Machine learning to power a fintech revolution, with Jeff Keltner

The first really big event that I spoke at during my time in Asia was in Shanghai. I was at Lendit China speaking about the emergence of fintech lending in Hong Kong, following a template my US colleagues had first designed to explain the rise (and rise, and rise) of lending fintech there. So I’ve always been interested in the forces that came together to give fintech lenders their first footholds in the American personal loan space, and then allowed them to power the subsequent growth in that sector.

For a decade, Upstart has been one of the forces powering that growth I’d previously just studied from the outside - which is why I was so happy to get Jeff Keltner onto the show. Jeff has been with Upstart pretty much from the start, so we talk about their philosophy, the emergence of fintech lending, and where innovation is happening today.

And Jeff has his own lending podcast, Leaders in Lending, which you should listen to wherever you’re listening to this one or find its links at https://www.upstart.com/lenders/podcasts/

Then, if you’d like to learn more about Upstart and the work they’re doing, head over to https://www.upstart.com/lenders/

You can learn more about myself, Brendan le Grange, on my LinkedIn page (feel free to connect), my action-adventure novels are on Amazon, some versions even for free, and my work with ConfirmU and our gamified psychometric scores is discussed at https://confirmu.com/ and on episode 24 of this very show https://www.howtolendmoneytostrangers.show/episodes/episode-24

If you have any feedback, questions, or if you would like to participate in the show, please feel free to reach out to me via the contact page on this site.

Regards,

Brendan

The full written transcript, with timestamps, is below:

Jeff Keltner 0:00

We started with one, which was 'how risky is it to lend Brendan a certain amount of money?' Obviously, one of the key questions people always ask about new approaches to underwriting credit is what happens during moments of stress. And it's a question you can model but ultimately the answer will be, 'we'll find out when there's a moment of stress'. And we saw that the correlation for a need for a hardship programme was much more tightly correlated with the risk assessments of our models than they were with traditional metrics like credit score.

So if you wanted to limit risk in your portfolio, as defined by impairments during that period, you were much better off saying I want a low risk Upstart loans than you where saying 'I want high credit score' loans.

Brendan Le Grange 0:42

I probably shouldn't have used up that 'why I like podcasting' intro the other day, because that's exactly how Jeff Keltner first came onto my radar, via his own excellent show Leaders in Lending which is, of course, on all major platforms and well worth checking out.

But that's okay, Jeff has the sort of career that allows for many angles of introduction - and in today's case, we're going to concentrate on his work with Upstart. Jeff stepped onto the FinTech train just as it was gathering speed, so we chat about his career, how the lending landscape has changed in America over the last decade, and the role that Upstart played in shaping what we see today.

Now, I have worked in consumer lending in Africa, Asia and in Europe, but my knowledge of the American markets is purely as a spectator, so I'm really looking forward to this one. Welcome to How to Lend Money to Strangers with Brendan Le Grange.

Jeff Keltner, welcome to the show. You're currently Senior Vice President for Business Development at Upstart where you've been for the last 10 years, so pretty much from day one. But before that, you were with some of the biggest names in tech, vintage names like IBM and internet giants like Google. So for me as an outsider, you look like the poster child for 'tech guy becomes a banker', you know, the the origination of fintech. Before we get into that, though, would you mind kicking us off with just a bit more on that background? How did you come into financial services?

Jeff Keltner 2:22

Yeah, so I studied Computer Engineering and college in quickly determine that I liked it, but wasn't as good at it as other people were. So I built my career on the sales, marketing, business development side of technology. And particularly, I think, looking back, where technology meets other industries and applications, not the pure tech.

So I spent some time in IBM and sales, I spent some time at Google before cloud computing was a thing, selling what became Google Workspace and Google Cloud, so in the very early days of that. And then when Dave Girouard left Google and founded Upstart, that was the same kind of thing: we said, these technologies exist, and they have an opportunity to make such an impact on financial services. In our initial belief, really, we were focused on credit. And we just said, 'hey, there's a such a massive opportunity to take these things that are being used for ad targeting and driverless cars, other kinds of technologies and apply them to a foundational element of the economy, like the way we decide who can borrow money, and how much and at what cost'.

So that was really what drag us into the financial services space.

Brendan Le Grange 3:17

That was 10 years ago now and Upstart has grown alongside, really, the upswing years of FinTech lending in the US. (That's right).

The way I've been told this story - I've never worked in the US, but I used to work with TransUnion and I'd speak to my colleagues there, they would tell it like this - in the last financial crisis, we saw, obviously, property prices collapse; and so home equity lines of credit disappeared from the market; this created a set of demand from low risk customers who now needed credit in a different way; but the traditional personal loan lenders had either all been wiped out, or just had their fingers burnt and so had stepped back from the market. So there was this moment of a vacuum where there were no traditional personal loan lenders but also there was fresh demand, and that demand wasn't really willing to put up with the old difficult situation that used to be involved in getting a personal loan. And so that allowed fintechs to step in - they stepped in, but more than that, it's been - don't quote me on these numbers, but - essentially the fastest growing product in the consumer credit space.

You would have been right at the heart of that, so what was that like? I imagine there was a little bit of uncertainty when you started into financial services, and then you step right into this rapid change?

Jeff Keltner 4:26

Well, I mean, I think that our core belief from the very beginning was a little different than many other fintechs. So I think your story is broadly right, but let me give you a different angle to think about this, which is it's very true that there were some unsecured consumer debt pre-crisis and that the banks more or less exited that business entirely during the last financial crisis, and that that left a vacuum for FinTech player. This had been, I think, the other part of that you can miss is that the technology that was emerging enabled a new way of transacting, right, like it wasn't possible before the mid 2000s to issue a purely digital loan, right?

And so almost all the processes were manual and not because people were behind the curve, but because it wasn't really possible yet. And I think part of what we see is that the personal loan was the first part that you could really do it, because you didn't need electronic lien and titled transfer, you didn't need that appraisal of a home the way you might in a HELOC or mortgage, it was really a product that could be end-to-end completely digital - and fintechs met that need. But I think it was as much the capability of doing things in a totally different way that was unlocked by the reality of the internet and the reality of an unsecured product, as it was purely just a void left by the banks.

It was very interesting to step into that. We were not the first player in the space, we actually entered the market with a relatively novel financial product that we didn't see as much traction as we wanted, and so pivoted into loans when there were already other FinTech lenders there that were bigger. And I think the thing that we saw, that I think most others didn't focus on as much, was the ability to better underwrite credit, the belief that more people are credit worthy than have traditional prime credit scores, and that if you could use modern data science, you could find those people and you could extend credit to more people and lower the price of credit to some of the people that were already being approved.

Because the world is is actually less risky than we think if we can really tell the difference. And that was, I think, less core to many fintechs, who focus on the digital cost reduction and less on the ability to actually underwrite credit.

Brendan Le Grange 6:16

And it's probably worth pausing here to talk about about what Upstart is and what you do.

Jeff Keltner 6:21

Yeah, you're right, what we really do is power banks and credit unions to offer loans on a common technology stack. And it's interesting, because I hear these FinTech partnerships, and that banks must choose between 'build versus buy'... and I think it's a bit of a red herring, frankly. Just because, if you look at the history of technology in the banking industry, they've almost always been partner or buy, the core providers, there's a core set of them, and everybody buys it, nobody writes their own core providers. So I think there certainly is a moment in the emergence of a new category when there aren't as many providers that have the robust capabilities that people build from scratch, but I think it's very natural that we see a common platform.

So we use a common technology stack across a number of partner lenders who actually originate and then hold the credit risk. So we're not a lender in that traditional sense, where we're making the loans or holding the loans on our balance sheet, we're really working with financial institutions to help them you know, better reach and better serve the consumers with you know, unsecured personal loans, auto refinance loans, increasingly auto purchase loans as well. So you know, we're, we're expanding from that initial core of the unsecured loan and taking it into other kinds of consumer asset categories.

Brendan Le Grange 7:22

You're listening to How to Lend Money to Strangers, with Brendan Le Grange, and today's guest, Jeff Keltner. If you're enjoying it, now's a perfect time to hit that little plus button to subscribe. Now, let's get back to that show.

But you filled that gap in, so you haven't just handed over a bit of data to say here, we're really good at alternative data, here's an improvement on a traditional FICO score, you've built that platform, basically taking all that friction - or as much as possible - out of the process?

Jeff Keltner 7:49

That's right. Well, I think we've really applied machine learning to data.

High level, I would say there are four elements of the lending process, and we started with one: which was how risky is it to lend Brendan a certain money? Like what's the likelihood of repayment or default, and then really allowing our lenders to specify, are they comfortable with that? How do they want to price it?

Increasingly, I think to the point you're making, we are also applying it to the second area we really focused on, how do I reduce friction in the process, right? We didn't start with this insight. This is kind of one of those you learn in the market. When we started, every borrower that our lenders were onboarding, we did a phone call, we asked for an ID to be uploaded to verify identity, that standard kind of KYC stuff that you do. And we had this insight, like, for the small loans, it costs too much money to get on the phone with people, maybe we could just use automated signals to do fraud prevention and not get on the phone, just for small loans, just for a few, to see what happens.

And so we tried it. And we saw this 2x to 3x increase in pull through and actually equal or positive credit performance. We went, 'oh, that's interesting, if I can take a certain amount of demand and turn it into twice as many loans, that's really valuable'. So we started the process of saying, can we use machine learning to get to a place where we're comfortable with more loans of larger sizes of longer durations that we can approve without that human intervention, because it both lowers the cost, but it reduces the friction.

And it turns out, consumers are not only rate sensitive on the loan side, they're also friction sensitive, they don't like putting in a lot of effort. So we are now at a place where our lenders see 70% of loans coming through the platform, having no touch origination - with ID verification, income verification done in automated ways, with very high NPS and very low cost and high conversions as a result of that.

So I mean, think of that as like the key two ways we've applied machine learning and then it makes such a tremendous difference and the experience for the consumer. It's amazing. I still talk to financial institutions today that maybe are running digital originations, but they say 'hey, we only support current customers on the digital channel'. Why? 'Well, because our fraud prevention mechanisms are based on in person interactions or based on branch experiences'. And so it's a whole different thing and it's easy to say you want to digitise the things, but then to take it you know, the first step of digitization from it was like, 'well, I had this paper form so now it's a digital form, but when you finish it, it's still like emails to the person you would have put it on the deskt of and we wait for them to call you back'. And, you know, process was still the old process, we just had a digital front end to get into it, right?

And I think that's, that's the first step. But there's so much more we can do. And the world has evolved a lot since then. But it's a hard problem, you know. I think banks that have gone to digital depository account opening have seen really high rates of fraud and struggled with that, and it is obviously, in some ways, a harder problem when you're lending because you're not taking their money, you're giving them your money. In that case, you know, it's a little harder, and you're out the money if you're wrong. And so the risk is a little higher. So I think it's, it's something that a lot of institutions are still, frankly, struggling with.

Brendan Le Grange 10:30

And it is more risky, and they've got the risk of the compliance as well. But for us, as customers, we seeing one click to buy now on Amazon, like a single click button, you get your iPhone, it takes a look at your face, it opens and it accesses your accounts, we've become so familiar with that, that the banks are, they are in a difficult position. They've got more restrictions than most, but the same time expectations. Yeah, very different from from the ones where we could shut down internet as a channel. (Yeah, the world has changed).

Given the timing, you would have been founded coming out of the last financial crisis and fully going and a big part of lenders lives when we hit the COVID times, which obviously started with a lot of uncertainty. And bankers being conservative, I imagine several people trying to get hold of you in the first couple of weeks, wondering what's happening? How did you see your portfolios perform during the worrying first part of COVID? And what have you seen in terms of a score stability or performance just of these modern models, did they hold up to the first big crisis?

Jeff Keltner 11:32

Yeah, it's a great question. I will say COVID is a pretty unique experience. I mean, it's a once in a century pandemic, a kind of unprecedent level of government stimulus, you saw, you know, hundreds of 1,000s a week unemployment filings. So we've seen some really strange things.

I'll give you the kind of early story and then maybe what we see now. So obviously, one of the key questions people always ask about new approaches to underwriting credit is what happens during moments of stress. And it's a question you can model, but ultimately, the answer will be, we'll find out when there's a moment of stress. it's like the Warren Buffett quote about how we 'find out who's swimming naked when the tide goes out'. And so it looked like the tide was going out. And what happened for a lot of lenders, and what all of our partners chose to do, was to put very generous forbearance and hardship programmes in place for people who couldn't make payments. And so we saw what we would call impairment and the portfolio, right, there wasn't really losses or delinquencies, they were just people who were no longer making payments for some period of time.

And we saw that the correlation for need for a hardship programme was much more tightly correlated with the risk assessments of our models than they were with traditional metrics like credit score. So if you wanted to limit risk in your portfolio as defined by impairments during that period, you are much better off saying I want low risk Upstart loans, than you were saying I want high credit score loans.

And that stress did not generally result in delinquencies over the medium term, because government stepped in. It was a very odd thing to see very high unemployment rates, and extremely low default rates on loans, which is like not usually think of unemployment being very tightly correlated with risk in the credit market.

And so you saw strong over performance on almost all portfolios - I don't think that's unique to Upstart. But we did see, again, correlation there with our risk tiers being being better predictors of that.

And now you're seeing that the stimulus has been generally come to an end, you're seeing what I'll call more of a reversion to the mean, the way I would describe it, it's pretty rapid, right? Unlike most recessions that kind of start slowly, you saw the opposite here, which was a very immediate shutdown, in a very immediate withdrawal of the support that had been maintaining really high credit performance, I think we're, we're seeing something more like what you might have called pre COVID, levels of losses, and the the models have continued to be much better separators at risk pre and during the stress than more traditional metrics. But you're certainly seeing this really interesting credit cycle.

Brendan Le Grange 13:37

I suppose it is in some way connected to COVID and then the costs of all the government programmes, but also everything else happening in the world, but now, as we were coming out of COVID, we've now got bigger financial issues going on in the in the market muddying the waters.

Jeff Keltner 13:51

This is definitely an interesting time. I've made the comment to people lately - and we'll see when this is like when this airs - but I've not lived through as many times where your 50% variance in likelihood of what the world looks like in six months has been wider, like normally, it's kind of a relatively narrow tier... now, I mean, it's a pretty wide spectrum for what I think is somewhat likely to be the good and the bad, the upside and the downside of that. So it's a really interesting time to be managing a lending portfolio or doing anything really in the financial sector.

Brendan Le Grange 14:20

So you've come through your first big crisis with your models. You've also in those 10 years moved from the startup to now public traded company, $800 million a year in revenue, 2000 employees, a big part of the American lending economy. Has that changed how you at Upstart look at your business, the sort of customers you're bringing on board? Are you seeing now the big establilshed names moving in this direction or have you seen that landscape change now that you're so well established?

Jeff Keltner 14:49

Yeah, I guess maybe this is an early employee/ founder bias, but I don't think it ever feels as well-established to those of us building it, as it does to the outside. But we tend to look at what we built as a really great set of technologies, but our partners are still a minority share of the personal loan market, which is maybe the smallest consumer credit market in the United States. So we kind of look at it as still so much work to do to take these technologies and apply them to a broader set of lenders that are touching a larger set of consumers, right?

And then to take it to... so we've kind of gotten started in auto refi, we bought a company that has auto retail, and so we've got auto retail lending, kind of in-dealer purchasing, but then obviously, we're still not touching real estate, we're not touching SME, you know, we think these technologies will have a lot of ability to provide products for people using payday loans. I'm not saying we're going to be a payday lender, but people who have short term liquidity issues, who need a more short term shot, smaller dollar loan, these technology should enable a much better option for them than maybe what you see in the payday industry.

And so we just look at it and say, Hey, we've proven that we know how to do what we do. But we feel like we've done it in a pretty small segment of the market. And so mostly, we just see how quickly can we go out and start applying this technology to help more consumers on more kinds of products in partnership with more lenders. So it still feels like what Jeff Bezos says 'it feels like day one at Amazon', I think we we still see so much more ahead of us than behind us that our executive team founding team tends to be what's next? Where are we going next? What's the next challenge to tackle kind of team so it doesn't feel that established?

To me, it just feels like hey, we've got a good foothold. And now what are we going to do with it?

Brendan Le Grange 16:21

Yeah, and I think that goes back to your point early on that personal loans is that sort of most simple of lending, not simple in who do you lend to, but simple in terms of it doesn't need the credit card systems or mortgage systems and such in place. But all those other systems are either becoming much more digital themselves, or possibly being replaced - or look to be replaced by - buy now pay later and things coming in and taking away some of the card market. So I think anything based on making it seamless for consumers is going to win.

Jeff Keltner 16:50

I'd say I agree and disagree with you, depending on your perspective. Like, personal loans is the easiest in terms of there's no merchant you have to pay, there's no house to value, there's no car, there's no repossession, I mean, it's kind of like the simplest on that level. It's also in many ways the most challenging to underwrite. Because there's no car to take back. There's no strong incentive to pay, there's no house, there's you really the only thing you can do is hit somebody's credit report, right.

And so we felt like it was the perfect place to build and test our underwriting engine because it's kind of the truest test of lending, you've got to determine ability to repay and likelihood to repay and you've got no backup. So we like the starting point there.

But you're right, there's in so many industries, there's a shift happening in terms of seamlessness of transaction we saw during COVID, many geographies started to allow supporting more electronic lien and title transfers for automobiles, supporting at least digital signatures for some documentation that used to require wet signatures - because that was kind of needed in COVID,and now you're seeing a lot of go 'oh, that worked pretty well, maybe we should just stick with allowing digital signatures', which of course simplifies the process for getting a home loan or an auto loan where were those kinds of transfers are required.

So I do think there'll be a lot of wood to chop, so to speak, in the making the experience better, which is kind of where we're focusing a lot of those other products, in addition to the 'can we take the consumer population and approve a larger number of them'. And that will always be part of our TrueNorth.

Actually, I should have said, we did a study with TransUnion that said 80% of the American population has never defaulted on a credit obligation. And yet less than half has a traditional prime credit score. We just look at that and go 'man, we should be able to approve so so many more people than what a traditional credit score would allow you to do if that's what you're relying on as a lender'.

We still see that as a massive part of the opportunity in every sector.

Brendan Le Grange 18:28

It's a great point, because I do some lending in pretty high risk markets - we talk about 30% default rates, but still, that still means 7 out of 10 people are paying their loans back!

Jeff Keltner 18:37

That's right, that means you're declining two and a half times the people that are going to default who are good borrowers, because you didn't know which ones they were - so that's crazy. Like, you got to think 30% on the test, like that's bad, right? That's to me, where there's so much opportunity to actually improve the, you know, the accessibility without introducing, I'm not saying banks should be take on more risk. It's like there's really low risk stuff, if we can identify it in my my suggestions, because we can it's identifiable, it just you got to get much smarter understanding which of the 30 and which are the 70 in that high risk market.

And when you do that to 70 look like low-risk people.

Brendan Le Grange 19:14

Yeah, and it is such a more pleasing approach. When I started high risk lending was solved by just adding more APR to the mix and saying, 'well, we'll lend to you at this price'.

And we used to do quite a lot of modelling and see what the risk would be, but the limits of the technology meant that you would adjust the loan size and things but essentially say 'yes' or 'no' and 'yes' meant at the APR we could do it for, take it or leave it.

Whereas finding new ways to get rid of those blind spots in our data to say 'here's a low risk one, here we can get more and more people into lending at affordable prices'. And sure, still try and catch those people where the loan is not going to do them any good, where the risk of them is too high ,but get some more people through the door.

Jeff Keltner 19:55

Yeah, my co-founder Paul this great kind of construct that really drives home to me, the reality is "every loan that defaults shouldn't have been made". And it's as bad for the borrower as it is for the lender. And every dollar of interest that's above the kind of fixed rate of return, the cost of funds for bank, is really a cost that the good borrowers bear because we're not good at figuring out who they are.

And when you think about the delta between the average consumer's price for a consumer loan, and the cost of funds of a bank, you go, that's a big price that a lot of people are paying. If we had a perfect model, no one would have to pay you to prove, only those who repay you so you could give them all the lowest rate.

And that's how you can think of our model as being a lot better, but then we look at in that context and go, man, it's got so much distance between us and that perfect state of a model that can approve 70% of the people in that high risk pool you have and charge them all the lowest interest rate. That's a tremendously different world than the one we live in. It just kind of shows you how much opportunity there is.

Brendan Le Grange 20:45

The headline stealer over the side of the Atlantic is all buy now pay later. And for me, one of the upsides of the model I see is that essentially it's free, as long as you pay it, so it's kind of another way around that same problem that well, as long as you're paying, you're getting interest free. And then we hit you with the late fees, if you miss it, which in some ways is saying well, the right people are paying.

But of course, if you can't afford your your instalment, you almost certainly can't afford your late fees. Yeah, ideally, would have just said no to that person. But hopefully, through tech and interesting business models, we can come through a route that's better than ours, where we would start with what are you essentially what's the government allowed maximum and work slowly downward from there?

Jeff Keltner 21:26

Yeah, I think by now pay later is a really interesting innovation in certain ways. I think there's certain questions about affordability in the right business models, but you're taking merchant value and using it to fund the financing for consumers, which is great. And the merchants are paying for a lot of that, and I think is the sharp end of the spear that we'll see in other places, which is the embedding of the financial transaction more deeply with the transaction, that it's fine.

I say this to banks all the time, but most of the loans you make to consumers, they didn't want the loan, they wanted something else. They didn't want a car loan, they wanted a car, they didn't want a home loan, they wanted a And you know, the idea that we would embed those inside the home purchase or the auto purchase transaction in a more deep way over time, particularly as those transactions shift to a digital medium. Just make sense. And buy now pay later is maybe the sharp end of that. But I think the same capabilities are going to be coming to you know, digital retailing for auto in, you know, as you find the big home shopping sites are trying to integrate financing. And I promise you, they're going to start making a more integrated experience for someone who's shopping for a home to finance that home at the same time.

Whether that's a free product or an interest bearing product over a longer period of time is I think that's semantics, but kind of details because that larger trend of hey, we can embed the financing into the the commercial transaction, and it gives a better experience for the customer. I think that's a real trend that's going to be solid and cross a lot of boundaries in terms of the markets it hits.

Brendan Le Grange 22:46

Yeah. And I saw that in our numbers early in COVID, where every time there was a hard lockdown in the UK, the applications for personal loans disappeared. So when there was nothing to do, nobody wanted a personal loan, but when we opened up a bit and people could travel, people could go out again, personal loans returned, and then we went to our second lockdown, and they disappeared again. So I mean, it sounds obvious, but you're right, that you're basically lending the money to your customer to give it to the shopkeeper. So, yeah, take out the middleman essentially.

Jeff Keltner 23:15

I also think that this may be different in different geographies. So maybe we're speaking different languages in Europe versus the US. But I do think there's also a secular shift among particularly younger consumers to be more intentional about their borrowing decisions. And I mean that to say like credit cards, I think you're often the consumer is not always aware when they crossed the line between transacting and borrowing and what the timeline and cost of the borrowing they may be doing is, I have the credit line, I swiped the credit card at IKEA. I know I can pay it back over time, but like, how long will it take? How much will it cost, how much - that stuff was kind of, you know, ambiguous to them was difficult to tease out, and I think BNPL also represents this desire.

I think you see it in the shift to debit card usage versus credit cards for some younger consumers. That's to say, I want to know what I'm transacting, I can do that on a debit card, I want to know when I'm borrowing and when I borrow, I want to know what it's going to cost me and how long it's gonna take me to pay it. If you're borrowing for two weeks, three weeks, a credit card is great. But if you're borrowing for six months, credit card is not the best way to do that. In many instances, it's difficult to understand the cost to you, the younger generation is saying, 'hey, I've seen my parents and credit card debt that they don't quite understand and how to how to get out of and like how do I set myself up where I understand the debt obligations I'm taking on as I take them on and that's a it's a little different thing. But I think a really interesting general trend for what's driving some of that behaviour.

Brendan Le Grange 24:32

Yeah, no, it's certainly a theme echoed by a lot of businesses here talking to young borrowers that there's really quite a well educated group of young customers coming up here yet much more intentional about it. And early on a credit card was the simplest spending tool. And for a long time, the only really convenient spending tool if you wanted to be able to make relatively large purchases, and you could borrow from it, and the danger came from the fact that you could go out intending just to spend and end up borrowing through the temptation or just a little bit of lack of discipline.

Jeff Keltner 24:59

And you never really knew when you were crossing the line. It wasn't clear, though I think it's an interesting problem.

Brendan Le Grange 25:04

You kind of mess up your future month as well, because if you paying minimum payment this month, I mean, you always could go and you could move things around. But it would have taken a lot of work to avoid maybe booking some international travel, you put your plane tickets on to pay down, but then you go to the shop to do groceries just afterwards, that's probably going to get on minimum payment, too, unless you're really being careful. So yeah, there was a lot of uncertainty.

Jeff Keltner 25:27

Yeah you got to really look at like, 'how much am I intending to spend, what is am I but like it there, it's mixed, right? That the two and that's I think we're for consumers, that confusion can come in where a lot of consumers I think like to separate like to go, hey, it's my transacting account, this my borrowing account, I want to know what I'm borrowing how much it costs when I'm paying it off. And I want to transact on the other one.

Brendan Le Grange 25:42

So it loops all the way around back again to the personal loan industry in that it is the simpler in terms of conceptually way to lend money, here's how much money I'm lending to over this time with these terms. And now we can do that, whether it's finding a wallet, whether it's finding a merchant directly, I think that side, the lending side, making it simple, making it easy fits a number of different business models on the front end.

But in terms of you guys at Upstart, that work you've done to do it quickly, too, because that's the other side of it, to be able to lend quickly to a customer much more sort of scoring them or giving their score more often for smaller amounts, is going to just be the normal way for lending, I guess for everything other than mortgage and auto loans for for a long time.

Yeah the credit card you would be that's the other thing is there a lot of today, you know, I got my credit card, the credit card I have now was approved when I was living in Hong Kong. And I was in Hong Kong for eight years. And I moved to England three years ago. And when I moved to England, American Express just asked you what is your old credit card number, they ping the guys in the other country. And based on your performance in the other country, they give you a new card? Yeah, so I guess technically three years ago, they checked my credit here, but in some ways it was 11 years ago, when last I went through the approval process. And of course, every month, they can see I'm paying and they can see my transactions, they can keep modelling me in the background.

But really, it was a long time ago, they did all their real careful checks. Whereas we were moving to a world where every transaction may be an individual credit, which means you need to be a lot clearer about the score, and you can't rely on things that take three months to filter through and only get updated every month. You want data that's fresher and more nuanced.

Jeff Keltner 27:15

Yeah, and I think this is a slightly off topic from what you're saying, but I do think one of the weaknesses with the score as a concept is that it doesn't scale well to different kinds of transactions: my credit score is my credit score, if I'm applying for $1,000 credit card or a $1 million mortgage. And sure you have different policies for those two, but there's no way that I can be an exceptionally good risk for one and an exceptionally bad risk for the other and blending them into a unified score doesn't make sense.

So detaching the credit scoring from the transaction for which you're asking about credit is kind of an odd concept to me, and I understand how we got there. But I do think, you know, we've moved to more scoring individual loans. And you're right, that there are companies out there that are already looking at how do i on a credit card transaction model like that?

How do I score the individual transaction, one of the things that I find now pay later guys will tell you as they get access to SKU level data to underwrite, which is different than the credit, they know what you're buying. And they can actually use that for some advantage and underwriting versus just knowing the merchant thinking about how granular can you make the decision you're making? And how much better a decision can you make about the level of risk it represents? On a tighter level? It's a really interesting kind of thing that we're moving to as data becomes more realtime.

Brendan Le Grange 28:21

Yeah, very interesting. And bringing up its own issues or complications around things like the the messaging to that customer who tries to buy a pair of sneakers and gets declined because it's the wrong brand of sneakers or sneakers are too highrisk....

Jeff Keltner 28:33

Just to be clear, I've never had to deal with and issues around declines for SKUs, not in my world but I think it is part of what BNPL does. And that's, yeah, it's an interesting element. And it's real question like it does correlate to loss.

Brendan Le Grange 28:48

One of the reasons that I sort of hear lots of different parts of BNPL. And whatever's happening in the marketplace today is from this podcast. And you know, I found you, from your podcast, Leaders in Lending, which I recommend to everybody. What was the story behind that? How did you get into podcasting? And I think you've been about two years that you've been going how's that journey been?

Jeff Keltner 29:11

A year and a half, something like that, a year and a couple months.

It's been fun. It was actually not an intention of mine. I sometimes get asked is like a career desire and is not really, but as we were taking our technologies into the banking sector - and as I said, Upstart primarily partners with banks and credit unions - we wanted a way to talk to them, to hear their perspectives, and to share some of what we had seen or we're hearing in the industry.

And we thought a podcast was a really interesting way to do it.

And so I somehow was tapped as the host. So it's sort of an accidental thing on my part, but it's actually one of the things I enjoy most right now of my responsibilities because they are typically really interesting conversations. I learned a lot from my guests. It's always interesting to be the host and try to figure out how much you're going to say and how much you're going to ask. It's interesting for me today to be on the other side of the microphone and being the guest but we really enjoyed it. It's been a took me a while I think to find my sea legs in terms of how to how to host and ask questions or whatever. I enjoy. And hopefully my goal every week is just put out content that gives people a new insights, interesting perspective, something that they can take back and make a practical use in their actual lives.

Brendan Le Grange 30:08

Yeah, I'm probably giving away too many secrets of the consulting world. But when I was a consultant, I felt like at least 25% of my job was just sitting down and saying "when I was working in this market, they used to do thi"s, or "I remember this from that place".

And it was before you started actually doing the real work. But just sitting around the table, it gives you this sort of sense of authority. And it gives you some context what's happening. And I think that's what these podcasts, can you just hear a little bit about what's been done differently and why on things that you would sometimes it's entirely unrelated, but sometimes will give people a little bit of an idea or a little bit of comfort and new idea to pursue. So it's obviously some of the style to my shows, I'm biassed, but I really like hearing it from people that are doing the work and able to talk about what's happening in a less structured formal way than the conference presentation where we're on our best behaviour. You're also obviously by the looks of it very busy outside of work with hiking and fishing and scouting. What else are you doing to sort of keep yourself occupied in these troubling times.

Jeff Keltner 31:12

I am a father of two young boys, my boys are 13 and 10. So pretty much whatever they're into is what I'm into, which now means scouting. So we're on a backpacking fly fishing trip next weekend with them. So fingers crossed, that goes well that's a that's a big one up so a lot outdoor stuff in my youngest is playing a lot of tennis. So I've had to refresh my tennis game, figure out how to be a good hitting partner, but mostly things with the family, the wife and the kids and the boys dictate the schedule these days. We just follow him around and try and be value additive to whatever their whatever they're doing.

Brendan Le Grange 31:41

As sounds great. I've got two younger girls and had got the golf putter out and some balls and trying to teach them to play crazy golf because we played for the first time the other day, and there was some interesting techniques.

Jeff Keltner 31:53

I tried the golf, right, and it was like that was my main sport and my kids just didn't it wasn't their thing. So now we're doing a bit of volleyball and a little bit tennis and I put the golf sticks away for a while but I tried. I wish you good luck.

Brendan Le Grange 32:05

I think it's the first price as a parent, it's a sport they can't really get injured in and there's a lot of upside if they're good at it. My eldest, well they both now love their horse riding there. So I fear I'll be in for the most expensive of hobbies

Jeff Keltner 32:19

Expensive and dangerous, yeah, perfect.

Brendan Le Grange 32:23

Jeff, thank you very much. It's been fantastic chatting to you and to sort of hear from from the inside about that rise of FinTech and enablement of FinTech in the US. It's been a pleasure having you on the show. It was great being here.

Jeff Keltner 32:35

Tahnks for having me, Brendan, I really appreciate it.

Brendan Le Grange 32:37

And thank you all for listening. If you enjoyed that, please do rate and review on your preferred podcast platform and share widely including on LinkedIn and while you they send me a connection request. The show is written and recorded by myself Brendan Le Grange, in Brighton England and edited with assistance by Kane Hunter, show music is by Iam_Wake and you can find full written transcripts now in several languages, show notes and more content at www.howtolendmoneytostrangers.show

And I'll see you again next Thursday.