Upstart Holdings, Inc. (UPST) Earnings Call Transcript & Summary

Let's get started. Welcome to Day 2, Citi Fintech 13. I'm Pete Christiansen, Citi's payments processors and IT services team. Great to have you here. Great to have Sanjay Datta, CFO of Upstart, join us. Upstart has been really supportive of our Axis event. So I appreciate your continued support. Great to have you here.

Good to be here. Thanks for having me.

Lots to talk about. I think Upstart really sits at a very interesting intersection of consumer and particularly what's going on in terms of credit performance. Let's start off a little bit more broadly. I want to get your view on where we are with the credit cycle. Obviously, this no credit cycle is the same. I think we can all agree on that. But the consumer has been through quite a bit, elevated comp being flushed dealing with cumulative inflation. And then you've seen different dynamics across those who are nonprime, "those who are prime." You've even seen average credit scores increased 30, 40 points for the average consumer during the pandemic now that perhaps is normalizing. So a lot of variables in that question. But generally -- and maybe we'll dig into some more of those throughout this conversation. But where would you think we are in the current credit cycle?

Yes, sure. It's been -- first of all, thanks for having me. It's going to be here. It's been a very unique credit cycle phase. In broad strokes, the credit performance in the wake of COVID and the stimulus in particular, was as good as we are seeing it largely attributable to the stimulus. At some point in maybe mid to late 2021, the, let's call it, the less affluent segment of the borrower base started to get very stressed. So they sort of reverted to pre- COVID default trends and then blew right through that and sort of skyrocketed over the course of 2022 is very unique and strange because normally, when a large sort of macro movement to default trends happens like that, it's because of employment stress. And Lord knows there's been no stress in the labor market. So it's been very, I think, interesting and confusing trying to understand why, and then we can get into some of the reasons. But the fact of the matter is that the less affluent borrowers in the country sort of blue pass their pre-COVID default trends. And by our measure, was somewhere on the order of 60% to 70% above it by late 2022. The more affluent borrowers in America have been more resilient for longer, but they've sort of operated on a bit of a phase delay in the end. So maybe a 6- to 9-month phase delay, but slowly, but surely, they were sort of following the path of the less affluent borrowers. These less affluent borrowers peaked sometime in 2023. And in late '23/coming into this year have started to show signs of normalization where normalization would be, call it, reversion to pre-COVID default levels. It is the primer folks now that have sort of crossed paths and are now driving the stress in credit performance. And what's interesting is -- so you asked me, so where do I think we are in this? When it was the less affluent borrower base that was really driving the stress in defaults? I didn't see it getting talked about too broadly because there's not a lot of public financial institutions that bank those people. We, I think, are one of them. So we were talking about it a lot, but I don't think it was sort of broadly in the narrative. But now that the primary folks are starting to sort of drive this stress, and I just got off the red eye from 2 days in Las Vegas is the annual Structured Finance Credit Conference. So everyone is talking about this there. So this is very much in the narrative now. And it's interesting because, obviously, these are much more visible borrower segments from the perspective of the markets, and everyone's talking about this happening. My view is that we're in the last stages of it because I feel like this is just something that is propagated from the stimulus and [ that this ] stimulus [ walked ] through the entire population. And now you're seeing it reach the sort of least risky, more segments and products because you've also seen it gone from unsecured lending to credit card and then auto and then more secured products. But I do think that from a narrative perspective, it's really somewhat nascent.

It's -- I think that, that thinking is probably very fair. I mean, you're right, job growth has been super resilient. You still have had pretty decent wage growth at the same time, but consumption patterns were elevated. It took time for the consumer to adjust to that, particularly with cumulative inflation, not expediting that whole process. And now it's them recalibrating their budgeting and things like that. So -- and working off that savings rate that they built up originally during COVID. But I think that does sound somewhat fair. You and I, we've talked about this, I guess, in a typical credit cycle, you see the less affluent borrowers get hurt first, then the more affluent. Do you think it happens, the recovery? Does it happen first with the less affluent side versus the more -- can that cadence happen? Or does it -- prime needs to get better first before the lower-end consumer can start seeing some improvements?

Well, I guess I'd say, first of all, I'd be careful to generalize too much because the shape of each recession is so different. I mean if you look at what happened to the GFC, it was the primary guys, the homeowners that got hit. So I don't know, in this one, in particular, the lesser prime, let's call them, the near-prime segments are already on their way to recovery. Early stages, but it's clear. I don't -- so I think it's -- this is just -- I think everyone is operating on a phase delay as the primer you are, the more resilient you were to the initial shock, it took longer for you to sort of go down that path, but the longer it will take for you to recover because this is more inertia. You're talking about bigger dollars. They did not get their initial wealth through government checks, they got it from asset wealth. And they did not go and buy electronics with it. They levered against their asset wealth and upgraded their homes. So it just took a lot larger -- sort of larger transactions and then, of course, the unwinding of it takes longer as well. So yes, that's how I view this one, but I wouldn't necessarily generalize from this one to whatever that happens next.

I want to get into some of the capital supply issues that Upstart has been dealing with the last couple of quarters. But first, before we get there, I was hoping you could share some of the lending perspectives you're hearing from some of your partners, your banking network, lending network there. And it seems that particularly since some of the volatility we saw roughly a year ago from today, banks have been a bit more incrementally cautious having more liquidity, that sort of thing. Do you see that dynamic playing into available funding that you have to offer on the platform? And how do you see that playing out?

Sure. Yes, the dynamics on the funding side are very interesting right now, and it's hard to know exactly where they're going to go. Banks are, just to crudely generalize, they're largely under stress for 2 reasons from a deployment perspective. One is they've had to contend with a contraction in the deposit base, which I think is historic in its sharpness, like the scramble to unwind the assets to contend with the contraction in the deposit base. And I know there's been a lot of talk about deposits moving between banks, but it's contracted in the aggregate, like this goes back to the credit problems that the consumers are having and the relationship between consumption and income and savings. But the savings based in this economy contracted very quickly, certainly in real terms. And I think it exposed a lot of vulnerabilities in the banking system's ability to unwind assets quickly in response a decrease in the deposit base. And then the second thing, of course, is the regulatory scrutiny, not just because of the bank failures, which has increased the scrutiny, but as Basel continues to get implemented in various forms, it's just increasing the punishment of the capital charges. And so you combine all of that, and you're right, the outcome has been that banks have been very conservative. In fact, they've been unwinding their asset base. So they've been shutting assets. We continue to hear that in the secondary markets, a lot of the buyers that we will be talking to about forward purchasing are very distracted by asset sales that are happening on the secondary, and we keep expecting that to run its course and it still has not. So it means that banks continuing to shed. And then I think there's even a shift in the strategic tone coming from banks, at least the ones we talk to where it's like one path, I guess, is that this will run its course and they'll go right back to being the large direct lenders in the economy. But that's not the signal we're getting from the folks we talk to yet. A lot of them are now increasingly focusing their direct lending activities on their customer base, their consumer base, so they're not necessarily buying assets broadly. And particularly because of the capital charges, there's a world in which they -- a lot of them seem to be signaling a pivot from direct lending to maybe looking a bit more like the insurance companies and taking senior security positions. So look, they have a deposit base, they need to earn yield, that's their business. But I think now there's an increasingly tendency to shy away from the full risk of the loan, which means they will deploy their capital as senior lenders and finance years and someone is going to have to take the risk of the loan as they do with insurance companies. And that's what's opening up the opportunity for this phenomenon in private credit you're seeing. Those are private equity funds that are building assets to deploy into credit, and they're looking for equity type returns. And in order to do it, I suspect there's a nice marriage there where they'll be willing to take the risk and take the high returns and the banks will be probably willing to take senior finance -- finance your positions. So that's a potential interesting model that I think...

Okay. So private equity, private credit players really filling that gap? Do you see that dynamic really starting to come out more? Has it been more spotty here and there because of maybe some of the pricing opportunities that are opening up in this area?

Yes. I think we're still in the turbulent storming phase of this because there is -- yes, there's a lot of money that's been raised. There's a lot of fire sales happening in the secondary, which I think are opportunistic in nature, but they're attractive. And we're still not -- we still don't have line of sight to where the credit performance itself is going because now the primary folks seem to be degrading and that's being noticed as well. So in the credit conference, I just came from, everyone seems to be gearing up for mobilization. But I think we're still in the phase where people are trying to pick the right funding with.

Yes. Time is everything. I would imagine, though, is at some point, there's going to be an opportunity, particularly when you think about like debt consolidation, right? I think the average American right now has $8,500 in credit card variable interest debt that's -- I think instant approval credit card for near prime is close to 29%, 30%. So obviously, there's going to be some opportunity for credit consolidation to be an important lifeline for many consumers. Did you hear any wranglings about that dynamic? Is that the opportunity that we should be thinking about once we get through working through some of these issues?

Not specific to that dynamic. But I would say, in general, demand for credit on the borrower side is high. And it's because their consumption is straining their income. I mean the core problem with the American consumer today to generalize that, they have no money in the bank at the end of the month because of all the things that have happened. And so demand for credit is high. The constraint is our ability to approve it because the loss rates are so high. So look, in general, higher credit card balances will lead to greater refi opportunities that will be constructive. But at the end of the day, the loss rates need to stabilize and get them approved before the approvability is there.

And then, it's funny. I ran into a gentleman earlier today, and he was asking me what panels you dealing today. I mentioned Upstart. He said, wow, that stock has really been around the horn, to say the least. It's been an interesting chart to look at. And Dave has talked about this a lot, building a more resilient, sustainable operating model. A lot of that involves branching out into different product sets. How are you thinking about changing the level of sustainability in the business model overall to have this all-weather kind of operation?

In the business model, yes, I think there's 2 components of that. One is on the funding side, developing a more resilient funding model. I think that in hindsight, pre-COVID in the times of plentiful capital, we are working with large number of funds and institutions on a very much [indiscernible] basis. And I think we had the illusion of diversification. Of course, it turns out when the markets turn, everyone acts in the same way at the same time. And so the strategy on the funding side is to develop much more resilient capital sources. It involves working with slightly different counterparties who have -- they themselves have a more resilient capital base, a more locked-up capital base, if you will. And it involves us putting some skin in the game in order to make that commitment committed. And then the second promise what you talked about, which is we are -- we, by design, the inception of our company was lending the riskiest product to the riskiest borrowers. That's, by its nature, the most volatile. As we get into more secured products and different economic structures that have more to do with -- less to do with upfront take rates and more to do with sort of ratable revenue over the life of the loan, I think these will serve as a natural buffer to volatility. That said, I think if you are in lending, and you're trying to apply technology to lending, there will be some twists and turns along the way. You can't completely isolate yourself from the macro. But I think we can significantly reduce the nature and the sharpness of the volatility.

I have to ask, since you just came from an interesting conference and you have some unique insights into the credit cycle itself, what are some of the things that you're hearing people are pointing to as potentially, all right, once we get to this point, we're out of this home, things are starting to get normal, is there like that? One data point is, we keep on hearing peak defaults are going to be mid-24. In your mind, what are some of the things people are looking for as a signal?

On the -- so on the consumer side, I think there's a recognition that this is about the imbalance between consumption and income and residual savings. And so I think -- I haven't heard many people sort of try and forecast the timing, but what people are looking for is an improvement in underlying personal fiscal health. And that needs to involve some combination of reduction in real spend or return to work or a real wage increase or something like that. It needs to result in the savings rates going up. On the institutional side, it really -- I do think the level of macro anxiety versus last year at this time at this conference is much reduced. I think the signal of obviously, the Fed cutting the rate for the first one. We'll take that anxiety completely off the table. So that would probably be a catalyst.

We've talked about it on a couple of calls benchmark rates changing also changes your available market, right, considering the 36% cap nationwide on unsecured lending. Should we think about that relationship now potentially going the other way as incremental same opening up more TAM, more availability to improve more borrowers?

As rates come down?

Yes.

Yes, absolutely. I mean you just kind of want to summarize the challenge our business has had in the last 2 years. It's that APRs have gone up significantly and it's pushed a lot of people above the 36% cap. Well, that's a self-imposed cap, by the way, it's not a regulatory one, but we don't go above that rate and because APRs have gone up so much, we've probably kicked 60% to 70% of the population out of the approval box. Now why have the APRs gone up? I would say 80% of the reason rough numbers is because of these loss trends we've talked about. And 20% of the reason is the base rates. So the base rates going back down would certainly be constructive. It's not the lion's share of the problem. But if they were to go back down to a couple of hundred basis points, each 100 basis points would probably be at a 10% to 15% conversion gain at this point. So it would definitely be meaningful.

Okay. That's super helpful. I want to talk about the evolution of credit decisioning and obviously, some of the technology advancements Upstart has had with its own platform, and we have addressed this topic a number of times at various events. But I would love to hear any updates you can provide to us. There's upgrades that are always happening, the core decisioning system is always being improved or it continually evolves. Just if you could take us through where you've come from a tech stack, tech capabilities, point of view in the last few quarters?

Sure. I guess the best way to -- the best framework for describing that is there's sort of 2 concepts when you're trying to price a loan and discern between borrowers. One is what we would call risk ranking or separation. And what that means is with a set of borrowers that otherwise look identical, can use alternative variables to try and relatively rank who is a good credit and who's a bad one. And then there's the concept of calibration, which means that all of those bars will get impacted if the macro changes and you have to recalibrate all of them. And the guts of the machine learning that we've been working on since our inception really goes to the first one of those 2. It's trying to distinguish within a group of indistinguishable borrowers or indistinguishable. A group of borrowers that have the same traditional characteristics, credit score, for example, how can we use their educational background, their employment history, other variables, things we observe about their digital interaction with us, how can we use those data points to really separate good from bad risk. And that's an exercise that has only gotten better interestingly. So if you get problems with credit performance in general, throughout all of that, our models have only gotten better in their ability to rank borrowers on a relative basis. The problem we run into, of course, is that we've fallen out of calibration because loss rates in aggregate have changed so much because of macro things. And so the second thing we've done and the majority of the investment we've done through these past 2 years is to get better at calibration. It's not a machine learning problem, unfortunately. So it doesn't play to our AI strengths. It is a data problem. But I think the best you can hope to do is detect and react to changes in the environment as they happen. We don't have predictive models, obviously, help with that. But I would say that over the past 1.5 years, all of the things that have gotten allowed us to get better at risk discernment and borrower level evaluation have gotten better. So more alternative data sets, more powerful algorithms, accumulation of training data over time. And the second thing is that we've -- I think we've got a lot of tools now that when the next recession happens, whatever form it takes, our ambition is to be able to detect and react to it more quickly and more precisely. So those are, I would say, the 2 sort of dimensions of technological improvement.

We've had a number of panels this conference so far talking about AI in general and the term that keeps on coming up is human in the loop. And it sounds like there was a degree of that no credit cycle is the same. You can't predict the future, but having a human in the loop this time, recalibrating the -- and adding more attribution to borrow statistics and those things has helped you enhance the platform quite a bit.

Yes. I mean I guess the way I would describe it is a machine model will fundamentally -- we use a particular flavor of machine learning or AI. There's -- AI has become such a buzzword. It's technology that I would -- the best way I can think of to describe AI is to take the word -- letter A out. It's -- take the word artificial out. It's like how would you describe intelligence. There's many different types of intelligence. It's used to do many different things. You can do -- use it to reason, to predict, to decide, to create. Artificial intelligence is just a way of doing that without a human brain. We use ML to predict, right? A lot of the AI we're talking about in the public consciousness is about creations, it's about content. We're using AI to predict. If you're using machine models to predict the future, they are using past data, and they're going to assume that that's the pattern that will hold in the future. If you, as a human have a reason to believe that that's not true, you have to tell the machine, right? Like it will assume -- whatever macro environment exists today, the machine will assume that, that will be the future because there's no good macro data on -- there's no good historical data on recessions, right? If you were to use the recessionary data you learned through the through the GFC...

It's all black swans.

Yes. But it also would be like all homeowners that would be, right? So to the extent you think the future will be different from the past, not at a borrower level because that's what machines are for. But in aggregate, you need tools to introduce that into the data sets. So that's -- those are things we think we're agnostic to before that we are not agnostic to right now.

I do want to get back to supply of capital, forward flow agreements you discussed before, talking to a lot of private entities and some of the opportunities there. But at the same time, risk sharing has become a bit more a part of the model. How do you think about balancing those 2 elements taking on more risk personally -- not personally, but from Upstart's balance sheet perspective and showing that you're willing to take on some of this risk to lure in and to attract some of these forward flow agreements? What are some of the pluses and minuses that we should think about as that becomes more as the business develop?

Yes. I mean, the rough equation is, as I said before, all of our funding was for lack of a better word, just that will come in and come out. I think on the go-forward, maybe a rough rule of thumb is, I would like half of our capital base to be locked in and committed. On that capital base, we will need some skin in the game. If you look at the rough numbers of the deals we've done so far that we've disclosed, we're sort of roughly 5% of that -- of the basis of that capital, okay? And so if we're 5% of half of our capital base, you could think of us as being maybe a low to mid-single-digit risk provider of our total platform. And I think that's a good trade. I would -- that would be acceptable risk tolerance in exchange for a resilient capital base. And then I think there will always be an important place for what you might think of as the spot market capital, capital because capital that has the ability to come in and out on the spot market will help create price discovery, and that will inform the larger strategic relationships over time in terms of "where the market is."

Yes. So there's a little bit of an intangible benefit to the market, having more liquidity in there and getting -- contracting potential on the market players. So it's not so much at least how I would think about it from like a cohort basis, all right? So this forward flow something happens in the economy that's unfortunate and maybe you have to take some losses on the risk-sharing side. You're still from a diversification in terms of funding flow overall for a particular quarter or a year, you're still profitable, right? Do you think about the scenarios right there?

Well, let me -- so I think the core concept behind these strategic partnerships in which we have a risk basis, the core underlying belief that we have -- so they're designed to not just be risk sharing to the downside, they're sort of reaching to the upside. So the core question for us is can we create as much upside in benign periods as there is downside in troubled periods. And that's an equation we think any given vintage that we originate next month, God knows that, anything could happen to create underperformance. But over a cycle, I think we're very confident that overperformance will pay for underperformance. So the mechanism we're trying to set up here, and that's the reason why these agreements need to be durable. They need to span a cycle. I think we believe that we can create almost a version of a macro insurance policy. That's what we're doing here. We're putting risk at work. If the period is benign, our capital will grow. We're putting that away for a rainy day so that when there's underperformance, and we need to make buyers all to some extent, that money is there. So this is really about whether we underperform over cycles or not. If that happens, there will be financial risk. But I think the history would show that even though we've had a period of underperformance because of all the things we talked about, I mean if you go back to 2018 and you look at what somebody would have yielded by investing ratably across that period, you're still coming out ahead. So yes, to me, it's less about the risk of any one period and it's more about how we are able to move money across the cycle from good or bad.

That's interesting. Also, what helps in the cycle is getting into new products. You've certainly, in the last 2 years, started to build up your auto program, now HELOC. First, on auto, can you give us a sense of where you are, obviously, what's happened with consumer lending in the last year? How has that impacted some of your growing plans in the auto side, where you are versus kind of initial planning? And how should we think about that portion of the business evolving over the next 12, 24 months?

Yes. Auto lending has been impacted by the same phenomenon that our core business has been. It's -- as we said, it sort of started in the riskier corners of the economy from a product and a borrower perspective and sort of washing its way through. So what happened on unsecured lending is now being fast followed in auto. So we unfortunately tried to incubate the business at a time where the environment is very much a moving target. As a result, we're sort of hunkered down. As we are in our core business, we're sort of letting that wash through. We don't have the history that we do in unsecured lending to be able to work through it with funding counterparties. So I think there'll be a period of time where we let things stabilize, and then we're still very bullish on that business. In the meantime, the distribution model for auto lending has been going really nicely. And the distribution model, in particular, is us giving software to dealerships that they use to actually sell the cars. So if there's not a lot of lending happening, there's a lot of GMV flowing through our footprint. And when we think the environment is conducive enough for us to turn on the lending, the volume is there. So that's exciting. HELOC is much more nascent, but it's also something that I think is a very strong fit for the environment that we're in, in the sense that the one thing the funding markets do want right now is loans backed by home equity. So that's a product that...

As it relates to HELOC, those, specifically, it sounds like there's a lot of inefficiencies in that distribution model currently today. Like what does Upstart bringing to that market to differentiate itself and hopefully grab share?

That's a great question. Maybe just backing up, if you were to try to boil down what value we're creating as a company at all, I think you could sort of point to 2 things at the end of the day. One is we aspire to have a more astute evaluation of the borrower level risk by having better algorithms and data. And then the second one, which is related, but it manifests itself differently is we want to have the best least frictionful consumer experience. That's a part we don't talk about as much, but the manifestation of it is I think now 89% of our loans don't touch a human. And I'm not just talking about the pricing of the loan, I'm talking about the fulfillment of it. So normally, you apply to it for a loan, you give a bunch of information, they'll price it. And then when you say you want it, you have to go through a process called verification where they -- you have to sort of make sure they're not lying to you or defrauding you. And that tends to be very cumbersome. Our fraud models are now at the level of sophistication such that 89% of the loans you apply for it, you get the price. You push I want it, you get the money with no degradation in credit quality. So that's very unique. I don't think anyone else can really claim that. So when you think about HELOC, there's not as much value to bring in terms of the underwriting because the loss rates are already very low. It's not like we're going to have a brilliant sort of lower assessment of the risk. But there the value is that it takes -- the industry average to get a HELOC is 39 days. We're down to 9, and our target is 5. I think the theoretical limit is 3 because when you involve home equity, there's a regulatory cooling off period that's required after you accept a loan. So we're already like, I would say, substantially better from an experience perspective than the industry, but we're not where we want to be. So every month, we try to cut off half a day or something like that. But yes, the play there is to just make it very seamless and fluid. But of course, it takes a lot of automation and a lot of good fraud handling to be able to automate disbursement of funding.

Sure. We're almost out of time, but I just want to finish off with how you're thinking about go-to-market in this environment, particularly for some of these newer products, obviously, auto? You're kind of keeping things steady as it goes right now until the environment improves. But certainly about HELOC, what's the next evolution of that category ramp for Upstart in terms of go-to-market and becoming more known for that product?

Yes. I mean the initial for HELOC -- I guess, certainly, first of all, it's limited to certain geographical regions because it's a product where you have to go state by state and get the right licenses. So we're sort of in the process of that. But from a marketing perspective, we just limited it to our own CRM database because we have this large list of people who have applied for personal loans and we could see who has home equity. So we can just reach out to them and based on their profile, we can try -- yes, who might be a good candidate. We're not at the stage where we're starting to do actual outbound marketing. Those will take the same form as our core business. There's going to be a mailer program, a digital program. There's some aggregators that have traffic that we can tap into. So I think it will look fairly similar...

Learn along the way before something more broader outreach...

Exactly.

Exciting. Well, we're looking for that turnaround. And hopefully, we'll see things improve on the consumer front. Sanjay Datta, thank you so much for coming. Great to have you as always.

Thanks for having me. Thank you.

Thank you.