Fintech Companies and What to Do About It.
0:00:13 It originated as a YouTube video.
0:00:18 You can watch all of our videos at youtube.com/a16zvideos.
0:00:19 Hope you enjoy.
0:00:22 – Well, hi, welcome to the A16Z YouTube channel.
0:00:24 I’m Frank Chen, and today I am here
0:00:27 with one of our general partners, Alex Rampel.
0:00:29 I’m super excited that Alex is here.
0:00:32 So first fact, we both have sons named Cameron.
0:00:33 – We do.
0:00:34 – So affinity there.
0:00:36 And then two, one of the things
0:00:38 that I really appreciate about Alex,
0:00:39 and you can sort of see this
0:00:41 from his young chess playing days,
0:00:46 is he understands fintech and incentives
0:00:48 and pricing backwards and forwards.
0:00:52 And so fintech has this hidden infrastructure
0:00:54 on how do credit card transactions work?
0:00:56 How do bonds get sold?
0:00:58 How are insurance policies priced?
0:01:02 And there’s deep economic theory behind all of these
0:01:04 and Alex understands them all.
0:01:05 So you’re gonna have a fun time
0:01:08 as Alex takes you through his encyclopedia of knowledge
0:01:10 of how these things are put together.
0:01:12 And so excited to have you.
0:01:13 – Yeah, it’s great to be here.
0:01:15 – So what I wanted to talk to you about is
0:01:19 I’m gonna pretend to be in the seat of a,
0:01:21 let’s call it an incumbent fintech company, right?
0:01:26 So I’m a product manager and visa or a Geico.
0:01:30 And I am looking in my rear view mirror
0:01:33 and there are startups in the rear view mirror.
0:01:35 And I’m very nervous that the startup
0:01:39 in the rear view mirror, exactly as the mirror says,
0:01:41 objects in mirror maybe closer than they appear,
0:01:44 is like, wow, they are catching up to me faster
0:01:45 than I really want.
0:01:49 And so I wanna understand like, what are startups doing?
0:01:53 Like how would they mount an attack on me, the incumbent?
0:01:56 And we’re gonna talk about sort of wedges they can use.
0:01:58 And then that’s sort of the first half,
0:02:00 like how are they coming after me?
0:02:02 And then the second half, let’s talk about like,
0:02:03 and what should I do about it?
0:02:05 So that’s sort of the premise for our,
0:02:09 so why don’t we start with the attacks?
0:02:11 Like how would a startup come for me?
0:02:14 And one way they come for me is they come
0:02:16 after my best customers.
0:02:19 – Well, so this is the interesting thing
0:02:21 about financial services in general,
0:02:25 because there’s a sharp television hanging on the wall.
0:02:27 And Sharp knows that they make more money
0:02:30 every time they sell an incremental television.
0:02:33 So more customers equals more money, cause, effect.
0:02:35 And the interesting thing is that for many kinds
0:02:37 of financial services, that is not true.
0:02:39 Because what you’re really trying to do
0:02:41 is assemble a risk pool.
0:02:43 And the best example of this is insurance.
0:02:45 So what is car insurance?
0:02:48 Car insurance has good drivers, okay drivers,
0:02:50 and bad drivers.
0:02:53 And effectively, your good drivers and your okay drivers
0:02:56 are paying you every month to subsidize the bad drivers.
0:02:57 So the same thing goes for health insurance.
0:02:58 You have people that are always sick,
0:03:00 you have people that are always healthy.
0:03:02 And if you are an insurance company
0:03:05 that only provided insurance for very, very sick people,
0:03:07 or if you’re a car insurance company
0:03:10 that only ensures people that get into accidents every day,
0:03:13 there’s no economic model to sustain that.
0:03:16 You actually have to accumulate the good customers
0:03:18 and use them to pay for the bad customers.
0:03:20 And the interesting thing about this is that
0:03:22 from the perspective of the good customer,
0:03:24 it’s not fair.
0:03:26 And I’m not talking morally or philosophically,
0:03:30 but just from a capitalist or economic viewpoint,
0:03:32 it’s like, okay, I want life insurance
0:03:35 and I eat five donuts a day.
0:03:35 I just had a donut today.
0:03:36 I don’t eat five a day.
0:03:40 But I have one donut every Friday as you can testify.
0:03:43 And then I have a friend who goes to the gym five times a day,
0:03:45 never eats a donut.
0:03:47 That guy’s probably gonna live longer than me.
0:03:50 Hopefully not, but probabilistically,
0:03:53 he’s probably going to have a better time than I am
0:03:54 in terms of life expectancy.
0:03:57 So why is it that we both pay the same rate?
0:04:00 And that just seems unfair to him.
0:04:03 It seems great to me because he’s subsidizing me.
0:04:05 – Yep, Jim Guy, subsidizing donut guy.
0:04:06 – Exactly, exactly.
0:04:08 And that seems unfair.
0:04:11 And then the startups can sometimes exploit
0:04:13 that psychological unfairness,
0:04:15 like that feeling of unfairness.
0:04:16 So, and it kind of does two things
0:04:19 because from the big company perspective,
0:04:20 if you were to take away,
0:04:22 think of it as a normal distribution.
0:04:24 So most people are in the middle
0:04:26 and they’re just gonna live whatever
0:04:30 to the average of 79.6 years or whatever it is right now.
0:04:31 Some people are gonna live forever.
0:04:34 They’re the ones that have the olive oil go to the gym
0:04:36 and do whatever it is that they do
0:04:38 that makes them live a long time, great genes.
0:04:40 And then some people are gonna die early.
0:04:43 And from the perspective of the startup,
0:04:44 if you can get all of the people
0:04:47 that are going to live much, much longer,
0:04:49 you’re going to be more profitable.
0:04:50 The same thing for car insurance.
0:04:52 If you can get all the people on the good end
0:04:55 of that distribution curve, you’re going to make money.
0:04:57 And then the nice thing is that
0:04:59 if you’re starting a brand new company and saying,
0:05:01 “Hey, I give you a loan if you can’t get a loan.
0:05:02 “Who’s gonna sign up for that?
0:05:04 “People who might be bad.”
0:05:05 If I say, “I’m gonna give you insurance
0:05:07 “if you can’t get insurance.
0:05:08 “Who’s gonna sign up for that?
0:05:10 “The people that are eating all the donuts.”
0:05:11 And that might not be very good.
0:05:15 So it actually has this nice kind of symbiosis
0:05:17 between if you do it correctly,
0:05:19 you get positive selection bias
0:05:22 and that you establish a new criteria.
0:05:24 Part of that new criteria is based on data,
0:05:26 but part of it is based on psychology.
0:05:28 The psychology is I’m treated unfairly.
0:05:30 I want to be treated more fairly.
0:05:32 That yields a lower price for people
0:05:35 for a pretty demand elastic product.
0:05:37 So I say, “I can get life insurance at half the rate
0:05:38 “because I’m going to the gym.
0:05:39 “That sounds great.
0:05:40 “That sounds fair.”
0:05:41 But to answer your question,
0:05:44 what the incumbent might be left with
0:05:47 is not half of the number of customers.
0:05:48 Like that could be the case.
0:05:50 It could be half the number of customers,
0:05:51 but it could be half the customers
0:05:54 and all of them are entirely unprofitable.
0:05:55 – Right, they took all the profits.
0:05:57 They didn’t have to take all your customers.
0:05:58 They just had to take the good ones.
0:06:00 – Right, so actually, if you just take,
0:06:03 and the funny thing is that because it’s not like
0:06:05 I want to get, “Oh, Geico has X million customers.
0:06:07 “I want X plus one million customers.”
0:06:08 You actually might want one tenth
0:06:10 as many customers as Geico.
0:06:12 Because if you can just get the good ones,
0:06:15 I mean, what if you give people a 50% discount,
0:06:17 not a 15% discount, like Geico always advertises about,
0:06:20 but a 50% discount on their car insurance,
0:06:23 and these are the absolute best drivers in the country,
0:06:26 how many claims do you have to pay out on the best drivers?
0:06:30 You might have to pay out nothing, literally nothing.
0:06:31 And if you have to pay out nothing,
0:06:33 and there are all these mandatory loss ratios
0:06:34 for different insurance industries,
0:06:35 so I don’t want to get into that.
0:06:39 But imagine that unregulated, you can pay out nothing.
0:06:42 Consumers feel like they’re treated very fairly.
0:06:44 They’re rewarded for better behavior.
0:06:48 This begets positive selection and not adverse selection,
0:06:51 then you’re going to have the most profitable lending company
0:06:53 or insurance company in the world,
0:06:55 because it really is a unique industry
0:06:57 where more customers is actually worse
0:07:00 than less but more profitable customers
0:07:03 because each incremental customer is like a coin flip
0:07:04 of profit or loss.
0:07:07 Might generate profit, might generate loss.
0:07:09 And that’s not true for the vast majority of industries.
0:07:11 Like Ford never sells a car saying,
0:07:13 “Maybe we’ll lose money on this customer.”
0:07:14 – Right, right.
0:07:17 They just like, “I need everybody to buy a Ford F-150.”
0:07:20 And if you don’t buy an F-150, I need you to buy.
0:07:22 There’s other thing that said, the expedition or whatever.
0:07:24 – They might lose money on the marginal customer
0:07:26 until they hit their fixed costs.
0:07:28 But they’re never going to have a coin flip
0:07:29 of when they sell the car.
0:07:31 Hmm, maybe we shouldn’t have sold that car,
0:07:33 but that’s what every insurance company has
0:07:34 when they underwrite a policy.
0:07:37 And that’s what every bank has when they underwrite a loan.
0:07:39 – Yeah, so auto insurance companies need to find people
0:07:41 like me, I have this old Prius, right?
0:07:45 First, it’s hugely reliable car.
0:07:47 And then I drive like a grandma
0:07:49 because I’m optimizing for fuel efficiency.
0:07:52 So I rarely go above 65.
0:07:54 And so really safe, I’ve never filed a claim.
0:07:57 They need more customers like me.
0:07:58 And that’s what drives the profits.
0:07:59 – Yes.
0:08:00 – ‘Cause there’s no payouts.
0:08:01 – Well, not only does it drive the profits,
0:08:05 it actually subsidizes the losses.
0:08:08 Because there are a lot of people who are the inverse of you
0:08:09 and you’re paying for those people
0:08:11 and the transfer mechanism is through GEICO.
0:08:12 – Yeah.
0:08:16 I saw an ad in my Facebook feed recently
0:08:17 for HealthIQ and I think they’re doing something
0:08:18 like this too, right?
0:08:20 So the, I think the proposition was,
0:08:22 hey, can you run on my own in less than nine minutes?
0:08:24 Can you bench press your own weight or something like that?
0:08:27 There’s all these like, oh, healthy people.
0:08:29 And is that the mechanism they’re exploiting?
0:08:30 – It’s exactly that.
0:08:32 I would say the first company to probably do this
0:08:36 on a widespread basis in FinTech land was SoFi.
0:08:38 And SoFi said, hey, you’re really smart.
0:08:39 They actually coined this term.
0:08:40 They called it the Henry.
0:08:42 High earning, not rich yet.
0:08:44 Because if you look at how student loans work,
0:08:46 it’s like everybody gets the same price
0:08:48 on their student loan, right?
0:08:49 It doesn’t matter what your major is.
0:08:54 It doesn’t matter what your employment prospects think.
0:08:56 What your employment prospects are,
0:08:57 everybody gets the same rate.
0:09:00 You get this rate, you get this rate, you get this rate.
0:09:01 Because a lot of it is effectively underwritten
0:09:02 by the US government.
0:09:04 And that’s not, so think about it again
0:09:07 from the twin pillars of psychology.
0:09:09 Where, I mean, psychology of the borrower.
0:09:12 Like how come I’m paying the same rate
0:09:14 as that person who’s going to default?
0:09:15 That’s just not fair.
0:09:17 I’m never going to default.
0:09:20 In fact, I’m gonna pay back my student loans early.
0:09:22 So that helped.
0:09:24 And then again, positive selection
0:09:26 versus adverse selection because,
0:09:29 and actually refinance has this concept in general.
0:09:30 Because I would say, if you’re planning
0:09:33 on declaring bankruptcy, or if you’re saying,
0:09:36 I’m going to, I’m gonna join Occupy Wall Street
0:09:38 and never pay back my loans and I hate capitalism,
0:09:40 why would you go refinance?
0:09:41 It just doesn’t make sense.
0:09:42 – Right.
0:09:44 – Because you’re just gonna default.
0:09:44 – Right.
0:09:46 – So if you raise your hand, and actually it’s interesting,
0:09:48 even on the other side, there are a lot of companies
0:09:50 in what I would call the debt settlement space.
0:09:52 And this is something that most people don’t know about.
0:09:55 But if you listen to like some interesting talk radio,
0:09:58 you’ll hear all these ads for debt settlement.
0:09:59 And what is debt settlement?
0:10:01 It’s saying, hey, do you have too much debt?
0:10:05 If you call us, we will negotiate on your behalf
0:10:08 and pay off your debts, and then you just owe us.
0:10:10 And you kind of need this intermediary layer
0:10:13 because imagine that you owe $10,000 to Capital One
0:10:15 and you can’t pay it back.
0:10:16 You call it Capital One.
0:10:18 It says, press one for your balance.
0:10:20 Press two to get a new card mail to you.
0:10:22 Press three if you don’t want to pay us the full amount
0:10:23 and want to pay us less.
0:10:26 Everybody’s gonna push through, right?
0:10:27 – This is why they don’t offer that option.
0:10:29 – They don’t offer that option, nor will they ever.
0:10:34 However, on talk radio, and this is very big in the Midwest,
0:10:37 like you’ll hear, you know, Freedom Financial.
0:10:39 Go call Freedom Financial and we will settle
0:10:40 your debts for you.
0:10:42 So they call Capital One and say, look,
0:10:43 Alex can’t pay you back.
0:10:45 We’ll pay you $2,000 right now,
0:10:47 and then you’re gonna get rid of the loan.
0:10:48 And you’re like, well, we’re not happy
0:10:50 taking 20 cents on the dollar,
0:10:52 but it’s better than zero cents on the dollar, fine.
0:10:53 We’ll take it.
0:10:57 And then you owe Freedom Financial the 20 cents.
0:11:00 But why do they feel comfortable underwriting that?
0:11:03 Because you rose your hand, you said,
0:11:05 I want to get out of debt.
0:11:08 And that’s positive selection bias right there.
0:11:10 Because people who are just deadbeats,
0:11:12 because behind every credit score,
0:11:13 if you think about how that works,
0:11:15 it’s willingness and ability to repay.
0:11:18 And the psychological trait of the willingness
0:11:20 is in many cases as important
0:11:22 as the financial constraint of the ability.
0:11:25 Because if I owe a million dollars to somebody,
0:11:28 and I only make $100 a year,
0:11:29 it doesn’t matter how honest I am,
0:11:31 I can never pay that back.
0:11:33 It doesn’t matter how long I’m gonna live 10,000 years
0:11:34 and I guess I could pay it back.
0:11:36 But otherwise I can’t pay that back.
0:11:38 But the willingness to repay is interesting.
0:11:40 And that’s very important.
0:11:42 And that’s again, this kind of psychological trait
0:11:45 that’s captured in this idea of positive selection.
0:11:46 So what does SoFi do?
0:11:48 They kind of again hit this twin pillar,
0:11:52 which is I want to only get the good customers,
0:11:55 I’m going to reprice them and steal them
0:11:58 from the giant pool that again, normal distribution,
0:12:00 these are the losers, these are the whatever’s,
0:12:03 and these are the people that you have no risk on whatsoever.
0:12:06 Let’s steal all of these people over here.
0:12:07 And it makes them feel good.
0:12:09 It’s a better marketing message.
0:12:11 At least it’s differentiated.
0:12:13 How do you compete with everybody?
0:12:14 It’s like, hey, we’re just like Chase,
0:12:17 but smaller and a startup and not profitable
0:12:20 and you probably shouldn’t trust us, bad marketing message.
0:12:22 Good marketing message is you’re getting ripped off.
0:12:26 We’re going to price you fairly, come to us.
0:12:29 So if I did this for lending.
0:12:30 – And what a health IQ do?
0:12:32 – So a health IQ did this for health,
0:12:33 really for life insurance.
0:12:35 So they started off with a health quiz
0:12:38 because I mean, it seems almost self-evident
0:12:40 that healthy people are health,
0:12:42 I mean, it’s a tautology,
0:12:44 like healthy people are healthier than not healthy people,
0:12:47 but can you actually prove this
0:12:48 from a life expectancy perspective?
0:12:50 So they started off with just recording data
0:12:53 and then building a mortality table.
0:12:56 And it turned out that what I would assume
0:12:59 is a prima facie case turned out to actually be correct,
0:13:01 which is these healthier people do live longer
0:13:03 than not healthy people.
0:13:06 And then they turn that into both a positive selection
0:13:07 advertising campaign,
0:13:09 which differentiated them from a brand perspective,
0:13:12 but also left them more profitable.
0:13:13 So what they do is they say, yeah,
0:13:16 can you run a nine or an eight minute mile?
0:13:19 Can you do these things to prove
0:13:21 that you’re better than everybody else?
0:13:22 And why is that important?
0:13:24 Well, from their own balance sheet
0:13:25 or profitability perspective,
0:13:27 they want to get these good customers
0:13:30 versus a brand new life insurance company
0:13:32 that said, hey, life insurance takes too long to get,
0:13:34 it’s a big pain and it’s expensive,
0:13:36 we’ll underwrite you on the spot in one minute,
0:13:39 no blood test, that’s gonna be adverse selection.
0:13:42 That’s like, ooh, I think I’m gonna die soon.
0:13:44 I want to get, everybody rejected me for life insurance,
0:13:45 I’m going to that company.
0:13:48 As opposed to here, they’re only getting the customers
0:13:50 that kind of hit,
0:13:52 that think they’re gonna hit the underwriting standard,
0:13:53 which is great.
0:13:55 They think it’s fair.
0:13:57 So it’s a differentiator from a brand perspective.
0:14:00 And then it turns out that, again,
0:14:02 each marginal customer in insurance
0:14:03 is kind of a coin flip.
0:14:05 They’re getting a weighted coin
0:14:06 because they’re only getting people
0:14:09 on the far right side of this normal distribution.
0:14:14 – So wedge number one is exploit psychology, right?
0:14:16 Positive selection rather than negative selection
0:14:18 and what you’ll end up with
0:14:20 because of this sort of unique dynamic
0:14:21 of the FinTech industry
0:14:24 is you’ll end up with the most profitable customers.
0:14:25 What’s wedge number two?
0:14:27 We’re gonna talk about sort of new data sources
0:14:29 and what startups can do
0:14:31 to sort of price their products smarter than incumbents.
0:14:36 – Right, so imagine that you have a group of 100 people
0:14:38 and of the 100 people,
0:14:40 half of them are not going to pay you back.
0:14:43 So think of this as the old combinatorics problem
0:14:44 of bins and balls.
0:14:46 So you’ve got this giant ball pit,
0:14:48 you scoop up 100 balls in your bin
0:14:50 and half of them are going to be bad,
0:14:53 half of them are going to be good.
0:14:56 So what’s a fair rate of interest if you’re a lender,
0:14:59 that you have to charge this whole bin
0:15:01 if half of them are going to default
0:15:03 and you assume that you can’t lose money?
0:15:05 The answer is going to be 100%.
0:15:08 – Oh right, because half of them you have to make up
0:15:09 for all the deadbeats.
0:15:10 – So half of them, you lose all of your money,
0:15:12 half of them you double your monies,
0:15:13 you’re back to square one.
0:15:14 – Now you’re even.
0:15:15 – Now you’re even.
0:15:18 So the problem is that that’s not good
0:15:19 because well in the United States
0:15:21 you can’t charge 100% interest.
0:15:24 It’s called usury, there are other parts of the world,
0:15:28 again, illegal, step one, Europe, so that’s a problem.
0:15:32 But what if you can use different data sources to,
0:15:36 again, it’s not positive versus adverse selection
0:15:38 as in some of the insurance companies,
0:15:41 but it’s saying can I collect more forms of data
0:15:43 so that instead of saying the only way
0:15:45 that I can make my operation work
0:15:46 is to charge an interest rate
0:15:49 which actually turns out to be illegal,
0:15:51 can I come up with more data sources
0:15:53 that effectively, even though discrimination
0:15:55 sounds like a terrible word,
0:15:57 and it’s normally used in that construct,
0:15:59 if you discriminate against criminals that’s fine.
0:16:01 I mean some of the people that try to take advantage
0:16:04 of lenders are actual organized crime.
0:16:05 You don’t want them in your bin,
0:16:06 you want to throw them out.
0:16:09 How do you take more data sources
0:16:11 and actually start measuring this?
0:16:12 And the interesting thing here,
0:16:13 and it’s somewhat unfortunate,
0:16:17 but you have a giant market failure happening
0:16:18 in many different regions of the world
0:16:19 because in the United States,
0:16:21 like the top interest rate that you can charge,
0:16:22 it’s regulated on a state by state basis,
0:16:25 but Utah has a 36% usury cap,
0:16:28 so a lot of people export that cap.
0:16:30 That’s a lot less than 100% that I was mentioning.
0:16:32 And there are lots of ways of kind of gaming that system
0:16:34 and you charge late fees and you charge this fee,
0:16:36 so it actually might end up looking more like 100
0:16:39 or 200%, but so you can charge more than 36%.
0:16:42 And then you actually can’t use certain types of data
0:16:44 if they are prone to having an adverse impact.
0:16:46 So if you think about how machine learning works,
0:16:47 I always kind of describe it
0:16:50 somewhat oversimplistically as linear algebra,
0:16:53 where I have, here’s every user that I’ve ever seen,
0:16:55 here’s every attribute that I’ve ever measured,
0:16:57 and what I’m looking for is like strange correlations
0:16:59 that I can’t even explain.
0:17:01 So I’m gonna ask you,
0:17:02 I’m not even gonna ask you a lot of these things,
0:17:04 it’s like, how long did you fill out this field for
0:17:06 on my loan application?
0:17:08 Did you enter all caps or not all caps?
0:17:09 Just all of these different things.
0:17:11 – Did you take the slider on how much do you want
0:17:12 and jam it all the way to the right?
0:17:13 – Right, all of these things.
0:17:16 – I can ask you, do you have a pet or not?
0:17:16 That might be interesting.
0:17:18 I don’t know if that’s a leading indicator
0:17:20 of default or not, but I wanna collect
0:17:21 all these different variables,
0:17:22 and then at the end of the day,
0:17:24 I’m going to see default or not default.
0:17:26 That’s the output, and then I’m going to see
0:17:27 what’s correlated with that.
0:17:29 And it’s a little bit of this, it’s a little bit of that,
0:17:31 I can’t explain it, but the computer can’t.
0:17:33 Now the problem is that in the United States,
0:17:35 you actually can’t do this,
0:17:37 because it might have an adverse impact.
0:17:39 And what does an adverse impact mean?
0:17:42 There actually was outright and terrible discrimination
0:17:44 in lending in the United States,
0:17:46 where there’s unfortunately terrible discrimination
0:17:47 in many things in the United States,
0:17:50 but lending was one of several or one of many.
0:17:54 So imagine that I said, are you married or not?
0:17:57 Oh, you’re not married, I’m not gonna make you alone.
0:17:59 Well, that’s illegal now.
0:18:00 Are you this race?
0:18:01 Oh, I’m not going to make you alone.
0:18:02 Well, that’s illegal now.
0:18:03 So what did people do to get around,
0:18:06 the people that were actual racists,
0:18:08 or actual, like maybe they weren’t racist
0:18:09 or discriminatory at heart,
0:18:11 but they were picking up on cues.
0:18:13 They’d say, oh, what part of town do you live in?
0:18:15 Well, you live on that part of town.
0:18:19 Well, that’s like 100% correlated with this race,
0:18:20 or this gender, or this, that.
0:18:22 I’m not going to make you alone.
0:18:23 So the law was strengthened,
0:18:26 so there’s a law called Fair Lending in the United States.
0:18:27 And then one of the components of it
0:18:30 is this idea of called adverse impact.
0:18:31 And it’s different than adverse selection.
0:18:34 It’s saying, I don’t care what you said you did
0:18:38 for why you rejected Frank for a loan.
0:18:41 If it turns out that everybody in your reject pile
0:18:45 has a disproportionate gender ratio, race ratio,
0:18:46 something like that,
0:18:49 I’m going to assume that you’re underwriting standards
0:18:50 or having an adverse impact.
0:18:53 So you as a bank couldn’t say,
0:18:55 hey, look, I asked him if he had cats.
0:18:58 And I’m using that to make the loan decision.
0:18:59 If it turned out that having cats
0:19:04 was correlated with being in particular race,
0:19:09 they couldn’t use the cat’s answer to deny you a loan.
0:19:10 – Correct, because that was,
0:19:12 and in all fairness to the law,
0:19:15 this is what people use with your geography.
0:19:16 What zip code do you live in?
0:19:18 Oh, you live in that zip code?
0:19:21 100%, you were a member of this particular race,
0:19:23 and the intent all along was to discriminate
0:19:25 against people of that particular race.
0:19:28 But now instead of using loan officers that use,
0:19:30 God knows what to decide.
0:19:31 Do I want to make you the loan or not?
0:19:33 You’re using a computer, you can look at the code.
0:19:35 So I think there is a lot of,
0:19:38 there are some anachronistic laws
0:19:39 that have to catch up here,
0:19:42 but let’s take an area outside of the US
0:19:44 to answer your question,
0:19:47 where perhaps you don’t have interest rate caps,
0:19:49 because the thing that a lot of people say,
0:19:51 oh, 200% interest is terrible.
0:19:55 500%, that sounds awful, you should go to jail for that.
0:19:56 But what does APR mean?
0:19:59 APR stands for annual percentage rate.
0:20:01 And what if I’m giving you a four day loan?
0:20:04 So I say, okay, I’m gonna loan you $9 right now.
0:20:07 You don’t look very trustworthy.
0:20:09 I want you to pay me back $10 on Monday.
0:20:11 – Yeah, that doesn’t sound so bad.
0:20:13 – Yeah, it’s like, you’re gonna pay me a dollar.
0:20:16 But what is that on an APR basis?
0:20:19 That’s like 9,000% made that up.
0:20:20 But it’s probably about that, right?
0:20:23 Because it’s 10% every four days, or every three days,
0:20:25 10% every three days, and that cumulates.
0:20:28 Like that’s a lot of money or a lot of interest
0:20:31 on an APR basis, but it’s the wrong metric
0:20:33 because effectively it’s like trying to figure out
0:20:37 what your marathon time is based on your 100 meter dash.
0:20:40 Like the winning marathon time would be an hour.
0:20:41 And that’s not true.
0:20:42 We know that nobody can run on that,
0:20:44 two hour marathon right now.
0:20:45 Yeah, so maybe Angela can.
0:20:46 – Maybe Angela.
0:20:50 – So there’s a company that we invested in called Branch.
0:20:52 And what they’re doing is they just collect
0:20:54 every form of data possible.
0:20:57 And they look for these strange correlations.
0:21:00 And the interest rates on an APR basis might be high,
0:21:02 but they’re really charging like a dollar.
0:21:04 – And these are small loans, right?
0:21:05 – Very, very small loans.
0:21:07 So I loan you, and actually the other interesting,
0:21:09 like one of the nice data points
0:21:11 that they’re accumulating over time
0:21:15 that is a really interesting idea, I think.
0:21:16 It’s not new.
0:21:18 In fact, it’s almost back to the future old
0:21:19 where they loan you a dollar.
0:21:20 If you pay it back, they loan you $2.
0:21:22 If you pay it back, they loan you $4.
0:21:24 If you pay it back, they loan you $10.
0:21:26 And they ladder up your credit
0:21:29 and they keep that information proprietary to them.
0:21:33 Because induction turns out to be a pretty good formula
0:21:36 for figuring out not so much the ability to repay,
0:21:37 but the willingness to repay.
0:21:40 You’ve established a pattern of willingness to repay,
0:21:43 but they also look at where were you today.
0:21:46 And again, you provide all this information
0:21:47 in order for them to crunch this,
0:21:49 in order for them to give you a loan
0:21:50 at ideally a lower rate.
0:21:51 Because the more information,
0:21:53 because it’s kind of twin pillars, right?
0:21:55 The less information we have,
0:21:57 the higher the rate that we have to charge.
0:21:58 Not because we’re evil,
0:22:00 but because otherwise you’re gonna have a market failure.
0:22:01 Like you have in lots of-
0:22:02 – You have the bin ball problem, right?
0:22:03 – Exactly. – Because you have no idea
0:22:04 how many deadbeats.
0:22:05 – Exactly.
0:22:07 And if I don’t have any idea,
0:22:08 I either have to charge a high rate
0:22:10 or not charge anything at all.
0:22:11 And not charge anything at all
0:22:13 doesn’t mean like everybody gets a 0% loan.
0:22:14 It means I don’t make any loans.
0:22:16 And like both of those are bad outcomes.
0:22:19 The better outcome is you accumulate more data
0:22:21 and you figure out here are the good people,
0:22:23 let me not accept the bad people.
0:22:25 Because again, the way that the good people
0:22:27 end up paying more money
0:22:29 is if the company starts accepting more bad people
0:22:32 because it goes back to what I said at the beginning,
0:22:35 which is more customers in this unique industry
0:22:38 often is bad if you don’t understand
0:22:40 how to select them correctly.
0:22:42 And for many of these new fangled lending
0:22:43 and insurance companies,
0:22:46 the default customer is going to be adversely selected
0:22:47 because if you’re a new lender
0:22:49 and you have no underwriting standards,
0:22:53 you’re basically advertising free money never pays back.
0:22:54 And those are the people that will be attracted to you,
0:22:57 both the criminals and the non criminals in droves.
0:23:01 – Yeah, so this is sort of startup attack wedge number two,
0:23:03 which is I’m going to generate a new data source
0:23:06 that allows me to price my product in a way
0:23:09 or reach a customer that a traditional company
0:23:11 would never even try or they don’t have the data source
0:23:13 or they have the bid and ball problem.
0:23:16 So what are the types of data that branch went to go get
0:23:17 to try to figure out,
0:23:19 should I give you a loan of a dollar or two?
0:23:21 – Well, the other type of data,
0:23:23 so branch was somewhat unique
0:23:27 in that they said we’re going to get data from your phone.
0:23:29 And it seems odd.
0:23:33 He’s like most lenders in the developed world
0:23:35 or not developed versus undeveloped,
0:23:37 it’s really like with developed credit infrastructure.
0:23:39 – Yeah, if there’s a credit bureau.
0:23:41 – They look up your credit report, if it’s good,
0:23:43 they make you a loan, if it’s bad, they don’t make you a loan.
0:23:45 It’s actually not that hard.
0:23:47 And there are all sorts of nuances that you can layer on top,
0:23:49 but this is how it’s been working for a long time
0:23:51 in the United States as an example.
0:23:55 Whereas there, it was like, okay, where did you work today?
0:23:57 Did it look like you worked today?
0:23:59 So it was stuff like that
0:24:01 and even like how many apps do you have on your phone?
0:24:04 Like weird stuff that you would never assume
0:24:07 actually has any kind of indication
0:24:09 of willingness or ability to repay,
0:24:10 but in many cases it does.
0:24:12 Like are you gambling?
0:24:14 Well, if you have a gambling app on your phone,
0:24:17 you’re probably gambling, maybe that’s good.
0:24:18 Yeah, maybe it’s bad.
0:24:20 It’s actually not making human judgments.
0:24:21 And it’s also not looking at any one
0:24:24 of these unique variables as a unique variable.
0:24:26 It’s looking at them in concert
0:24:29 and then correlating them with these outcomes
0:24:31 or really observing the outcomes
0:24:33 and then linking them back to all of these different inputs.
0:24:35 – Yeah, I remember talking to the team
0:24:38 when I was researching my last machine learning presentation
0:24:41 and the fascinating things that I found were
0:24:45 if you’ve got more texts than you sent,
0:24:46 you were more credit worthy.
0:24:47 If you had the gambling app,
0:24:49 you were more credit worthy rather than less,
0:24:52 which is not kind of what you would expect.
0:24:54 If you burned through your battery,
0:24:56 you were more likely to default, right?
0:24:59 So like all of these things where human alone officers
0:25:02 would never really guess,
0:25:03 and they probably would guess the wrong way.
0:25:05 – Right, because many of them are counterintuitive.
0:25:08 And then many of them are not, they’re not unilateral.
0:25:10 Like so it’s not just, I mean, I don’t know,
0:25:12 but it’s not just the battery thing,
0:25:14 it’s the battery thing with this, with that, with that.
0:25:15 – It’s the combinations.
0:25:17 – And it’s like humans can only really observe
0:25:20 three dimensions plus time, so I guess four,
0:25:23 and these are 9,000 dimensional problems.
0:25:25 So it’s just, it’s much, much more challenging
0:25:27 for humans to really grok.
0:25:28 – Yeah, got it.
0:25:31 So that’s the sort of the second category of attack,
0:25:33 which is you generate a new data source
0:25:36 and then that allows you to price or find customers
0:25:38 in sort of a more cost effective way.
0:25:40 Let’s talk about the third,
0:25:45 which is around sort of fundamentally changing behavior.
0:25:47 So why don’t you talk about,
0:25:49 maybe Ernie is a good example of this?
0:25:52 – Yeah, so if you assume that humans are static,
0:25:55 so they’re born, both of our cameras were born,
0:25:58 and their DNA is set upon birth.
0:25:59 Maybe it changes a little bit with some mutations
0:26:01 from some gamma rays here and there,
0:26:03 but it’s set upon birth,
0:26:05 and then human behavior never changes.
0:26:06 And that’s one way of looking at things.
0:26:08 And then you think about adverse selection
0:26:09 versus positive selection.
0:26:11 Good drivers are always good drivers.
0:26:12 Bad drivers are always bad drivers.
0:26:14 Let’s just get the good drivers.
0:26:16 So the other category,
0:26:17 and it’s not to say that these other two groups
0:26:18 don’t do this,
0:26:20 but if I look at a company like Ernie,
0:26:24 most payday lenders are reviled
0:26:25 because they charge high fees,
0:26:29 they don’t educate their borrower very well.
0:26:31 Now it actually provides a valuable service
0:26:33 because if I’m getting paid next Friday,
0:26:36 but my rent is due today and I don’t have money,
0:26:37 do I want to get evicted?
0:26:41 No, I want to get paid right now,
0:26:43 and the only person that does this is the payday lender,
0:26:46 but the payday lender is competing with other payday lenders
0:26:49 for advertising in the local newspaper or something,
0:26:51 and if they’re able to rip me off more,
0:26:52 not because they’re evil,
0:26:55 but because they have to afford the advertising spot,
0:26:56 they’re now ascended to do so.
0:26:59 So it’s just, it’s a vicious cycle.
0:27:00 So let’s talk about Ernan.
0:27:02 So what Ernan does is they say,
0:27:05 okay, we know that you’ve worked this long.
0:27:09 So again, new data source because the phone’s in your pocket
0:27:12 and you work at Starbucks and you’re getting paid hourly
0:27:13 and we’ve seen the phone in your pocket
0:27:16 or in your locker in the Starbucks office
0:27:19 and you’re by the barista counter for eight hours.
0:27:21 So you worked, we saw your last paycheck,
0:27:22 hit your bank account,
0:27:24 we know that that’s where you work,
0:27:25 we’re not taking your word for it,
0:27:27 we have real-time streaming information about this,
0:27:33 and now we will give you your money whenever you want.
0:27:35 Not money that you haven’t earned yet,
0:27:36 but money that you have earned,
0:27:38 but you actually haven’t gotten paid for yet.
0:27:41 And then you can tip us, there’s no cost.
0:27:44 If you want, you can get us– – No interest, no fee, no–
0:27:45 – If you want to pay us nothing, that’s fine.
0:27:47 I mean, we would appreciate if you pay us something
0:27:49 because obviously we’re providing valuable service for you.
0:27:52 And then you can even give tips for your friends,
0:27:54 there’s this community that’s really emerged
0:27:55 of people on Ernan.
0:27:57 And actually, if you look back at different business model,
0:28:00 but this idea of microfinance in general,
0:28:04 so if you think about Mohammed Yunus and what he did,
0:28:07 this idea of can you encourage people
0:28:11 to pay back loans using social pressure.
0:28:13 So again, not adverse selection versus positive selection,
0:28:16 but actually trying to force everybody
0:28:18 down positive behavior.
0:28:20 – Let’s get the community to encourage repayment.
0:28:24 – Right, because then saying,
0:28:27 or let’s get the community to encourage people
0:28:29 actually driving safely.
0:28:33 Because there’s underwriting at the time of admission,
0:28:36 there’s underwriting based on ongoing behavior.
0:28:38 So like many of the car insurance companies
0:28:41 that are brand new are saying we will re-underwrite you.
0:28:44 Like, yeah, if you drive like Frank when you signed up,
0:28:48 great, but now you switched into like race car driver mode
0:28:49 and you were trying to hack us,
0:28:52 but we’re actually monitoring your speedometer at all times.
0:28:53 So guess what?
0:28:54 You got a higher rate now.
0:28:57 So that might encourage you to drive safely.
0:29:00 If I’m Frank and I drive safely in my Prius,
0:29:03 but then I decide and then I got a really good rate
0:29:05 on my car insurance as a result.
0:29:07 And now I’m like, aha, I game the system.
0:29:10 Now I’m going to drive like a maniac.
0:29:12 Well, the nice thing is that you can make
0:29:14 underwriting dynamic and you can say, all right,
0:29:17 we’re actually going to re-underwrite you every day.
0:29:18 So we had the positive selection
0:29:20 to try to attract the Franks.
0:29:24 We have the continuous evaluation to try to encourage
0:29:27 the right behavior post Frank sign up
0:29:30 and also to stop the gamification of it’s like,
0:29:33 I’m gonna pretend to be safe and then be like a maniac.
0:29:35 But then how do you actually get?
0:29:38 What if Frank was a bad driver initially,
0:29:42 doesn’t fall into my positive selection loop,
0:29:47 but I still want to try to make Frank a better driver.
0:29:48 – If I could turn him into a good driver,
0:29:49 he’d be profitable.
0:29:52 – Right, so because that’s the flaw
0:29:55 with kind of wedge one and wedge two
0:29:57 of like creaming the crop, really wedge one,
0:29:58 which is we’re going to cream the crop.
0:29:59 We’re going to do what SoFi did.
0:30:00 We’re going to do with health IQ.
0:30:03 I mean, it’s a great strategy,
0:30:07 but the rest, again, if you assume that it’s all nature
0:30:08 and there’s no nurture,
0:30:10 then perhaps there’s nothing you can do.
0:30:13 But if you can actually try to nurture better behavior,
0:30:17 you actually see better, you do see better behavior
0:30:19 and then the profitability goes up.
0:30:21 So, and the interesting thing there
0:30:24 is that you’re still finding mispriced customers,
0:30:26 but you’re actually helping turn them
0:30:28 into correctly priced customers.
0:30:31 So somebody like a bank would turn away that customer
0:30:33 and say, we don’t want them
0:30:37 because they have a 500 FICO, which is really bad.
0:30:39 And then you have to figure out,
0:30:42 and as with all of the new startups
0:30:43 that are saying we only want the best customers,
0:30:46 we want to leave the banks with the bad customers,
0:30:48 but it’s kind of the twin pillars of
0:30:52 can you identify something that’s below that credit score
0:30:54 or below that driving score or something,
0:30:56 and then can you encourage positive change?
0:30:58 And if you can, then you can start actually
0:31:02 creaming the crop of the bottom half of the customers,
0:31:02 right, not even the bottom half.
0:31:04 It’s the customers that are just neglected
0:31:06 because nobody wants to underwrite them.
0:31:09 And then you do that, you take them on
0:31:12 because you have a secret to change their behavior.
0:31:14 – Right, you’re seeing a lot of companies
0:31:18 that sort of are using behavioral economics research
0:31:22 to figure out how do I nudge people into better behavior.
0:31:23 And so this would be an example
0:31:26 of how you’re trying to change behavior
0:31:28 to get the profitable customer.
0:31:31 – Right, so there is one company
0:31:33 in the lending space a while ago called Vouch.
0:31:35 I think ultimately it didn’t work,
0:31:38 but when you apply for a loan,
0:31:41 it actually kind of taps your social network
0:31:44 and it requires that they do a reference for you.
0:31:45 Either a reference in terms of like,
0:31:49 yes, Frank is a good customer, you can trust him.
0:31:52 And even kind of a co-commit.
0:31:56 So I’m getting a loan for $1,000 and you say,
0:31:59 yeah, Alex is okay, or I’m saying Frank is okay.
0:32:01 And if he doesn’t pay you back,
0:32:05 I will put $100 in, because that’s how confident I am.
0:32:07 And it’s not all 1,000, but it’s 100.
0:32:11 And then you’re my friend, I go bowling with you.
0:32:13 We go take our cameras out together.
0:32:16 And if you don’t pay back this $1,000
0:32:19 to this kind of faceless, large, evil corporate entity,
0:32:21 not really, but if you don’t pay that back,
0:32:23 I’m on the hook for a hundred bucks.
0:32:25 I’m not going bowling with you anymore.
0:32:27 So there are other things that are really interesting
0:32:30 to try to encourage the correct form of behavior.
0:32:33 And actually, part of it is just making it personal.
0:32:35 Like this was the whole Eunice theory,
0:32:37 which is if you are kind of held accountable
0:32:41 by your peers, that is so much more powerful
0:32:44 than getting a collections call from Citibank.
0:32:46 Like you’re like, ooh, that’s the collections never,
0:32:48 iPhone block, done.
0:32:50 But how am I going to block my friends out?
0:32:53 If Alex calls me and says you really got to pay
0:32:55 that loan back, otherwise I’m out a hundred bucks, right?
0:32:56 That’s much more powerful.
0:32:58 I mean, this has worked great for a lot of health
0:32:59 in a different domain, right?
0:33:03 Which is if you are trying to get a pre-diabetic patient,
0:33:06 not to get diabetes, the most effective thing to do
0:33:08 is lose something like six or 7% of your body mass.
0:33:11 And the way they do it is they get you into a group.
0:33:14 They mail everybody a scale.
0:33:16 Everybody sees your weight in the morning, right?
0:33:18 Like that’s a powerful motivator.
0:33:20 – Yeah, I mean, this stuff, psychology is very powerful.
0:33:24 So there are a lot of tricks that you can use here.
0:33:27 And if you understand the impact of them,
0:33:29 you actually have to reassess your entire branding
0:33:31 and customer acquisition strategy, right?
0:33:32 – Right, right.
0:33:35 All right, so remember I opened up pretending
0:33:38 to be the product manager at Visa.
0:33:41 And now we’ve gone through all of these three categories
0:33:42 of how the startups are coming for me.
0:33:45 And like, I’m starting to sweat here, right?
0:33:47 They can come get my best customers.
0:33:48 They can generate new data sources
0:33:50 that like I would have a hard time doing.
0:33:53 They can actually even go after sort of worse customers,
0:33:54 change their behavior,
0:33:56 turn them into profitable customers.
0:33:57 I’m scared now.
0:33:59 Like what in the world should I do?
0:34:02 Like you’re in my seat, you’re the head of innovation
0:34:04 or head of strategy or head of digital
0:34:06 at one of these big FinTech companies.
0:34:10 What should I do with respect to startups?
0:34:11 – Well, I think it’s actually very hard
0:34:13 for a company that’s trying to be all things
0:34:15 to all customers.
0:34:16 Because if you look at what SoFi is,
0:34:18 look at SoFi’s brand.
0:34:20 Brand is, you know, we are the high,
0:34:21 like if you’re great, you’re good enough for us.
0:34:22 – If you’re Henry, right?
0:34:24 – If you’re a Henry, you’re good enough for us.
0:34:25 Health like you.
0:34:27 If you’re healthy, you’re good enough for us.
0:34:30 So on that sector of the curve,
0:34:33 how does GEICO say, hey, if you’re a good driver,
0:34:35 go to this special part of GEICO.
0:34:37 If you’re a regular driver, you still save 15%.
0:34:40 If you’re a bad driver and you had a DUI,
0:34:41 well, we can cover you over here.
0:34:46 It’s just, it’s lost in this kind of giant GEICO,
0:34:48 you know, GEICO marketing message.
0:34:52 So in many cases, it actually helps to have sub-brands
0:34:55 and divide this up, which is somewhat anathema
0:34:56 to a lot of companies that want to say,
0:34:59 how do we get as much efficiency and synergy as possible?
0:35:01 We’re gonna have one overarching brand.
0:35:03 And, you know, one of my favorite examples
0:35:04 of this kind of different industry,
0:35:07 but the highest end of the highest end of jewelry
0:35:08 is Tiffany and Co.
0:35:10 Or one of the highest end of the highest end.
0:35:13 And for a long time, it was owned by Avon.
0:35:14 – No, really?
0:35:15 – You know, the Avon lady, Avon.
0:35:19 So, and if Avon bought Tiffany, which they did,
0:35:20 and they said, okay,
0:35:23 we’re gonna rebrand Tiffany and Co as Avon,
0:35:25 like that doesn’t work.
0:35:27 Like you’re not gonna get, you know,
0:35:30 80% gross margins on whatever they sell at Tiffany and Co.
0:35:33 – Prefisted Avon’s just doesn’t have quite the right ring.
0:35:34 – It doesn’t work.
0:35:36 And then for Avon to say, okay, you know,
0:35:39 the door-to-door salesperson or sales lady
0:35:41 with the pink Cadillac that’s going around,
0:35:45 like we’re now going to have her push $2,000 bracelets
0:35:48 as opposed to the normal $10 fare,
0:35:49 like that’s not gonna work either.
0:35:51 But it actually can make sense
0:35:54 if you want to just appeal to more customers,
0:35:55 you have different brands
0:35:57 and you don’t wanna all suck them together.
0:35:59 So you can imagine instead of having, you know,
0:36:01 Geico could be your generic brand,
0:36:03 but then you could have,
0:36:04 I think I mentioned this to you once before,
0:36:05 a friend of mine is Mormon,
0:36:07 doesn’t drink alcohol,
0:36:10 and says we should have Mormon insurance for cars
0:36:11 because it’s just totally unfair.
0:36:13 Again, going back to the psychology point,
0:36:16 like why is it that I’m paying for the drunk idiot
0:36:17 that goes through the stop sign,
0:36:20 I don’t drink, I can prove that, I will never drink,
0:36:21 I have a million friends just like me
0:36:22 that will never drink,
0:36:23 we should all get car insurance,
0:36:25 we should all get a 40% lower rate.
0:36:28 Do they think of Geico when they go there?
0:36:29 Maybe they could,
0:36:32 but it could be like Mormon car insurance,
0:36:33 or something, I’m not good at branding,
0:36:35 but you could have a separate brand
0:36:37 for all of these separate subgroups
0:36:40 and have the same underlying infrastructure
0:36:41 behind all of them,
0:36:44 but again, part of this is just how do you brand
0:36:45 and how do you market effectively?
0:36:48 Because if you look at the efficacy of health IQ ads,
0:36:50 or the efficacy of SOFI ads,
0:36:51 they’re so much higher,
0:36:53 because again, you have this large group of people,
0:36:56 or in many cases small but valuable groups of people,
0:36:58 that feel like they’re being treated unfairly.
0:37:01 So yeah, Geico is save 15% on auto insurance,
0:37:03 click here.
0:37:08 Mormon car insurance advertised to LDS members in Utah,
0:37:11 shooting fish in a barrel,
0:37:13 that’s gonna have a dramatically higher click rate,
0:37:14 and then many of these products
0:37:16 are also very demand elastic.
0:37:19 So I’m not saying save 15% on car insurance,
0:37:22 I’m saying save 80% on car insurance,
0:37:24 it’s very easy to do, click here,
0:37:26 positive selection bias,
0:37:28 that’s gonna work better than like Geico,
0:37:30 but we also have something for Mormons too.
0:37:31 – Right, yeah.
0:37:34 The goal is to find the LDSers and the hyper-myelors
0:37:36 who are really safe, et cetera, et cetera, right?
0:37:38 And so it’s very counterintuitive,
0:37:39 because if you’re at a big company,
0:37:40 you’re thinking scale,
0:37:42 how do I get the next increment
0:37:44 of revenue growth or profit,
0:37:46 and you’re saying actually go the other way,
0:37:48 don’t try to make your single brand bigger,
0:37:50 try to think about a dozen sub-brands,
0:37:54 each going after sort of the perfect market for them.
0:37:57 How do you positively select into a sub-market?
0:37:59 – Well the other side effect of this is that,
0:38:02 part of the asymmetric warfare
0:38:03 that some of the startups have,
0:38:05 is that if you wanted to kill Geico,
0:38:08 you wouldn’t steal 100% of their customers.
0:38:10 I mean if you did that, that would almost be too obvious,
0:38:12 you’d steal 20% of their customers,
0:38:14 but only the good ones.
0:38:17 So imagine that Geico could actually devolve or evolve,
0:38:19 depending on your point of view, into 10 sub-brands.
0:38:21 There’s no more Geico,
0:38:24 but it’s just like the 10 sub-brands basically select
0:38:25 for the right types of customers,
0:38:29 or even help judge and improve behavior
0:38:31 from other subsets of customers,
0:38:35 and then expel the 30% that are just bad news.
0:38:38 And if you can expel the 30% that are bad news,
0:38:41 you might say okay, well all of this dis-sinergy
0:38:43 of going from one brand into 10 sub-brands,
0:38:45 well that was idiotic.
0:38:46 Because now I have fewer customers,
0:38:47 but actually no it isn’t.
0:38:49 Because you might have fewer customers,
0:38:50 but it’s not like selling widgets.
0:38:52 You’re selling probabilistic widgets,
0:38:55 where in many cases you have negative gross margin
0:38:57 when you sell a widget.
0:38:58 So it’s important to figure out,
0:39:02 how do I get the good ones, keep the good ones,
0:39:04 and then get rid of the bad ones.
0:39:06 – Yeah, so that’s one strategy,
0:39:10 which is sort of sub-brands and sort of customer segmentation.
0:39:13 What if I’ve been told by main management team,
0:39:16 go find a bunch of startups to work with, right?
0:39:18 Sort of somehow figure out a marketing
0:39:23 or co-selling relationship so that we can start experimenting
0:39:24 with some of these new models,
0:39:26 and we can keep an eye on the startup community
0:39:29 so that maybe we can put ourselves in the best place
0:39:31 to buy them if it turns out working.
0:39:33 Is there a way to do that?
0:39:34 – Well there are many ways to do that.
0:39:37 Probably the easiest way that is often counterintuitive
0:39:39 for a lot of big companies is I call this
0:39:41 the turn down traffic strategy.
0:39:44 So Chase turns down a lot of people for loans,
0:39:47 either because, again, it’s the bin and ball problem
0:39:50 where it’s like, well, you might be good, you might be bad.
0:39:51 Sometimes it’s not even that.
0:39:52 It’s like, we think you’re good,
0:39:56 but we just can’t profitably underwrite a $400 loan.
0:39:58 But Chase has all the traffic.
0:39:59 So what is turn down traffic?
0:40:02 It’s saying, okay, we rejected you.
0:40:04 Hey, here’s a friend that you might like.
0:40:05 So this is not cream of the crop.
0:40:08 This is the bottom tier on the ingestion point
0:40:12 for a big financial institution saying we don’t want you,
0:40:14 which kind of is kind of a mean thing to say.
0:40:17 A way to ameliorate that potentially is saying
0:40:19 we don’t want you because we’re not smart enough to,
0:40:22 hey, sorry, we’re working on it.
0:40:23 All our systems are down.
0:40:25 But here’s a great startup that does.
0:40:27 Now why would you send customers to a startup?
0:40:29 Well, the number one thing,
0:40:32 that Geico spends $1.2 billion a year on advertising.
0:40:34 It’s really hard to compete with that from a,
0:40:37 so if I could not spend a dollar of advertising
0:40:42 but give 90% of my net income to Geico as a startup,
0:40:43 I still might make that trade.
0:40:44 I mean, we don’t always like this
0:40:45 because we wanna see do you have
0:40:47 your own acquisition strategies,
0:40:49 your own acquisition channels,
0:40:51 you’re not dependent on the big company.
0:40:52 But from the big company’s perspective,
0:40:55 turn down traffic is often brilliant.
0:40:57 Because it’s saying here’s somebody
0:40:59 that knows how to underwrite better than we do
0:41:01 or more profitably than we do,
0:41:03 we’re going to send our customers,
0:41:05 otherwise what happens?
0:41:07 And this is what I think Amazon got right
0:41:10 in an era where everybody else got this wrong.
0:41:12 Amazon said, okay, you’re on Amazon’s website
0:41:14 and you’re looking at the Harry Potter book.
0:41:17 And then right next to our Harry Potter book
0:41:20 is an ad for Barnes & Noble for the Harry Potter book.
0:41:21 Barnes & Noble is like, this is amazing.
0:41:23 We can buy ads on Amazon’s website.
0:41:24 They’re so stupid.
0:41:26 We’re buying ads and stealing their customers.
0:41:29 But every time you click on that Barnes & Noble ad,
0:41:31 Amazon made a dollar and it’s 100% gross margin.
0:41:32 They share that with nobody.
0:41:33 There’s no cogs on that.
0:41:36 And then they can use that dollar of pure profit
0:41:39 to lower their cost of their Harry Potter book.
0:41:41 Which actually made more people wanna go to Harry Potter,
0:41:43 we’ll go to Amazon to look for Harry Potter
0:41:45 than go to Barnes & Noble that said,
0:41:47 we’re locking you within our walls.
0:41:49 It’s like a casino with no clocks.
0:41:50 And we’re gonna pump oxygen in.
0:41:53 So because what a lot of big companies don’t get
0:41:54 is that Google is just one click away.
0:41:57 Like why give all the excess profits to Google?
0:42:00 When I go to Chase, I get turned down for a loan.
0:42:01 And then I go back to Google and I say,
0:42:03 where else can I get a loan?
0:42:05 Well, Chase should be sending you there.
0:42:06 And actually they’re starting to do this.
0:42:08 So that’s one strategy that I think has a lot of legs.
0:42:09 – Yeah, so turn down traffic.
0:42:10 That’s super interesting.
0:42:13 Look, you spent all the money to bring them to your site
0:42:15 and otherwise you would have just lost them.
0:42:17 That sort of sunk cost.
0:42:19 So you get something out of it.
0:42:20 That’s fantastic.
0:42:22 Well, why don’t we finish this segment out?
0:42:24 I wanna do a lightning round with you.
0:42:26 Which is I want sort of instant advice
0:42:27 for somebody in this seat.
0:42:29 I’m an executive visa or a geico.
0:42:31 And so I’m gonna name a category
0:42:34 and you sort of just have how to deal with startups
0:42:35 and you can react to it.
0:42:38 All right, so category one is
0:42:41 you should always invest super early
0:42:43 as early as you can into a startup.
0:42:45 – So again, remember adverse selection
0:42:46 versus positive selection.
0:42:48 So I would say the competence.
0:42:50 So this is what you have to get right,
0:42:53 which is if you take nine weeks to make a decision
0:42:55 and like, you know, we’ll decide within a day
0:42:56 or if Sequoia or Benchmark
0:42:58 or some other great venture capital firm
0:42:59 will decide within a day.
0:43:01 Like you’re not going to get good deals
0:43:02 if you take nine weeks.
0:43:05 So it can be very, very important to invest early,
0:43:07 but like the best things always seem overpriced.
0:43:09 Like this is something that we’ve learned
0:43:10 and it’s the same thing
0:43:12 with underwriting your own customers,
0:43:14 which is like if something’s too good to be true,
0:43:15 it probably is.
0:43:17 So some of the best things are actually very expensive.
0:43:18 Yeah.
0:43:19 All right.
0:43:21 – Just given those dynamics,
0:43:22 just wait for the later rounds.
0:43:24 Let all the venture guys take all the risk
0:43:26 and then like you plow in late.
0:43:28 That should be my strategy.
0:43:29 – I think in general,
0:43:30 that’s probably a better strategy.
0:43:31 But again, saying like,
0:43:33 ooh, we’re getting a great deal on this one.
0:43:34 That’s probably,
0:43:37 then you know that you’re the adverse selection
0:43:40 a source of capital as opposed to, okay,
0:43:41 here’s something I can’t believe
0:43:43 we’re paying this much money for it.
0:43:44 We have to fight our way in.
0:43:46 There are 10 other people that want it.
0:43:48 You probably know you’re on to a good customer,
0:43:50 if you will, or a good investment.
0:43:54 – All right, partner with as many possible startups
0:43:56 as you can, ’cause you don’t know who’s gonna win.
0:43:58 So let’s open up a marketplace.
0:44:01 Let’s 100 startups that I have either turned on
0:44:03 traffic relationships or something.
0:44:05 – I think that actually that does make sense.
0:44:08 I mean, there should be some kind of gating item
0:44:10 to make sure like maybe not a hundred.
0:44:14 But how do we stay close to different models
0:44:14 that are working well?
0:44:17 Because the main advantage that the incumbents have,
0:44:20 again, depends on like lending or insurance,
0:44:23 but it’s typically something around cost of capital
0:44:25 and something around distribution.
0:44:27 So if you have both of those
0:44:29 and you’re not using it to the fullest extent,
0:44:30 like you turned down a lot of customers,
0:44:33 like you should try to find an intelligent way
0:44:36 of using this and using that’s your unique thing,
0:44:38 like venture capital firms don’t have that.
0:44:40 I can’t fund somebody and send them
0:44:43 a million customers tomorrow, but Geico could.
0:44:45 So, but you can’t do that a hundred times.
0:44:48 You can probably do that some sub-segment of times
0:44:51 according to how much additional traffic
0:44:53 or whatever it is that the unique advantage
0:44:55 that you want to bring to bear.
0:44:58 All right, now on M&A strategy.
0:45:02 M&A strategy one, buy super early
0:45:05 before it’s proven to work
0:45:07 because presumably the prices are lower.
0:45:12 So M&A strategy early, focus on early stage companies.
0:45:14 – I’m a big fan of what Facebook’s done with M&A
0:45:15 and I encourage everybody
0:45:16 and pretty much every other industry to do this.
0:45:18 So Facebook has two formats for M&A.
0:45:22 One is we buy the existential threat that could kill us
0:45:24 and we price it probabilistically.
0:45:28 So surrender 1% of our market cap to buy Instagram.
0:45:29 That was way overpricable.
0:45:31 – Everybody said that.
0:45:32 – Wow, there you go.
0:45:33 – There’s a one in a hundred chance
0:45:35 that this is gonna be bigger than Facebook.
0:45:38 We should probably surrender 1% of our market cap.
0:45:41 WhatsApp, 7% chance or whatever it was.
0:45:43 I think it was 7% of Facebook’s fully diluted market cap
0:45:45 was spent on WhatsApp.
0:45:47 These were brilliant acquisitions.
0:45:48 Oculus, I mean Oculus hasn’t turned out
0:45:50 the same way that WhatsApp has perhaps,
0:45:51 but like same idea.
0:45:52 This could be the new platform.
0:45:54 If we don’t buy this and Apple does,
0:45:57 we are subject to their random whims and fancies.
0:45:58 So that’s category one.
0:46:01 Category two, and this is super counterintuitive
0:46:02 for a lot of companies,
0:46:04 buy the guys that failed trying.
0:46:09 Because they had the courage and the tenacity
0:46:12 to try to go and build something new
0:46:14 and that’s what you want in your company as well.
0:46:17 And then this is the most counterintuitive part.
0:46:18 Is like take the person that failed
0:46:21 and put them in charge of the person that was successful.
0:46:23 And that’s the part that, that’s breaking glass.
0:46:26 – For a big company, that’s so hard.
0:46:28 You reward your execs on success, not on failure.
0:46:30 – Right, but in many cases it’s like you have a big company
0:46:33 that’s been trying to build this thing for 10 years.
0:46:35 And if they build it, they will get one billion customers.
0:46:37 Because they have, or I’m making that up,
0:46:38 they have the distribution.
0:46:40 Then you have the startup that actually built the thing
0:46:41 in like a week.
0:46:43 And they built it for a million dollars.
0:46:45 And that would take the big company like a billion dollars
0:46:47 in 10 years to do.
0:46:48 But like, oh, the company failed.
0:46:50 Oh, that’s a bad company.
0:46:51 These are bad managers.
0:46:53 But actually you want to take them and put them in charge.
0:46:55 And the joke that I would always make is like,
0:46:59 if Amtrak buys Tesla, the worst thing that Amtrak could do,
0:47:01 because Amtrak is probably more profitable
0:47:02 than Tesla at this point.
0:47:04 But if Amtrak were to buy Tesla,
0:47:06 the worst thing they could do is say, okay,
0:47:09 all of you Tesla bozos, you work for us.
0:47:12 But the whole point of a lot of this other form of M&A
0:47:14 is you’re really trying to buy products
0:47:16 that you can push into your distribution.
0:47:19 And you’re trying to buy talent that wrote the products,
0:47:20 that built the products, that understand that.
0:47:22 And the only thing that they needed,
0:47:25 the only gap between them and actual huge success
0:47:27 is distribution, which these big companies have in droves.
0:47:30 – Yeah, so that makes perfect sense.
0:47:32 Maybe just a piece of advice
0:47:34 on how to actually make that happen.
0:47:36 ‘Cause you have this dynamic where you’re a big company,
0:47:39 you just bought a failing startup, right?
0:47:41 You have all of the execs inside
0:47:43 that have earned bonuses consistently over a year
0:47:45 for awesome performance, right?
0:47:46 You’ve rewarded success.
0:47:47 And now you’re gonna say,
0:47:50 I’m gonna take this guy that kind of failed.
0:47:52 And like, you work for them.
0:47:53 – Right.
0:47:55 – Like, oh, that’s hard to do inside a big company.
0:47:56 – It’s very hard.
0:47:58 But I mean, in some cases, you just wanna do it early.
0:48:00 I mean, I think it actually, where it works best
0:48:02 is where you say, we need this product.
0:48:03 – Yeah.
0:48:03 – We need this product to exist.
0:48:04 We don’t have it right now.
0:48:06 We haven’t spent eight years trying.
0:48:10 Rather than saying, let’s go assemble a team
0:48:11 and I’m gonna rely on like something
0:48:12 that’s just not in our core DNA,
0:48:14 here’s how we’re gonna go shopping.
0:48:16 We’re not gonna go shopping and value this.
0:48:18 And again, we don’t, this is not a self-serving comment
0:48:20 because if somebody buys one of our failing companies
0:48:22 for $10 million and we have a billion dollar fund,
0:48:23 it doesn’t matter, right?
0:48:26 Like we want the companies that actually beat the incumbents.
0:48:28 But the incumbents, the way that they can actually do great
0:48:30 is to adopt more of this Facebook mentality.
0:48:35 And like, the key thing is that many of these acquisitions,
0:48:36 these kind of aqua hire,
0:48:39 that’s the portmanteau of acquire and hire,
0:48:41 these aqua hire acquisitions that Facebook made,
0:48:45 these people now run big swaths of Facebook.
0:48:47 So I agree, it’s hard to do
0:48:49 if you already have a leader in place.
0:48:50 In that case, it just requires
0:48:52 a very strong-willed leadership team
0:48:55 and an actual overt strategy that this is what we do.
0:48:57 It becomes easier if it’s like, okay,
0:48:58 we’re trying to do this new thing
0:49:00 rather than assemble our own team
0:49:01 and they don’t know what they’re doing
0:49:02 but they’re well-intentioned.
0:49:05 Let’s go buy a company, but let’s buy a company
0:49:06 that hasn’t already done the thing,
0:49:10 but a company that tried and failed to do the thing.
0:49:12 But we’re pretty sure that these are the best triers
0:49:14 and failures in the business.
0:49:17 That’s the hard thing to really measure
0:49:19 because most people are used to measuring outcomes
0:49:20 and not process.
0:49:23 And the key thing to make this strategy work
0:49:26 is you actually want to over-allocate on process
0:49:28 and you want to wait outcome to almost zero
0:49:30 because you’re buying the outcomes that were in fact zero.
0:49:34 – Yeah, the mark is about to interview
0:49:37 Andy Duke, the thinking in bets, right?
0:49:39 And this is sort of the essential thinking in bets motion,
0:49:41 right, which is don’t confuse a bad outcome
0:49:45 with sort of a bad bet, right, right, exactly, awesome.
0:49:46 Well, thank you so much, Alex,
0:49:48 for coming in and sharing your thoughts.
0:49:50 For those of you in YouTube land,
0:49:52 please like and subscribe.
0:49:54 And for the comments thread on this,
0:49:57 I’d love to get your input on what you thought
0:50:01 of Alex’s idea that what you really should do
0:50:03 is not go after more customers,
0:50:07 but instead go after only the best customers.
0:50:09 So what are examples that you’ve been trying
0:50:11 in your own startup where you’re trying
0:50:14 to implement that idea?
0:50:16 So see you next time.
0:50:18 Go ahead and subscribe to the channel if you like it
0:50:20 and see you next episode.