Upstart Holdings, Inc.

Good afternoon, and welcome to the Upstart Third Quarter 2023 Earnings. Today's conference is being recorded. At this time, I would like to turn the conference over to Jason Schmidt, Head of Investor Relations. Please go ahead, sir.

Good afternoon, and thank you for joining us on today's conference call to discuss Upstart's Third Quarter 2023 financial results. With us on today's call are Dave Girard, Upstart's Chief Executive Officer, and Sanjay Datta, our Chief Financial Officer. Before we begin, I want to remind you that shortly after the market closed today, Upstart issued a press release announcing its third quarter 2023 financial results and published an investor relations presentation. Both are available on our investor relations website, ir.upstart.com. During the call, we will make forward-looking statements, such as guidance for the fourth quarter of 2023, relating to our business and plans to expand our platform in the future. These statements are based on our current expectations and information available as of today and are subject to a variety of risks, insurgencies, and assumptions. Actual results may differ materially as a result of various risk factors that have been described in our filings with the SEC. As a result, we caution you against placing undue reliance on these forward-looking statements. We assume no obligation to update any forward-looking statements as the result of new information or future events, except as required by law. In addition, during today's call, unless otherwise stated, references to our results are provided as a non-GAAP financial measure and are reconciled to our GAAP results, which can be found in the earnings release and supplemental tables. To ensure that we can address as many analyst questions as possible during the call, we request that you please limit yourself to one initial question and one follow-up. Later this quarter, Upstart will be participating in the JMP Securities AI Forum, November 29th. and Wed Bush's Disruptive Finance Virtual Conference, December 1st. Now I'd like to turn it over to Dave Girard, CEO of Upstart.

Good afternoon, everyone. Thank you for joining us on our earnings call covering our third quarter 2023 results. I'm Dave Girard, co-founder and CEO of Upstart. While 2023 continues to be a difficult environment for consumer lending, I can state with confidence that we're making rapid progress in building the world's first and best AI lending platform. We do this with a focus on our mission and with an optimistic eye toward the transformation of an industry that is inevitable over the next decade and beyond. In the third quarter, rates were at an all-time high in our marketplace, higher than we expected them to be, reflecting both decades-high interest rates and significantly elevated risk in the consumer economy. This is not a path we would have chosen and is obviously not constructive to our growth, but it reflects the reality of operating responsibly in this environment. That said, we believe the economic trends continue in the right direction. Inflation is waning. Interest rates presumably are peaking or have peaked. The jobs market remains strong, and many retailers are suggesting that consumer spending is softening. we continue to look for a return to normal for personal savings rates, which are highly correlated with risk and therefore with pricing on our platform. From a financial perspective, we'd of course prefer to be growing quickly, but this is a time when it's wise to be operating in a conservative mode. In that light, we were EBITDA positive for the second straight quarter. Our contribution margins are still near record highs. And finally, we remain confident that our personal loan models are calibrated And Upstart Powered Credit is performing as expected right now. Not only are we financially secure, we also continue to invest in our AI platform. Last quarter, we launched Model 15.0, the latest version of our core personal loan underwriting model. This new version increased our model accuracy by about 15%. The largest improvement we have seen since we began tracking improvements in 2018 by about a factor of 1.5. Our previous most impactful model launch added personalized timing curves, what we've come to call our loan month model, for a giant accuracy improvement. This version improves the accuracy and precision of these personalized timing curves and also adds personalized macro effects for the first time. Last week, we also launched an upgraded version of the Upstart macro index to account for seasonal patterns and repayment behaviors. Over the last decade, we've measured a distinct seasonal pattern with respect to loan repayments. The time from January to April represents the best seasonal loan performance in our experience, likely due to borrowers receiving extra cash in the form of state and federal tax refunds. Performance then generally degrades marginally each month until the early fall and then flattens or modestly improves through the end of the year. With a seasonally adjusted UMI, will have a more accurate lens into changes in the financial health of consumers, and that should result in less volatility in loan pricing and approvals from month to month. Our auto retail platform saw a huge boost recently as we partnered with a major OEM to implement our software in support of the launch of a new vehicle. Our technology powered their consumer reservation, deposit, and customization system for this amazing new vehicle, and was implemented quickly at more than 99% of all their dealerships in the US. We see this as a harbinger to a future where consumers can choose exactly the car they want online and have it delivered directly to them with none of the friction and inconvenience many associate with the car buying experience. We're partnering with leaders in the industry to unlock this future. Separately, we also expanded our roster of car dealerships that have gone live with upstart lending from 61 to 69. And we've added support for rooftops in Arkansas, Maryland, and Virginia, expanding our reach to 70% of the U.S. population. We also signed agreements with two of the largest national non-prime auto lenders to help fund our auto lending solutions. I'm also excited about the progress with our home equity product. As of today, the Upstart Key Lock is available to homeowners in Colorado, Michigan, Washington, and Utah, and we expect to be live in Alabama, Kentucky, Tennessee, and Washington, D.C. in the coming weeks. We've received encouraging feedback from applicants about how fast and easy the process is, even at this nascent stage. A home equity product helps diversify our business in two critical ways. First, it's a very prime product with annual loss rates expected to be 1% or less. And second, it's a counter-cyclical to a refinance product because it's an effective way to tap equity in a home during a higher rate environment such as we have today. More than 90 percent of HELOCs are offered by banks and credit unions today, so it's a good fit with the Upstart platform and our partners. We'll bring some pricing advantage to the HELOC market over time, but there are two predominant advantages we expect to see sooner. The first is speed and ease of access, because the banks and credit unions that originate most HELOCs today take more than a month on average for the applicant to receive funds. We're aiming for less than five days. Second, our existing platform unlocks customer acquisition advantages that others can't match. Each month, more than 80,000 homeowners apply for a personal loan on Upstart. Some large fraction of them can and will be better served with a home equity product that offers a lower rate. After all, personal loans in HELOCs are just two different ways to solve the same customer need. By integrating our personal loan in HELOC application processes, which we expect to do before year-end, will take a giant step toward becoming customer-centric rather than product-centric. The tradeoffs between price, time, and effort will change over time and will help applicants choose the best product for them. Now let's talk about the funding side. Despite the difficult lending environment, we've seen some great success with credit unions in the last couple of years. We attribute this to the fact that credit unions are extremely focused on delivering the products that their members want, with an intense focus on the quality of experience. They also map well into current and future Upstart products with approximately $29 billion in personal loans, $266 billion in auto loans, and $82 billion in HELOCs funded by credit unions each year. So we're doubling down on credit unions by building features and capabilities that will strengthen our partnerships. In recent weeks and months, we upgraded the Upstart performance console to enhance visibility into originations and loan performance trends improves connectivity to the core systems that power credit unions' financials and operations, delivered features that make it easier for new members to join a credit union, and enhanced our partners' ability to cross-sell other products to existing members. We're also unlocking loan participation, where a loan originated by one credit union can be fractionalized and sold to a network of other credit unions. This significantly improves liquidity in the system and allows us to reach the long tail of small credit unions in an economic and constructive way. On the capital market side, we continue to pursue a large number of committed funding partnerships in order to strengthen the liability of loan funding on Upstart. With banks retrenching, paying more for deposits, and likely facing even more imposing capital requirements, we believe it's important to find alternative sources of funding, even for the primus of loans. We're in discussions about partnerships and structures that can enable at-scale funding across the entire credit spectrum and are excited to innovate in this space. Lastly, we're investing significantly in servicing and collections, a vital part of our business where improvements can go directly to the bottom line. We recently launched a new version of our funded borrower dashboard, which is the experience borrowers see while in the process of repaying a loan. We also recently launched our first mobile application, which is initially focused on loan repayment. Surfacing and collections is clearly an area where AI can lead to better results, and we believe the surface area upon which we can apply our AI expertise is broad. We've brought some incredible new talent to this aspect of our business and are excited about its potential. To wrap up my remarks today, there are plenty of reasons to remain optimistic about Upstart. First, even with our rates at all-time highs, we continue to grow fee revenue and invest in our teams and core AI. Second, our models are learning and improving at an unprecedented pace, creating more separation from traditional approaches to lending. Third, the competition to serve mainstream American consumers with responsible lending products has waned considerably, given the challenges in the markets in recent years. And fourth, we believe there's an inevitable period of normalization when both rates and risk levels return to long-term averages. That will provide upstart with a tailwind over a multi-year period in the future. We aren't waiting around for that period, but we'll certainly be ready when it arrives. Speaking of the future, last week I had the privilege of participating in the US Senate's AI Insight Forum. It was a very productive bipartisan discussion on how to maximize the benefits of AI while mitigating risks. I focused my remarks on the lessons we've learned on our journey to use AI responsibly. to help establish America as a global leader in AI-enabled lending. Hearing about the challenges other industries are facing deploying AI reinforced to me that lending is one of the most compelling examples of how AI can clearly improve the lives of all Americans. I left more excited than ever about the opportunities ahead of us. Thank you, and I'd like now to turn it over to Sanjay, our Chief Financial Officer, to walk through our Q3 2023 financial results and guidance.

Thanks Dave and thanks to all of you for joining us today. We're coming off a quarter of mixed results in Q3 with narrow misses on revenue and margins against success nonetheless in maintaining positive adjusted EBITDA and sequential growth in fee revenue. We continue to operate in a challenging and fluid macro environment. Real consumption continues to hover at levels which we believe to be unsustainable. Incomes continue to lag consumption growth, despite the positive trend in labor participation, as the ongoing decline in government transfers, still receding from stimulus-era highs, offsets any wage gains from the return to work. The residual savings rates in the economy have consequently continued to languish at historically low levels. This consumer reality continues to be reflected in elevated borrower default trends in an upstart macro index. now adjusted for seasonality, which indicates a stable level of loan defaults throughout most of this past year, albeit one which remains remarkably high. Our ability to approve borrowers in this environment has remained the constraint on platform growth for most of the past quarter. On the funding side of our business, banks continue to manage balance sheets conservatively and seek to unwind existing asset positions in secondary markets. The decline in aggregate deposit base which started in mid-2022 has now thankfully eased, but anemic savings rates are so far hindering a rebound in liquidity. Correspondingly, the institutional funding markets remain distracted with the bounty of trading opportunities coming from the banking sector. On the other hand, significant amounts of institutional capital have been recently raised for upcoming deployment into credit and the volume of discussion and negotiation aimed at setting up for 2024 remains encouraging. With these items as context, here are some financial highlights from the third quarter of 2023. Revenue from fees was $147 million in Q3, slightly below our guidance of $150 million, and marginally up from $144 million last quarter. Our underwriting of primer higher income borrowers has become more conservative over this past quarter as their loss rates accelerate and converge with the broader default trends across the borrower spectrum. This has been a headwind for our volumes and fee revenues over this past quarter versus our contemplated guidance. Net interest income was negative $12 million in Q3, owing to continued elevated charge-offs in our R&D portfolio, as well as a one-time change in our charge-off process for loans and bankruptcy. that had the effect of pulling some charges forward into Q3. Taken together, net revenue for Q3 came in at $135 million, slightly below guidance, and representing a 14% contraction year over year. The volume of loan transactions across our platform in Q3 was approximately 114,000 loans, up roughly 5% sequentially, and representing over 70,000 new borrowers, Average loan size of $11,000 was up 9% versus the same period last year and sequentially flat. Our contribution margin, a non-GAAP metric which we define as revenue from fees minus variable costs for borrower acquisition, verification, and servicing as a percentage of revenue from fees came in at 64% in Q3, up 11 percentage points from 54% last year but one percentage point below our guidance for the quarter. We continue to benefit from high levels of loan processing automation and fraud modeling efficacy, achieving another new high in percentage of loans fully automated at 88%. Operating expenses were $178 million in Q3, down 17% year over year, but up 5% sequentially as increasing sales and marketing spend somewhat offset continued savings in engineering and product. Altogether, Q3 GAAP net loss was $40.3 million and adjusted EBITDA was positive $2.3 million, both roughly in line with guidance. Adjusted earnings per share was negative 5 cents based on a diluted weighted average share count of 84.4 million. We ended the quarter with loans on our balance sheet of $776 million before the consolidation of securitized loans. down sequentially from $838 million the prior quarter. Of that balance, loans made for the purposes of R&D, principally auto loans, sat at $447 million of the total. In addition to loans owned directly, we have consolidated an additional $196 million of loans that were sold from our balance sheet into an ABS transaction earlier this quarter, from which we retained a total net equity exposure of $43 million. As described in our prior earnings call, this transaction was somewhat unusual for us and designed to serve as a visible market reset as well as a public vote of confidence in our current degree of model calibration. Our corporate liquidity position at the end of Q3 remained strong with $517 million of unrestricted cash on the balance sheet and approximately $425 million in net loan equity at fair value. We continue to watch for signs of moderating consumption, improved savings rates, and reduced credit defaults in our economy as precursors to a broader normalization of consumer fiscal health. Until we see such signals, our operating assumption is that the macro environment will remain constant. And in such a scenario, our business growth will predominantly come from model upgrades and from improved underwriting accuracy, both of which we consider to be squarely in our set of core technical competencies. and ones in which we have demonstrated a strong historical record that's delivering growth over the years. With this context in mind, for Q4 of 2023, we expect total revenues of approximately $135 million, consisting of revenue from fees of $150 million and net interest income of approximately negative $15 million, contribution margin of approximately 62%, net income of approximately negative $48 million, adjusted net income of approximately negative $14 million, adjusted EBITDA of approximately zero, and a diluted weighted average share count of approximately 85.6 million shares. That is all for our prepared remarks this afternoon. A shout out to all of the upstart teams who have kept their heads down and their resolve strong over the course of a past year spent in an excruciating external environment. Our business today is a much stronger one in a number of very tangible ways than it was before the onset of this post-stimulus hangover. And when the winds eventually settle, your work will be manifest and brightly shine. With that, Dave and I are happy to open up the call to any questions. Operator?

If you would like to ask a question, please signal by pressing star 1 on your telephone keypad. If you are using a speakerphone, please make sure your mute function is turned off to allow the signal to reach our equipment. Again, press star 1 to ask a question. We will take our first question from Simon Clinch with Redburn Atlantic. Please go ahead.

Hi, everyone. Thanks for taking my question. I was wondering if I could start with your comments around private credit and the encouraging pipeline for 2024 in terms of long-term funding commitments. Could you just expand on that a little bit more in terms of I guess the scale of long-term funding that you would hope to get and how we should think about that in terms of visibility into the potential growth beyond this year.

Yeah, hey, Salman. Sanjay here. Thanks for your question. Yeah, I guess the comment was maybe an allusion to the fact that in the ongoing discussions that we're having with a number of counterparties, obviously in the vein of securing longer-term committed style capital. I think the volume and the depth of those discussions has been on a good trend, and it feels like there's a lot of counterparties that are exploring this asset class that are newer to the asset class that have freshly raised capital. It's probably related to what we all read about in the press at the private credit space is becoming very big and very interested. What that means with respect to 2024 and future plans or maybe sort of visibility we have into the capital base, nothing concrete yet, unfortunately. I think it's just an encouraging trend for us and it sort of reinforces our view of the direction we're taking.

Okay, thanks for that. And just as a follow-up, I'm just turning to the very high level of automation, 88%, coupled with the sort of very low conversion rates and the high contribution margins. I was just wondering, when we get to the stage where volumes start to recover, just for modeling purposes, should we effectively assume that as conversion rates rise, that level of automation is not sustainable? And so we would expect, I guess, the incremental costs to rise significantly and push that contribution margin down. Is that the kind of math that we should be thinking about?

Hi, Simon. This is Dave. I don't think so. There is an obvious relationship between the rest of the funnel and the sort of automation level. So it could go up and down because of just changes in who's applying and things like fraud rings that could drive the automation level down when our systems are kind of blocking fraud rings, things like that. So there are some reasons it can not always go up into the right and in fact has over time. Having said that, I don't think it sort of has that relationship that you are suggesting.

Okay, that's really helpful. Thank you very much, guys.

We will take our next question from David Scharf with JMP Securities. Please go ahead.

Good afternoon, and thanks for taking my questions as well. Maybe Following up the last round of questions, kind of similar vein, focusing on not so much funding, Sanjay, but I guess balance sheet retention. We've been obviously seeing a lot of non-prime lenders announce more kind of private capital partnerships lately, forward flow arrangements. The ABS market spreads are still wide, but they are tightening. Given all of that context, You know, irrespective of the demand environment, is there a planned trajectory in mind that investors should think about in terms of, you know, effectively winding down, you know, the balance sheet exposure? I mean, we realize there are always going to be so-called testing loans, but when the residual from the securitizations factored in, it looked like it was kind of flattish, you know, quarter to quarter and kind of wondering when you would imagine the loans held for investment or for sale to be substantially lower.

David, thank you for the question. I guess at a headline level, nothing explicit to really share with respect to how we would view the balance sheet evolving over the course of the next year. As we said in our remarks, the explicit constraint on growth right now is on the demand, is on the borrower side and the approvability of the borrowers. And as that relaxes, presumably over the course of time and as our platform rescales, obviously the goal is to use as much third-party capital as is available in that rescaling. And that will be our goal. But I don't think we have any explicit guidance or predictions on how that will evolve over the coming quarters.

No, understood. And maybe just a follow-up on credit. You know, it looks like the fair value adjustment, it was about another negative kind of $40 million this quarter. I would imagine that that's typically made up of current period losses as well as any mark-to-market adjustment. Can you kind of give us the breakdown of those? And also, as we think about sort of your guidance in Q4, for net interest income. I mean, embedded in that is another kind of fair value adjustment and loss figure. You know, is it possible to give us kind of the charge-off figures for the retained balances that was embedded in the Q3 number as well as the Q4 guide?

Yeah, I mean, I think I will just say that, well, first of all, the components of fair value are the actual charge-offs being incurred by the loans in the balance sheet. any unrealized fair value adjustments we're taking to those loans when we mark them up and down as the assumptions evaluation change, and then any realized gains or losses we take on sales or securitizations. And I'll just say that by far the biggest component in Q3 are the charge-offs themselves. And a lot of that, as we said, we've got about half a billion dollars of auto loans on our book that we're made sort of in the vein of R&D almost a year and a half ago now, back when we were in a very different environment. And that half of our loan book is unsurprisingly taken on excess defaults right now. And I think all of those facts will remain true in Q4 as well. The largest component of fair value in general is the actual charge-offs being incurred by the loan book. Got it.

Thanks, Sanjay. Thank you, Dave.

We will take our next question from Lance with VTIG. Please go ahead.

Thanks for taking my question, guys. You know, a lot of the pressure on the top line has been, you know, the rates coming in above 36%. As we come into an environment where, you know, it's more likely than not that we see, you you know, combined with UMI, it looks like your UMI is starting to kind of level off a little bit. How should we think about the interplay kind of between the two and how that impacts, you know, your conversion rate and how, you know, where the APRs are coming in below 36%? So, you know, is it just a linear like-for-like basis, whereas the rates start to come down and then the conversion rate starts to go up or, you know, what are the contribution between that and the UMI and where rates are?

Hey Lance, this is Dave. Great question. Generally speaking, both UMI and kind of underlying rates have both been, as we've said, moving up pretty much constantly for the last year and a half or even longer. And each time that happens to either of them, they are additive, in effect, to the price of a loan in the marketplace. And as you said, because we have kind of a sharp line at 36% where we don't offer loans above that, that means people are being knocked out of the approval box and declined increasingly as those two go up. I would say, generally speaking, the UMI effect is larger than the interest effect. the interest rate effect. But they both push in the same direction. They both work to make prices higher. So to your point, not to project where things will go, but assuming rates do come down and assuming over time UMI comes down, those will both certainly become tailwinds to us and essentially lower rates and improve approvability and thus the funnel metrics that's driven by those. So it's very fair to say as those things go down, it would be what we would consider a normalization of our funnel and our overall pricing and everything else. And we certainly look forward to that.

Got it. Appreciate it. And then, you know, pivoting a little bit to the HELOC product, I'd be interested to kind of hear the, you know, first reactions from essentially, you know, what is the approval to funding time? What are the margins looking like for the product? How's, you know, consumer feedback been? Any color that you can give there would be great.

I would say generally, again, this is Dave, I would say it's probably too early to quote conversion rates, things of that nature. It's still in its early stage. We are still kind of This is kind of the part of the product where we're very quickly making funnel improvements that can be leaps and bounds, but they're off of a very modest start. But I will say generally, the world expects key locks to take a long time and to be a very difficult process. That's just the backdrop. of us coming into the market. So it's not, we can look really good and people can be very pleasantly surprised. And we're still kind of saying, wow, we're just a shadow of where we're going to be in six months or a year with that product. And that's kind of where we are. I think we have a process that's probably still better than you can get elsewhere in the market. Maybe not everywhere, but what most people are experiencing when they go out looking for a HELOC. But at the same time, we're not even a shadow of the way where we're going to be down the road. So But, you know, it's working. It's good. We're launching more states. And, you know, each week or two of development is typically a big step forward for us because there's just so many obvious things that we're improving as we get started.

Got it. Thanks so much. Thank you.

We will take our next question from Peter Christensen with Citi. Please go ahead.

Thank you. Good evening, and thanks for taking my question. I was just curious, you know, with loan size being kind of flattish sequentially, just curious if we should think of average loan size as a function of available capital or capital supply or, I guess, more or less an issue of what's going on in consumer credit right now. And I guess as a follow-up, just curious if we should think of loan size impacting the contribution profit margin.

Thank you. Hey, Pete. This is Sanjay. Thanks for your question.

So I think your question was about loan size, not loan volume, correct? Correct, correct, yes. Okay, yeah. Loan size, let's see. I mean, there's a couple of different impacts. I would say the biggest one is borrower mix. so you know all else equal the sort of more low risk borrowers you could think of maybe in a traditional sense the primer borrowers that you have will tend to be able to to be approved for larger loan sizes and riskier borrowers in contrast lower loan sizes so if you have for example a macro event such as we've had where there's a lot of people being nudged outside of the approval box at the riskier side and then you know consequently have a A primer borrower mix than before, you would expect to have larger loan size and vice versa. So borrower mix is one. Overall sort of risk and approvability is another. All else equal, as loss rates get higher in the macro, UMI gets higher, we would generally sort of on the margin be approving folks for smaller loans than in a very constructive environment. So I would say it's really down to mix. I wouldn't necessarily view loan size as a sort of inherent or fundamental metric for our business in any sort of significant way. Although it is true, just answering your last question, it is true that all else equal, a larger loan size will lead to a healthier contribution margin.

That's helpful. Thank you. I'm just sorry, just as a quick follow-up, I guess, how should we think about what borrowers are requesting versus what they're being approved for, is there a meaningful difference there in this environment?

Yeah, I would say that in relative terms, they've gone very much in the opposite directions. I think the request for credit, the underlying fundamental demand for credit is as high as we've seen it in quite a while, and there's a lot of underlying demand from borrowers. Conversely, our ability to approve them is as limited as it's been in our company's history as a reflection of the UMI. And so you sort of have them at cross-purposes right now.

Thank you, Sanjay. It's helpful. Thank you, Pete.

We will take our next question from James Fawcett with Morgan Stanley. Please go ahead.

Great, thanks. I want to follow up on that last question. You know, clearly you guys had to tighten things up and, you know, given the low savings rates is that demand for credit seems understandable. But what are you seeing in terms of actual conversion, that is offers for loans that are actually converted into loans and what that origination looks like? Are you seeing any movement in that metric?

Agent, so you're asking about the conversion rates? Yeah.

Yeah, I think our metric for this quarter was around 8%, 8.5%. So meaning of all applicants who fill in an application and submit, about 8.5% of them become funded loans. I think at our peak, that number was closer to 24%.

Got it. And can I just ask for a clarification? What about, that's of all applicants, what about of approved loans that actually ultimately convert? Any sense what that looks like?

From loans that are approved and offered to funded loans, I believe is on the order of about a third.

About a third. Okay. That's helpful. And then I wanted to go back to the committed capital partners, just any additional color that you can provide in terms of the things that they are looking for, whether it be specific metrics or catalysts to get them moving forward. And do you feel like for them, maybe it's the same thing as we're talking about for borrowers that we kind of would like to, people would like to see underlying rates come down and then for your own metrics, UMI come down. Just wondering if your potential committed capital partners are kind of waiting for a similar relief across the market or if there are other issues that they're watching.

Sure, yeah, thanks, James. Let's see. Well, certainly for counterparties that are, so we say newer to unsecured consumer credit, there's a lot of investigation about the asset class itself. With respect to Upstart, I don't think it's as much about whether rates are going to move or UMI is going to move. It's more about developing a confidence in their diligence that we are going to hit the targets that we claim, meaning when we predict loss estimates, they are accurate. And so they care a lot about our accuracy rather than the absolute direction of rates in the economy right now. And then there's clearly a macro component as well, All of these counterparties have credit committees and macro committees that take point of view on broad asset allocation and what the right timing is for these investments. And so many of them are actively engaged with us in asking what we're seeing in our data with respect to macro trends and loss rates in the different segments of the borrower base. So that's sort of, I think, at the highest level, how they would think about the framework of investing with us. And then with respect to specific deals, As we've said before, you could sort of broadly class counterparties into some that are interested in maybe earning a premium on the return right now, and they're trying to understand where that can come from, certainly in a high-rate environment like today's, if they can lock something in. It can look very good down the road a year from now if the economy changes. And then there are others who are less interested in return premiums. They're interested in predictability. And that's where, as we've said in the past, in some instance, because we have a very precise understanding of our model calibration, we can sort of co-invest with them, put some skin in the game, give them some level of comfort on that dimension.

That's great. Hey, thanks for all the details.

We will take our next question from Ramzi Elisal with Barclays. Please go ahead.

Hi, thank you. This is John Coffey on for Ramzi. My first question is on slide 21, and you may have already answered this and I missed it, but when I look at the cadence of the auto-secured loans, it looks like that halved from Q2 of, well, last quarter to this quarter. I was wondering if you could just talk a little bit about the drivers there, what caused that?

Yeah, hey, John.

Thank you for the question. Yeah, auto lending, I would say, is currently subject to all of the same sort of dynamics and forces as our core business, meaning rates themselves have risen and default trends look very similar to what they look like in the unsecured world, maybe with some slight timing differences. But in terms of the sort of relative level of increase in default rates, they very much mirror what we've seen in our core business. And Just as our core business has had to retrench, as we've had to recalibrate our models, as a lot of people, as Dave described, have fallen above the 36% line and out of the approval box, a lot of those dynamics are very similar in auto as well. With the difference that auto being a more nascent product for us, we've been calibrating a newer model in real time in that, whereas with personal lending, I think we've got a very sophisticated model that we're just kind of make sure that it stays calibrated to the macro environment. So the short answer is a lot of the things that have happened to our core business have also conspired to restrain auto lending volumes a little bit as well.

All right, that's very helpful. And I just have one follow-up question. Last quarter when you reported in August, you already had a month behind you, that being July. So given that you had your Q3 guide, what was the biggest surprise given that you still had two months to go What was the factor that, you know, I know it wasn't a big miss of guidance, but there was a miss. What was the thing that you thought essentially was the thing you were wrong on or was the biggest surprise there versus what you originally thought of?

Yeah, thanks. Great question. I'll highlight two things that we sort of alluded to in our remarks. One of them is just with respect to, you know, the transaction volume, the transaction revenue. One of our model launches this quarter allowed us to essentially measure the sort of macro impact along the lines of what we do for UMI, but at a borrower and a segment level, which is a greater level of granularity in terms of macro sensitivity in the borrower base. And what that model quickly understood is that what's happening in very recent months is that what you might think of as the primer or the more affluent borrowers are sort of starting to accelerate in their default trends, whereas the less affluent borrowers who were earlier affected last year in the environment that we've been through are now much more stable. And it led us to be, I would say, a little bit more conservative in our approval of more affluent, sort of primer borrowers. And so that was one impact that I think versus what we had contemplated in our guidance was a bit of a delta. And the second one was just sort of a mechanical thing that showed up in net interest income, but it had to do with essentially a process change around the timing with which we were charging off bankruptcy loans. You know, we went from the resolution of the bankruptcy to essentially the announcement of the bankruptcy, and it had the effect of pulling some charges forward from future quarters up into Q3, and that's something that we weren't expecting when we announced guidance.

Great. Thank you very much. Thank you.

We will take our next question from Rob Wildhack with Autonomous Research. Please go ahead.

Hi, guys. I wanted to ask about the co-investment summary on slide 20. Last quarter, the $40 million co-investment was marked up $11 million, so up $30, and now the $66 million is up, but it's only up $7 million or like 11%. What were the drivers behind that markdown versus the second quarter?

Yeah, thanks, Rob. It's a great question. Yeah, what you're seeing there is maybe a bit of retrenching in how we're valuing that. So it's representative of what you might think of some headwinds over the past 90 days in how all of that system is performing. Look, one of the things that's on our minds that I think is relevant here is we're not – We're not necessarily adjusting this valuation for current seasonality. And it occurs to us we're in the worst seasonal part of credit right now. So credit, it is at its best from a seasonal standpoint in the sort of February to April timeframe. And it's sort of at its worst in the October, November timeframe. And so I think some of what's happening here is that like the seasonal headwinds, the credit performance are making this asset maybe look a little bit worse than it will in six months. And we haven't necessarily gone to the level of sophistication of adjusting for seasonality. So I think that's part of the story there. But some of it also may be durable. And therefore, what it would indicate is that stuff is performing marginally worse than it looked 90 days ago.

Okay, thanks. And then on the committed funding partners, those initial agreements, the ones in place now, I think they were described as $2 billion in funding for the next 12 months. So that was roughly spring 23 to spring 24. So of the existing agreements, could you remind us what's locked in for beyond spring 24? And for anything that's not, what's the process or arrangement for extending or renewing those agreements?

Sure. Yeah, thanks. I guess I would say we haven't necessarily given specific numbers beyond the spring. I do think there were some announcements out there in the public which alluded to some longer timeframes for some of those agreements. And so some of those we expect to continue. I think with the other ones, as we get towards the sort of one-year mark, and depending on how happy the counterparties are and how happy we are and how well everything is performing, you might imagine that there will be discussions around renewal. So I think we'll be undertaking those probably in the new year.

Okay, thanks.

We will take our next question from John Hecht with Jefferies. Please go ahead.

Afternoon, guys. Thanks very much. So I think you did about $1.2 billion of loan originations, and it looks like you cleared nearly a quarter of that through your on-balance sheet securitization. So for the remaining 75-ish percent, I'm wondering, could you just let us know what was the characteristic? Were they kind of bank buyers that have been around for a while? Were they more credit funds? Were they more tied to the counterparties in the forward flow agreements that you just were talking about? How do we think about the mix of that disposition?

Sure, yeah, thanks for the question. I mean, maybe just one clarification about the securitization. It was executed in Q3, but a lot of the loans that were sold into the securitization were originated in Q2. So it wasn't necessarily a component of Q3 origination. I think at a higher level, yeah, we sort of have these three channels, if you will, of funding. One is the bank and credit union channel where they're typically doing the origination themselves as a lender. There is our traditional forward flow channel, and now there's a newer channel of what you might think of as committed capital. I think in rough terms, The forward flow and the bank channel numbers are somewhat comparable, and the channel by which we are providing loans into committed partnerships where we co-invest is slightly larger than the other two, but there's a pretty good balance across the three.

Okay. Thanks very much. No hands.

We will take our next question from Giuliano Bologna with CompassPoint. Please go ahead.

Just going back to the committed capital co-investment disclosure that you have, I'm curious, it may have come up early in the call, roughly speaking, what the balances of loans that you have outstanding that you have that co-investment structure attached to at this point?

Yes, thanks, Giuliano. I think that's disclosed in the few – I don't have the exact number on my fingertips right now. Last quarter, if you recall, the co-investment represented approximately 5% of the total, and I think that's probably still a relatively consistent ratio.

That's very helpful. And this might be a similar brief question to have related to the revenue from fees. Do you have the numbers for the platform referral fees versus the servicing income for the quarter and how that breaks down for the period? And the genesis of that is to kind of back into what the take rate was on originations.

Yes, Julian. I think those numbers are also disclosed in the queue in one of the notes. And I don't have them Off my fingertips again, but I think you will see something like transaction revenue is maybe 3 to 4x the servicing revenue. Transaction revenue is up about 7%. Servicing revenue is declining slightly as the outstanding balance in the platform declines. But if you want the specific numbers, they'll be in one of the notes in the queue. We can point you to it offline if that's helpful.

That's helpful. I appreciate it. And then thinking about the cadence of where you are on the funding side, I realize this is a question that's been asked a few different ways at this point. But when you think about where you are from an operating perspective, you obviously need more volume to get incremental margin and cover your fixed costs. I'm curious how you're thinking about the cadence of how funding could come back or how you're preparing to manage the business for the foreseeable future. So if this environment persists for a number of quarters beyond 4Q.

Sure, yeah. Let's see. Well, as I said in the remarks, I think there's a lot of good conversations happening on the funding side. The platform constraint today on growth is on the borrower side and our inability to approve contracts. But in anticipation of that eventually changing, and as we said to one of the earlier questions, that could change because rates drop. It could change because sort of default trends normalize and UMI drops. In anticipation of that, we definitely want to have a few more agreements teed up and some partnerships ready to go at that time. But it isn't the gating item on platform growth currently.

I appreciate the answers to questions, and I will jump back in the queue.

Thank you.

We will take our next question from Reggie Smith with JP Morgan. Please go ahead.

Hey, good evening. Thanks for taking the question. Most of mine have been answered, but I was curious. You were talking earlier, I think a previous analyst had asked about and, Sanjay, I think you suggested that roughly a third of the loans that are approved are ultimately funded. Can you tell us kind of how that has trended and it changed over time? And then the second part of the question, do you ever approve borrowers for less than they request? And do people tend to take those loans if you do? No.

Sure. Thank you, Reggie. Let's see. The answer to the first question, well, the first question was what has happened to what we call acceptance rates, which is the instance where a loan is approved and an offer is made, and what percentage of those are accepted by the borrower, and those are currently at the 30% level. They've definitely trended down. I think they probably used to be in the 60s a couple of years ago before the environment sort of became very Very challenging. And the intuition is obvious. As rates have gotten very high, even though borrowers are getting approved under 36%, in some instances, the rates that they're being approved for are quite high. And their propensity to accept those rates is a lot lower than it was before. And the second question had to do with... Sorry, remind me of your second question, Reggie.

Yes, I was curious if somebody comes in with a loan request and it's totally out of whack with their income or debt levels, do you ever approve them for less? Or can you do things like that to improve acceptance rates? Or is there a side below it just doesn't make economic sense to do a loan? Any color there would be helpful.

That's a great question, RJ, and we absolutely do that. There are many applicants who are asking for loan levels that we are not able to approve them for, but what we can do is say, well, you know, we can approve you for a lower amount, and very often those are accepted instead. And so that is a component of our acceptance rates, and it is a business practice we have.

And do you find that those loans tend to perform as well as you would expect, or do they – What insights have you gleaned from those types of transactions?

They perform as expected on a risk-adjusted basis, meaning because we are approving them for a lower amount, that adjustment to the risk is commensurate with the risk that we perceive. At the end of the day, for investors, they perform as expected just as expected. full, you know, full offers do, um, compared to what we call, we call these counter offers.

That makes sense. Okay. And the last question for me, um, marketing, I saw it was up sequentially. Um, it had been, you know, obviously down a lot from the previous year, anything to call out there on the, uh, on the marketing side of, you know, competition is advertising. Like what, what's, uh, what's driven that?

Yeah. Hey Reggie. Um, I think the thing to call out is that we're at the point now where as we are, as we said, a borrower-constrained platform, and as we regrow into prior volumes, all of that growth will be paid growth. So obviously, as we've contracted over the past 18 months, we've prioritized organic and unpaid volume, repeat volume, if you will, We've got a great benefit from that, but I think the incidence of repeat loans will start to subside just given the recent volumes on the platform. And as we sort of regrow into higher volumes, such as where we've been previously, that you'll see a higher incidence of paid growth versus unpaid growth. And that will have the effect of bringing up the overall average acquisition cost.

That makes sense. Okay. Thank you. Thank you, Rachel.

There are no further questions at this time. Mr. Girard, I will turn the conference back to you for any additional or closing remarks.

All righty. Well, thanks all for joining us today. We continue to make strides in building the first and best AI lending platform in the market. There is not a company better positioned than Upstart to lead this transformation of the financial services industry. So thank you, and we'll see you next time.

This concludes today's call. Thank you for your participation. You may now disconnect.