Upstart Holdings, Inc. (NASDAQ:UPST) Q4 2021 Earnings Conference Call February 15, 2022 4:30 PM ET
Company Participants
Jason Schmidt – Vice President of Investor Relations
Dave Girouard – Chief Executive Officer
Sanjay Datta – Chief Financial Officer
Conference Call Participants
Simon Clinch – Atlantic Equities
Pete Christiansen – Citi
Arvind Ramnani – Piper Sandler
Ramsey El-Assal – Barclays
Andrew Boone – JMP Securities
Mike Ng – Goldman Sachs
James Faucette – Morgan Stanley
John Hecht – Jefferies
Nat Schindler – Bank of America
Operator
Good day, and welcome to the Upstart Q4 FY 2021 Earnings Call. Today’s conference is being recorded.
At this time, I would like to turn the conference over to Jason Schmidt, VP of Investor Relations. Please, go ahead, sir.
Jason Schmidt
Good afternoon, and thank you for joining us on today’s conference call to discuss Upstart’s fourth quarter and full year 2021 financial results. With us today are Dave Girouard, 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 fourth quarter and full year 2021 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 first quarter and full year 2022 related to our business and our plans to expand our platform in the future. These statements are based on our current expectations and information available to us today and are subject to a variety of risks, uncertainties 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, as we assume no obligation to update any forward-looking statements as a 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 non-GAAP financial measures and are reconciled to our GAAP results, which can be found in the earnings release and supplemental tables. To ensure that we 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 Jefferies Payments & FinTech Summit on March 1, JMP’s Securities Technology Conference on March 8 and Morgan Stanley’s Technology, Media and Telco Conference on March 9.
Now, I’d like to turn it over to Dave Girouard, CEO of Upstart.
Dave Girouard
Good afternoon, everyone. Thank you for joining us on our earnings call, covering our fourth quarter and full year 2021 results. I’m Dave Girouard, Co-Founder and CEO of Upstart.
We’re a month-and-a-half into the New Year, and I’m grateful finally to have the opportunity to reconnect with the investor community, some quiet periods just be longer than others. We have seen some epic progress at Upstart in the past few months, and I’m excited to share what we’ve been up to, the results we’ve been seeing and how we’re thinking about 2022 and beyond.
Let me state upfront that we’re in a multi-decade mission to put affordable credit within reach of every American. The price of credit is the price of opportunity and the price of mobility. And we want to ensure that opportunity and mobility are available to all Americans, particularly for those whom the financial system has failed in the past.
But like you, we’ve been eyes wide open, watching all that’s happening in the world in the last few months: the rise and fall of the Omicron variant, the clear signs of inflation and the Fed’s plan to counter it, and of course, the market rotation out of high-growth technology. But through all of it, our business continues to get stronger and my confidence in Upstart’s future has never been greater.
As the rare public technology company with triple-digit growth and profits, we’re confident that an economy and market in transition plays to our strength. I’d like to start by reflecting on 2021, which was a remarkable years for Upstart. We grew revenue from $233 million in 2020 to $849 million in 2021, while generating net income of $137 million. And with the fourth quarter surge, we’re now at more than $1 billion in revenue on an annualized basis. 2021 will be remembered as the year AI levying came to the forefront, kicking off the most impactful transformation of credit in decades.
To gain some perspective on what Upstart achieved in 2021, we looked for another company in the public markets with our combination of scale, growth and profits, but we were unable to find one. Our profits are neither marginal nor ephemeral. We generated more cash in 2021 than we burned in our entire eight-plus years as a private company. Profits matter for a reason. They allowed us to invest significantly in our future by more than doubling our headcount in product, engineering and machine learning in 2021. This unusual combination of growth and profits in a heavily competed industry is evidence of a distinct competitive advantage and clear operating leverage. It also suggests that you’re witnessing the creation of an industry-defining category, artificial intelligence lending and the emergence of the category leader, Upstart.
In addition to reaching $1 billion in annualized revenue and record profits, Q4 was special for other reasons. It was the first quarter with more than $4 billion in loan transactions on our platform, a record not just for Upstart, but potentially for the entire personal lending industry. Our bank and credit union partners originated almost 500,000 loans in the quarter. We also now have 42 banks and credit unions as well as more than 150 institutional investors funding loans on the Upstart platform, providing deep and diverse sources of liquidity to keep the engine humming and the AI models learning.
I’m also pleased to report that we now have seven lenders on the Upstart platform with no minimum FICO score required. But perhaps the most important achievement of the last quarter of 2021 was the incredible work done by our auto team. Through a relentless and determined cross-functional effort, this team put the last essential pieces in place necessary to begin scaling auto lending on the Upstart platform. I’ll come back to this topic in a moment.
I’d like to note that the Upstart team accomplished all of this during the second year of a global pandemic, while operating in an almost entirely remote and distributed fashion. We moved to a digital-first strategy while simultaneously implementing what we call a vertical team working structure. This new approach is unlocking Upstart’s ability to execute quickly and efficiently as a multi-product company. What’s really exciting is that we’re finding talent across the entire US. In fact, in Q4, more than two-thirds of our hires were made outside of our California and Ohio footprint. I cannot help, but express my amazement for all the Upstart team accomplished in 2021, particularly given the circumstances under, which they accomplished it. A sincere thank you to the entire Upstart team and also to the family and friends that support them.
Now, I would like to move on to 2022 and how we’re thinking about the year ahead. We find ourselves today in the strongest position Upstart has experienced to date and it’s our mission in 2022 to build on the many successes of the last year. At the beginning of each year, I’d like to clarify in my head and with the team, the handful of objectives Upstart needs to achieve to make the year an unqualified success. Right at the top of the list for 2022 was achieving meaningful scale with auto lending on our platform. We believe in our core that AI lending isn’t a one-category phenomenon, but will eventually transform virtually all flavors of credit. I’m happy to tell you that just 1.5 months into the new year, we’ve accomplished this goal. In fact, our auto refi funnel performance is now comparable to where our personal loan funnel was in 2019 on a channel-adjusted basis.
Based on this progress, we now expect $1.5 billion in auto loan transactions on our platform in 2022. Just as importantly, we now have the confidence to invest the resources necessary to unleash the model and technology improvements in auto lending that made Upstart the category leader in personal lending.
As I referenced earlier, this great leap forward was the product of an intense push by our auto team toward the end of Q4. There are many pieces and parts we needed to get right to enable a minimally efficient funnel and the team worked night and day right up through the holidays to make it happen. It’s worth stating that scaling the auto business from here is no simple task, more funnel and model improvements will be necessary, and distribution channels in auto refi aren’t nearly as well established as they are in personal lending.
But even though channel development will require significant time and effort, the good news is that we’re confident we’re in a class by ourselves. Upstart has a unique and proprietary auto refinance product with far less competition than we’ve had in personal lending. In truth, if you don’t have a certain level of funnel efficiency in auto refi, you really don’t have a product. Today, we’re confident that automotive lending is a category we can grow into for years to come.
We also continued to make rapid progress in the new product categories that I mentioned in our last earnings call: small dollar lending, small business lending and mortgage. In each case, we’ve established a core team and are making real progress toward entering the market.
In the case of small dollar and small business lending, we expect to have these products in market during 2022. In the case of mortgage lending, we hope to be in market in 2023. In each case, we anticipate a year or so of development, a year of feeding and testing and then a year to begin scaling. A home run success for Upstart would amount to a new product in-market and ready to scale in each of the next two or three years. Of course, it’s very hard to time innovation, much less market adoption, but this is the pace we’re aiming for. Overall, the categories we’re in today or expect to enter represent an addressable market of more than $6 trillion in annual originations.
Upstart is now about the size that Google was when I joined that company in early 2004. So I’ve seen this movie before and hope to use what I learned there to build Upstart into the most impactful fintech in the world. I have some specific personal goals for Upstart in 2022.
First, to transition into a multiproduct distributed company that can operate in parallel instead of in serial. Second, to break new ground in terms of quality of execution at the $1 billion-plus scale, with leaders such as Google, Amazon and Apple as my North Star. And third, to move aggressively to unlock Upstart’s addressable market while simultaneously upgrading our ability to pursue it. These challenges will keep our leadership team busy in 2022 and well beyond. Upstart is a unique company, both in terms of our technology and our business model. We don’t exactly look like anybody else. And for this reason, we’re often misunderstood.
So in closing, I’d like to share a few thoughts about Upstart that have struck me in the last few months as useful ways to understand who we are and what we’re building. First, Upstart is both a consumer Internet brand as well as a cloud software provider, delivering a deeply proprietary and technical product to our bank and credit union partners. This combination is entirely unique and is central to our competitive position today and in the future.
Were it not for the AI models at the core of Upstart, we would have little unique value to offer our bank partners. And were it not for our consumer presence and scale, we would not control our destiny, and our AI models would not be learning as quickly as they are.
This combination means we can dramatically strengthen the competitive position of banks who partner with us, while simultaneously helping consumers find the very best credit product available for them.
Second, choosing not to become a bank was the right decision for Upstart and it’s central to our world view. A very successful bank will serve a particular slice of America incredibly well, with a well-constructed portfolio of products, a trusted brand, durable relationships and a predictable business model.
We believe we can help forward-thinking banks succeed in their mission with better technology. We think of ourselves as a consumer Internet brand focused on personal finance. Unlike a bank, an Internet brand can seek to serve all Americans and eventually everyone in the world. This time with an incredible diversity of offerings from hundreds, if not thousands of partners, each of whom will benefit from leveraging upstart AI.
So in short, our goal is to become a technology partner to all the world’s great financial institutions, and through those partnerships, to enable the broadest array of financial products at the best price and with the best experience to everybody.
Finally, lending is a cyclical industry and always will be. Though Upstart is not a lender, we are a technology provider to this industry. So we expect our growth in transaction volumes to vary considerably from quarter-to-quarter. But at the same time, we represent a secular change that we believe is both inevitable and durable.
Our core thesis is that over a period of years, AI lending will rapidly gain market share over legacy approaches to credit and Upstart is in the pole position to benefit from that. In fact, economic volatility, such as we’ve seen in the last two years only serves to demonstrate the value of a modern AI-enabled approach to credit origination.
Thank you. And now I would like to turn it over to Sanjay, our Chief Financial Officer, to walk through our Q4 and full year 2021 financial results and guidance. Sanjay?
Sanjay Datta
Thank you, Dave, and thanks to everyone for joining today. I hope everyone had an accretive Valentine’s Day. Quickly running through our results, starting at the top of the P&L. Net revenues in Q4 came in at $305 million, up 252% year-over-year. Revenue from fees constituted $287 million of that amount, representing 94% of overall revenue and up 37% sequentially from last quarter.
The majority of our sequential growth came from additional top-of-funnel rate requests, which grew at 29% Q-on-Q. The balance of growth was driven by higher funnel conversion rates, which were up 140 basis points or 6% relatively Q-on-Q despite the significant expansion in final traffic.
The volume of loan transactions across our platform in Q4 was approximately 495,000 loans, up 301% year-over-year and representing over 400,000 new borrowers. This increase in volume is distinguished by participation across a widening swath of borrower segments.
At one end of the spectrum, attracting growing numbers of applicants meeting the traditional definition of prime, where we have historically not competed. At the other end, bringing more hidden prime borrowers into the lendable universe under the national bank rate cap, as partner banks increasingly eliminate hard eligibility criteria, leading our models free to perform their magic.
Accordingly, we scaled our marketing program spend in Q4 by 19% Q-on-Q, while simultaneously improving loan unit economics.
Our contribution margin, a non-GAAP metric, which we define as revenue from fees minus variable costs for borrower acquisition, verification and servicing, consequently improved through this expansion, rising from 46% in Q3 to 52% in Q4. Our improved contribution margins versus Q3 reflect refinements we’ve made to our digital and direct mail targeting models, take rate optimizations, improvements to life cycle marketing, which drove a higher proportion of low-cost loans, and shrinking operations unit costs as our automation rate recovered to 70%.
Operating expenses in Q4 were $244 million, growing 22% sequentially over the prior quarter. Spend on engineering and product development once again led the way as our priority investment area growing 25% sequentially despite slower hiring than desired. Growth in general and administrative spend registered at 22% sequentially as operating leverage continues to improve. Expenses in sales and marketing and customer operations, as always, grew in proportion to revenue, albeit in Q4 at a rate of increasing economy of scale.
Taken together, these components resulted in Q4 GAAP net income of $58.9 million, up 102% Q-on-Q and an adjusted EBITDA of $91 million, up 54% Q-on-Q. Adjusted earnings per share for Q4 was $0.89 based on a diluted weighted average share count of 98.8 million. On the full year scoreboard, we tallied $849 million in net revenue in 2021, which was a 264% growth clip over 2020, a contribution margin of 50%, up 400 basis points from the prior year, and adjusted EBITDA of $232 million, representing a 27% adjusted EBITDA margin versus 13% a year earlier.
We ended the year with $1.2 billion in restricted and unrestricted cash, up from $311 million ending the prior year. Of the net increase, approximately $855 million was raised in the capital markets, $266 million was cash earned from operations net of loan transactions, and $170 million was reinvested back into our balance sheet in the form of loans made in support of new R&D programs. Consequently, our balance of loans, notes and residuals at the end of the year was $261 million, up from $140 million in Q3 and reflecting the accelerated pace of R&D.
Most notably, auto lending has been funded since inception entirely from our own balance sheet. This is, as always, a temporary incubation period until we reach the point where the loans can be directed to our bank partners and institutional investors at reasonable scale, which we anticipate will begin to happen next quarter.
As we stare down the year ahead of us, we are cognizant of the fluidity in the macro environment. Over the past quarter, we have started to observe what we had long predicted, namely a reversal on the trajectory of default rates. Defaults have been at an unnaturally suppressed levels for more than a year. As we have consistently messaged, the fading of stimulus should presumably lead to normalization in default rates. And as of November, we believe we are seeing that normalization.
As we, along with our bank partners, investors, have been anticipating this shift and as the loans on our platform have been priced accordingly, we are not expecting any meaningful adverse impact from rising defaults on our volumes or economics. Note that this recent upturn in loan default is not to be confused with the longer-term secular vintage-over-vintage increase in absolute default profile on our platform, which has been alluded to in some public forums. This phenomenon is almost purely a function of change in borrower mix, as our models expand the frontiers of approvability and pull more applicants into the lendable universe over time.
Viewed in this context, rising absolute default rates that are correctly predicted and priced are not above but in fact, a feature of our platform and a trend we expect to see continue as we successfully progress against our core corporate mission of expanding access to credit.
A second macro topic de jure relates to rising interest rates and inflation. Our view is that a moderate increase in rates will not have a meaningful impact on our business for two reasons. An increase in the Fed rate does not translate directly into higher cost of funding for our bank partners. And to the extent it does, the floating rates on the credit cards that our loans are predominantly refinancing will move in tandem. This means that the savings that our borrower has realized, measured by the spread between our rates and the rates of the credit being refinanced, will remain reasonably constant.
Any decrease in loan demand at the margin from borrowers reacting to higher nominal interest rates will be more than offset by the growing demand for credit in the broader economy as stimulus it operates, as evidenced by recovering credit card balances.
As we look to Q1, we highlight the seasonal contraction we have historically observed between Q4 and Q1, which we have traditionally associated with tax refund season. While such seasonality has been attenuated more recently in the wake of COVID and the associated stimulus, we are expecting a return to the negative sequential pattern here in 2022.
With this as context, for Q1 of 2022, we are expecting revenues of $295 million to $305 million, representing a year-over-year growth rate of 148% at the midpoint; contribution margin of approximately 46%; net income of $18 million to $22 million; adjusted net income of $50 million to $52 million; adjusted EBITDA of $56 million to $58 million; and a diluted weighted average share count of approximately 95.9 million shares.
For the full year 2022, we expect revenue of approximately $1.4 billion, representing a growth rate of approximately 65% from the prior year; contribution margin of approximately 45%; adjusted EBITDA of approximately 17%; and an auto loan transaction volume of approximately $1.5 billion.
It is worth highlighting that the decrease in contribution and EBITDA margins we are guiding for 2022 relative to 2021 is intentional and controllable and largely a function of two levers. One, the speed of the ramp-up in auto lending, which will reduce our overall contribution margin by about 5 percentage points until it attains mature scale in operations and customer acquisition.
And two, the objective of growing our technical workforce by around 150% this year, which we view to be the most lucrative reinvestment opportunity for our corporate profits. Obviously, either or both of these investment decisions remain at our discretion and are susceptible to being revisited should any changes in our financial trajectory warranted.
Before I turn it over to Q&A, I want to highlight as a final note that we recently announced the authorization from our Board of Directors to repurchase up to $400 million of Upstart shares. With the volatility in the trading of our stock, we have seen what we believe to be attractive buying conditions at various times over the past year, and our profitability puts us in a position to be able to initiate this program and take advantage of those situations on behalf of our shareholders. Our thanks once again to all the talented Upstarters who are helping to build this company.
And with that, Dave and I are now happy to open the call to any questions. Operator, back to you.
Question-and-Answer Session
Operator
Thank you. [Operator Instructions] And we’ll take our first question from Simon Clinch with Atlantic Equities.
Simon Clinch
Hi. Greeting from the U.K. So first of all, congratulations on a very strong quarter. And I’m really interested in the economics of the auto business. I was wondering if you could perhaps walk us through sort of how the revenue model is going to work and how it perhaps differs or will differ in terms of the P&L impact, as we roll it through from loan assumptions.
Sanjay Datta
Yes. Hi, Simon, great to hear from you. So auto economics, I think, I would say, we’re not yet at the point where we’re ready to give a precise view on the unit economics. For the simple reason that almost all of the loans to date have been staying on our balance sheet, so currently, we’re earning net interest income off of those loans, which is obviously not our core model.
At some point, those loans will start to make their way to banks and investors. As I said in the prior remarks, we anticipate that happening over the next quarter or so. And when that happens, we’ll begin to pivot to a fee model that’s more akin to our core business model, but that’s still in front of us.
And then on the cost side, consumer acquisition and operations are still things that are — I would call them subscale. So we have targets for where we want them to get to, but not actual results yet. So taken all together, we don’t have precise guidance for you.
I guess, what I would say as a general statement, is that the overall take rate that we earn over the life of an auto loan, we anticipate at scale to be in the same ballpark to what we earn on personal loans.
I suspect less of it will be upfront on transaction and maybe more earned ratably over the life of the loan. But that’s what we expect to grow into as we scale. But as I said, we’re sort of not yet in that model where we can give you actual guidance yet.
Simon Clinch
Okay, great. And just a follow up on that as well. In terms of your rooftop expansion, could you talk about the pace at which you can expand and sort of what, I guess, the — where you think you might end up to in this year, or what kind of bottlenecks you might be experiencing, what the challenges are and actually ramping that up rapidly?
Dave Girouard
Hey, Simon, this is Dave. Yes. I mean, it’s not a specific number we’re giving guidance on. I would say, generally, if you looked at the numbers, and you can see them in the investment deck, we did see acceleration in the fourth quarter, which is nice. We actually rebranded from Prodigy to Upstart Auto Retail in the third quarter. So it was nice to see that, that didn’t cause any disruption. In fact, it was an acceleration in Q4.
I think generally, the biggest challenge for auto retail at the moment is the supply chain that car manufacturers and the auto industry overall is seeing, meaning it can be challenging to sell software to a dealership that helps them sell more cars when they don’t have enough cars to sell in the first place. But despite that and despite that headwind that we’re selling into, as you can see, we are expanding pretty rapidly. So our expectation is we’ll see rapid acceleration of that over the year. Certainly, as supply chains repair themselves and inventory levels on car dealerships, et cetera, begin to return to normal, we think that will be a tailwind that will just further accelerate. So we’re really pleased with the progress and we think we would expect to continue to accelerate adoption across rooftops through this year.
Simon Clinch
Okay, great. Well, thanks. Thanks David. I’ll jump back in the queue. Thanks.
Dave Girouard
Thanks Simon. Next question.
Operator
Thank you. Next, we’ll take from Pete Christiansen with Citi.
Pete Christiansen
Good evening guys. Thanks for the question. Impressive results. Dave, I know last quarter, you talked about expanding your credit — your target credits in both directions, both prime and more towards the lower end. Just wondering, how do you see yourself performing more towards the prime end where cost of capital is more of a competitive edge there? Are you seeing headway in that particular area? And where do you see your Upstart’s competitive advantage?
Dave Girouard
Yeah, Pete, thanks for the question. We’re definitely making pretty quick progress into the primary end of the market, which has historically not been in a place where we had a lot of presence. I think it has been relying on us bringing bank partners on board who have depository funding, the cost of funding that can compete for those borrowers. And that’s really what’s seen us begin to make headway there.
So I think all else being equal, the winner in any particular segment in the market is going to be a combination of the cost of funding available and then the quality of the model. And higher — the more you get into the primary segment, the cost of funding becomes more dominant, but still ability to model better, having a more efficient process, not asking people for documents, instant approvals. Those are all things that are helping us win in the primary end of the market.
Also, the truth about a lot of marketing isn’t as targeted as you’d like, meaning you can’t always target a specific part of this credit spectrum, if you will, in marketing. And that means it’s very important to be competitive across as much of that spectrum as you can because it allows you to do more broad-based marketing. And that’s really what we’re seeing as well. We’re seeing really great progress in digital, other channels that aren’t nearly as targeted as you might see elsewhere and being competitive in prime is really helpful for us because it allows us to use those types of marketing channels.
Pete Christiansen
That’s helpful. And then I’m just thinking about your rate requests, which were super impressive, I think, up 30% sequentially by my math. It seems like you are grabbing more eyeballs. Can you just explain a bit where — what marketing strategies you’ve expanded upon to drive increased eyeballs to the Upstart platform, where are you seeing greater success? Is this still in the affiliate marketing channel? Are you seeing more of a pickup in direct, that sort of thing? And then I’ll get back into queue.
Dave Girouard
I would say it was generally very broad-based. We saw improvements across every channel and pretty significant ones. Affiliate channels continue to grow. As I mentioned, digital has been very successful. We saw — in the fourth quarter, we had unprecedented performance in direct mail, which has always been an important channel for us. And then also very fast growth in kind of an organic users, which also part of which are repeat borrowers, so that’s also growing very quickly as we’re kind of ramping up our life cycle marketing efforts. So I don’t think it was any single channel. It really was very broad-based, and we think there’s a lot of upside in all these channels going forward.
Pete Christiansen
Great. Thank you. Great job.
Dave Girouard
Thanks, Pete. Next question
Operator
We’ll move on to Arvind Ramnani with Piper Sandler.
Arvind Ramnani
Hey – hi, thanks and thanks for posting another terrific quarter. Just a couple of questions on my end. Can you talk about some of the pricing dynamics with your partners? And given some of the normalization of consumer credit and sort of the impact you’re going to see on the business?
Dave Girouard
Sure. So the way to think about it is we have prices we charge to our bank partners, and those are costs that they absorb. But in terms of their pricing, the return targets that they are expecting in their loan portfolios and their loan programs, those are parameters that they control. So if for whatever reason, they decide they need a higher return on asset for any particular part of the risk spectrum, they can dial that up anytime they want. And that will result at the margins at a higher price to the consumer. So these are the kind of dynamics that the 40-plus banks and credit unions in our platform have available to them and can make decisions on their own.
And as usual, they’re trading off things like profitability, risk, volume, these are all sort of ways they can turn the dials to optimize the program for their business needs in a moment. So that’s the long and the short of it. These aren’t — we’re not changing our price to banks based on — what we charge to banks is not really a function of interest rates out there, anything of that nature. But what the consumer — the price the consumers may experience on our platform certainly are a result of what our bank partners are choosing to do in the market.
Arvind Ramnani
Terrific. And are you able to give us any color on sort of your kind of take rate, overall take rate as the year progressed? Just trying to figure out like, has it been trending upwards or has it been kind of flat through the year?
Sanjay Datta
Yes. Hey, Arvind, this is Sanjay. So take rates, I don’t think there’s maybe sort of one generalization we can say about the trend. They do go up and down as a function of other things. Mix is one example of things that will change the overall take rate on the platform.
More generally, I guess I would say this like when our models get better, it can result in one of two things. On the one hand, rates can get lowered for the borrowers and then volume increases. So that’s sort of one outcome. And then there’s another possibility, which is that the lower rates to borrowers are offset by higher take rates. And so our value manifests through a higher take rate. And which one of those to it is in any particular segment, sort of depends on how elastic the demand for our loan is.
So if small changes in APRs can result in large changes in volume, then that’s a great outcome for us. In some segments, our rates are getting to the point where they’re already so much lower than the market that lowering them further doesn’t really change the borrower’s propensity to take the lump and then so more of our value capture ends up materializing its take rate.
And so depending on which segments are growing because each one has sort of a different elasticity profile, some can sort of result in value as their models improve through the volume and others through take rate. It’s a bit hard to sort of generalize the trend overall in the platform now.
Arvind Ramnani
Terrific. And if I could squeeze one last in. Certainly, on the EBITDA compression, you provided some color in your prepared remarks on ramp-up in auto and hiring in tech. But if you think of like first for 12 months from now, what do you think may drive kind of upside? I mean, certainly, you have to invest in auto and you’re going to have hire tech folks. But is there any levers you have in place that can drive some upside in EBITDA margins?
Sanjay Datta
Yes, absolutely. And I would think that, if you think about the auto business itself, it’s going to go through a cycle much like the personal loan business did, where in the early days you’re ramping, you’re developing your sort of acquisition programs. They’re not quite at scale. Your operations are not finally tuned.
And so, in the early days of our personal loans, if you look back to our corporate history, we knew we had a lower level of profitability. And as that business scaled, it started accreting pretty directly to the bottom line. And I — we believe that auto is going to go through a similar cycle, maybe a more accelerated one, because we know the playbook now.
But as it is today, where we’re starting to achieve meaningful volume, our CACs are not as efficient as they are in personal lending, our operations unit costs are not as efficient as they are in personal lending, but they will get there. And as the models get better, the conversion funnel improves.
All of the things that happened to our personal loan business, we anticipate will happen on the auto business. And so whereas in 2022, it will be, maybe a bit sort of dilutive to our contribution margins at scale and at maturity, we think that it will have the same sort of profitability profile as our core business today.
So that’s just something — I think we’re sort of like incubating new businesses as we go. Dave sort of alluded to a pace of, you don’t want to read 6 to 12 months of new business, and it will go through an investment cycle.
But as we get more and more in our portfolio, that have matured and are sort of now ‘cash cows.’ I think the natural profitability of the overall model will just, kind of, trend to its equilibrium direction, which we believe is higher than where it is today.
Arvind Ramnani
Terrific. Thank you very much. I’ll just hop back in queue.
Jason Schmidt
Thanks, Arvind. Next question?
Operator
Thank you. We’ll take our next question from Ramsey El-Assal with Barclays.
Ramsey El-Assal
Hi. Thanks so much for taking my question this evening. I was wondering if you could share your early thoughts on the distribution strategy for the new products, you’ll be rolling out. Obviously, Prodigy really helps with auto. But should we also expect to see Karma — Credit Karma or other large distribution partners sort of playing a role in auto and also in these other new categories?
Dave Girouard
Hey, Ramsey, this is Dave. I think each of the products are very different in the nature of the channel development, I think, we’ll be pretty unique to them. So we will certainly use the relationships and the expertise we have.
For example, small business, there’s no doubt that, in our view, direct mail will be important to that. And we have what we would consider to be pretty exceptional skills in direct mail. There are some affiliate-type partners or aggregators in small business, probably not at the scale that some of them are in personal lending.
And likewise, in auto, auto clearly, direct mail is a great channel. It’s already proving to be the first channel that’s really taking off for us in auto. And there are some aggregators, but again, not as much a single point scale as we see elsewhere.
So they’re all different. But I think in almost every case, the channels that we have some footprint in today in personal lending will be meaningful probably with different weightings, if you will. And I think we’ll see a lot more diversity. And probably as important as anything is as we add subsequent channels, being able to cross-sell is going to be really important to us as well, and that becomes quite accretive to us.
So it’s a different sort of again, varies a lot by product. We love the partners that we have and would love to work with them on more products as we bring them to market and definitely anticipate doing that.
Ramsey El-Assal
Great. I know it’s early days. So I appreciate your read on that. I also wanted to follow-up here with a question on the repurchase authorization. It’s unusual to see a company — you guys are so solidly in growth mode. It’s unusually to see that in the capital allocation mix. How should we interpret that? Is this more a sign that you’re — you just feel that the share price is undervalued and you want to signal that to the market, or is there a decreased likelihood of capital allocation in other directions like M&A coming on the back of that?
Sanjay Datta
Yeah, hey Ramsey, this is Sanjay. Thanks for the question. Yeah, I would say, importantly, we have not run out of things to do by any stretch. As you know, we’re growing quickly and hiring a lot. So this isn’t a capital structuring decision, it’s economic opportunism. And it’s really a function of two things that are somewhat unique at our stage. First of all, well, one of them is the volatility of our stock is well known. You’ve seen it over the past year. And look, there’s — we have a conviction that there’s just been numerous instances over the past year, we’re knowing what we know about our business and our opportunity. We are of the opinion that it’s been undervalued.
And then the second component is we are actually profitable, so because of that it’s very unique feature, we have the actual ability to take advantage of that conviction on behalf of our shareholders. So it is opportunistic, and I think that the volatility in our stock continues, we’ll watch it, and we’ll be in a position to take advantage of that aspect. But it’s not really a conviction around returning capital to shareholders as much as it is taking advantage of the volatility of the stock and the profitability we have as a business model already.
Ramsey El-Assal
Great. That’s super helpful. Thanks so much.
Sanjay Datta
Thank you Ramsey.
Operator
Thank you. Next, we’ll move on to Andrew Boone with JMP Securities.
Andrew Boone
Hi guys. Thanks for taking my questions. I wanted to go first to default rates. So Sanjay, I think you talked about it being a feature, not a bug. But can you just give us a little bit more detail? Can you provide any incremental just pieces of data that give us more confidence there, talk about cohorts or anything else to just give us a little bit more confidence?
Sanjay Datta
Yeah, Andrew. Thanks. So I guess I was just trying to maybe draw a distinction between two different things that often get conflated. One of them, which is talked about a lot is, it is the fact with each successive vintage that’s originated on our platforms, the absolute level of delinquency default goes up, and that’s reflected in our securitizations. And so the first point I was making was that’s, in our view, not a bad thing. It’s happening because we are expanding our universe of approval of borrowers over time. And any time you go from a situation where you have a small amount of data and you’re acting conservatively, over time, having a lot more data and then relaxing your constraints around risk, then your average delinquencies will rise just mathematically.
And as long as you’re predicting that correctly and pricing the loans accordingly, our view is that this is a good thing. So in fact, I would say it’s maybe the best single distillation of our entire history of success in corporate value creation as a platform, right? Like that’s what we’re doing. We’re expanding the frontiers of approvability and making the universe bigger, whereas, we started from a position that was more conservative.
So I would say like put that aside. So that’s one thing that’s happening, but that’s just sort of a reflection of our business journey. But the second point, which is different, but equally about the delinquencies is that if you imagine that sort of delinquencies by vintage, which — where each vintage has a higher delinquency than the last, it is a true statement that every single one of those individual data points is lower than where we had expected it to be.
Now that’s a statement that’s equally true about all vintages and equal in magnitude about all vintages. And we are — we have been of the belief that, that’s because of the stimulus in the economy. And we’ve been consistently messaging that we have been predicting that, that would revert at some point and those little dots would return to the sort of position where we originally expected them to be.
And lo and behold, since October or November, each of those vintage curves is now reverting back to where we expected. So this is something that’s more of a local phenomenon. It’s not just about the secular vintage-over-vintage profile of our business, but it’s more about each individual vintage returning to a higher level of default. But the fact that we had been sort of predicting it for more than a year, and seeing it finally materialize separately is not a huge impact our business.
If you are in some temporary suspended state of abnormality, as long as you don’t dilute yourself into thinking that that’s the new normal, the eventual resumption of normalization shouldn’t be a surprise. And so we’re going through that shift, but that’s something that is new as of October or November. It’s not sort of a longer-term sort of increase in default profile, which I was just asking, too, because we see it discussed a lot in the public forum. So we thought it was worth clarifying.
Andrew Boone
Great. That’s helpful. And then my second question is just on the $1.5 billion auto goal. Can you just help us understand the potential upside as well as downside like where would that be higher? And why might that be lower as we think about that goal for 2022? Thank you.
Dave Girouard
Sure. Well, look, obviously, it’s early in the year, and we have sort of achieved lift off, if you will, with auto. So that’s why we feel comfortable presenting that number. But we have a long way to go, and it certainly depends on us continuing to make progress through the year. So there are certainly scenarios where it could be better than that and some where we would be less than that. And that’s our best view as to what we have sitting here in February. But just like in the personal loan world, for us, it comes down to our models improving as quickly as possible us for moving friction, us getting better at finding distribution channels and acquisition channels, getting better at cross-selling.
So there’s probably seven or eight key inputs to that formula of how good does that business look come December. And certainly, it is one of the most central areas of focus for the company in 2022. And I guess, we’re just sharing that, we have enough confidence to put a real number out there, a meaningful number and we’re going to go to work and take that business as far as we can this year. But we’re optimistic and we’re just really excited because there’s a certain threshold you cross where it becomes real and viable and there was a time when we really needed 50%, 100% improvements to the funnel in order to really have this thing start to scale.
And now we’ve done that, and we can sort of get to the place where we can get much smaller wins one at a time to really grow from here. And that feels like what personal loans felt like just a couple of years ago. In fact, one of the points we made is that our auto funnel today looks much like what the personal loan funnel look like in 2019. And that obviously was the beginning of a lot of growth. So that’s what gives us some confidence in that market.
Andrew Boone
No. Thank you, guys.
Jason Schmidt
Thank you.
Operator
Thank you. And next, we’ll move on to Mike Ng with Goldman Sachs.
Mike Ng
Hey. Good afternoon. Thanks for the question. I just have two. First, I was just wondering if I could follow up on the margin commentary. Could we expect Upstart to get back to 2021 margins in 2023, or what does that visibility look like?
And when you talk about hiring the technical workforce, could you just provide a little bit more color on what the key areas of investment there are? Is that simply a doubling of this engineering and product expense?
And then second, I was wondering if you could just comment on whether you’re seeing any changes in institutional investor loan demand? And could you just remind us how reliant you guys are on the securitization markets? And have you seen any changes there? Thank you.
Sanjay Datta
Yes. Hi, Andrew, [ph], this is Sanjay. So a multipart question. Let’s see. The first question is whether we may see a return to our current margin structure in 2023, I guess, I would say that — so it’s obviously a little bit hard to see out that far in terms of the investments we’d be making.
I would say this, there’s no fundamental reason why our business over time won’t return to our existing profile and, in fact, beat it. It’s really just going to be a function of how quickly we’re incubating and investing in new businesses. So auto is the one that we’re obviously investing in this year.
I suspect by 2023, it will be accreting to the bottom line, not reducing our margins. It may have a slightly different margin profile in terms of the timing of the cash flows, but we think it will be in a similar ballpark. And so, each new business we get into, we’re planning on getting into business lending, small business lending later this year, a small dollar lending.
Maybe 2023 is the year we get into home or mortgage. They’ll each have a slightly different margin profile, but more importantly, a cycle of investment. And so really, this begins to take the form of portfolio investment.
But I do think that, as I said, we know the playbook on these businesses now. We know how to get them to profitability and beyond, and we’ve proven that in our core business. And so I think that our ability to incubate new businesses and get them to profitability will improve over time. And in the long run, I don’t see why we would not sort of meet, if not exceed our current level of profitability as we scale multiple businesses.
You asked a little bit about the technical hiring, which is a big — sort of a big objective of ours in the coming year. I mean, it’s pretty broad, computer scientists, data scientists, machine learning engineers, product managers.
These are the people that are refining our models, leading our expansion into new areas, you’re building, accelerating strength in our core business. They are refactoring our platform from a single product platform into a multiproduct platform. They are rearchitecting our code base from a monolith into a suite of micro services.
They’re building out bank-facing consoles, an auto dealer-facing consoles and their readiness for forays in the small dollar lending and business learning and mortgage lending. So it’s very broad.
But as we’ve said, it’s pretty much to our — in our view, there is a direct line between the work that’s being done on the technical side and the bottom line of our business, because that’s ultimately what’s at the core of the value that we’re creating as a business.
And then last question was on the reliance that we have on institutional investors and securitization markets, which I view to be a little bit differently. So look, the supply chain of money more broadly is obviously very important to us. It includes banks that are using their own balance sheet to fund the loans that they originate and then the excess volume that we have gets funded in the institutional world. I would say that the direct reliance we have when you exceed the balance sheet capacity of your aggregate banking footprint is on the buyers of the loans as a what we call a forward flow buyer. And so they do absorb a significant amount of the capacity that we create.
The securitization markets are almost more of an indirect thing for us because it’s the loan buyers themselves that then securitize the loans. And so it’s more of a — we don’t touch the securitization markets directly other than we help run the deals that the investors themselves contribute into. So I think each of those investors has maybe a different answer as to what, kind of, liquidity they need from the securitization markets behind them. But I would say there’s a significant fraction of a matter more than happy to buy the loans and just earn the yield without looking for liquidity in the ABS market. So I would say the reliance on securitization markets is less relevant to us directly.
Mike Ng
Great. Thanks for all the color Sanjay. Very helpful.
Sanjay Datta
Okay. Thank you.
Operator
Thank you. Next, we’ll move on to James Faucette with Morgan Stanley.
James Faucette
Thanks very much. I wanted to ask a related question is we’ve seen the normalization of the lending markets and borrowing markets, et cetera. Can you talk a little bit about what your sense of your bank partners, et cetera, are right now to continue to increase their — the size of their loan books and what you think, generally speaking, is the appetite to do so this year?
Dave Girouard
Sure, James. I don’t think we’re seeing any particular trend in one way or another. I think we’re just still early in the game, meaning we’re bringing new lenders on the platform. They’re mostly in starting in growth mode. There’s a few that are at what they would think of as their own peak or run rate.
So — how — I will say for sure, last year, there was a very severe need for loans out there, almost unprecedented in terms of excess deposits, lack of loans in the banking world last year. And that certainly dissipated. I think there’s a lot more belief that, that situation is correcting itself. So maybe they won’t have as much demand. It’s really hard to say. Well, I don’t — I can’t say that we’re seeing a trend one way or the other.
I think mostly, our hope is we’re going to bring on a lot of bank capacity, and it’s going to continue to create a better net experience for consumers than the platform and we feel pretty good about that this year. So I guess the shorter answer is we’re still in the early days of this, and we don’t see a place where bank demand is going to drop off in our platform or anything we would certainly expect it to continue to expand.
James Faucette
Good. And then the other question I had was, as you look at the performance of your underwriting and that kind of thing in the normalizing market, what are you benchmarking against? And how are you — how quickly can you make adjustments where you deem necessary, et cetera? I think there — we certainly get a lot of questions on how the performance of Upstart loans are comparable to that of other underwriting mechanisms, especially in a changing environment?
Dave Girouard
Yes. I mean I think the thing that’s unique about us is there’s really two separate functions that operate very independently. The accuracy of the model is really the domain of the machine learning team and that sort of side of the house. Their whole goal is model accuracy, nothing more or less. They don’t want the models to be underpredicting or overpredicting defaults or prepayments. They want to be as accurate as they can. And there’s just nothing more to what they’re doing and then trying to upgrade the models and continually get better at that.
On the other side, you sort of have the business that is bringing more banks on, bringing more investors on and sort of feeding the engine, if you will. And again, those banks decide what return they need given what they’re seeing in the market, what other choices they have in terms of deploying their balance sheet or their deposits. So that just plays out in a business sense.
But — and that might mean again, if banks decide they want a higher return on the loans that are getting through the Upstart platform, then they can choose to do that. And that’s just the dynamic. It’s effectively a marketplace dynamic, where the choice of return versus risk and volume, et cetera, is really in the hands of the bank partners and they make those choices. And the core function of Upstart more than anything else is to be as accurate as possible on the risk models.
And that does mean as the economy is shifting, the environment shifting that our models are keeping up with it as quickly as possible and trying to get stay as accurate as possible. That is certainly trickier in times when things are changing very quickly as they did two years ago when COVID first hit. And as they have in the last few months, as really stimulus has gone away and we’re kind of returning to normal. But again, that’s – that’s the primary job of that model. But what the consumer ends up experiencing in terms of prices is really a function of all that coming together.
James Faucette
That’s really great color. Thanks a lot.
Dave Girouard
You bet.
Operator
Thank you. And next, we move on to John Hecht with Jefferies.
John Hecht
Good afternoon. Thanks very much for taking my questions. First one is maybe a few different questions on the auto side. One is — how — maybe can you talk about origination activity thus far, how many loans you have in your balance sheet? And then what’s the cadence of the $1.5 billion over the course of the year? And what’s the mix, indirect versus refi?
Sanjay Datta
Hey, John, this is Sanjay. I’ll take the first question. Auto loans on our balance sheet, I would say it’s — I’ll call it a majority, but it’s just maybe the most significant category of loans we have right now, the biggest category of new loans that we’re sort of running R&D on.
And as far as the cadence of the $1.5 billion over time, so we’re not giving that split explicitly between refi and retail other than to say that refi is the program that’s off the ground and up and running, and our retail program is still much earlier stage. So what that ultimate split will look like by the end of the year, we’re not really guiding to, but certainly in the early half of the year. And then this initial surge that I think is giving us confidence to be able to sort of telegraph these numbers, it’s really more about the funnel that’s driving the refi business. And as to the cadence over time, we’ll see. I mean, I think we’re not giving sort of sort of near-term or current numbers, but we’re — I would say, at a run rate that gives us a reasonable level of confidence that we’ll get there and it will grow hopefully linearly as we go through the year.
John Hecht
Okay. Thanks. And then, Dave, you — like last quarter, you were talking about different parts of the market at different parts of the year that were crowded. And so you guys, given your models, were able to find opportunity in other parts of the market.
Just as we get into early 2022 here and now you’ve got your goals, is there anything we should think about in terms of where you see opportunity now and where that might cause a mix shift to occur over the course of this year?
Dave Girouard
Well, I’d say, in the personal lending category, we’re pushing really across almost all parts of the credit spectrum, as I think we said in the last earnings call. And I think that will continue. We are definitely bringing on more banks that really trend toward the primary end of the spectrum and will make us more competitive there, and we would anticipate that will keep going.
At the same time, the core mission of the company is to make affordable credit available to everybody, and that means kind of continues to expand the perimeter of people that we can bring within the sort of national bank level.
So that effort is continuing as well, including the small dollar product, which is really going to help us move that part of the market even faster, I think. As well as the Spanish product, which is still nascent for us, but I think is showing promises, way to just bring more people in the fold.
So it’s really hard to say where that will go on balance. The personal lending product, we’re really in a strong position today and can continue to push on all parts of that market.
And then, of course, the newer products, I think the great thing that we’re excited about is, we are really comfortable now we have kind of crossed the chasm, if you will, on auto. We feel confident when building that product.
And the important thing about that is, the second one, at least in our view, is much harder than the ones that come after that. So proving that our models and our technologies and our skills and our teams can kind of adapt to a second very different product, just gives us that much more confidence as we get into small dollars, small business.
And eventually, a lot of people want to hear about the mortgage market. We just think that we’re building the skills and the confidence you’ve got to move to the subsequent products. Also building credibility with bank partners, with capital markets, investors, et cetera, that are necessary to make progress in those categories as well.
John Hecht
Great. Thanks very much.
Operator
Thank you. And we will take Nat Schindler with Bank of America.
Nat Schindler
Yes. Hi. Thank you. Thanks for taking my questions. So two quick questions. One, can you just explain a little bit — I understand the investment on — in the operations side [indiscernible] why would contribution margins come down from the Q4 levels? You what, 5 2% you’re going down to 46% and then 45% for the full year. Just wondering if there’s any detail there.
Sanjay Datta
Hey, Nat, it’s Sanjay. Yes, I mean, it’s almost single-handedly function of — well, it’s two things. One, when we do have a significant surge on the revenue side as we had in Q4, we do tend to overshoot our contribution margin, because we plan to spend against what we were expecting to do in revenue and when we have a big quarter.
So on the one hand, the Q4 number is maybe a little bit inflated. But I think the more important thing is as we get into Q1 and into 2022, the auto business is starting to scale. And that’s at a level where the contribution margin is today are much lower. And they’re much lower for three reasons. One, certainly for the period of time, where we’re putting this stuff on balance sheet, there is no fee revenue model. So contribution margin is all about fee — revenue from fees. And right now, as we originate the auto loans and put them on balance sheet, there’s no fees coming in, there’s interest income. And so for some period of time, there will be no fee revenue.
And then even when there is, as I said earlier, I think maybe more of it may be earned ratably over the life of the loan compared to personal loans, which is all earned on transaction. So the revenue profile will be different. But I guess, equally importantly, all of the unit costs to originate an auto loan are currently subscale compared to personal lending. So, whereas, our CAC is at a certain level, our customer acquisition cost at a certain level and personal loans is very efficient, in auto, it’s not there. We’re still building our programs, learning what works. The funnel is still getting better. As Dave said, it’s still sort of circa 2019 in terms of efficiency. And then our operations costs still equally are not at scale yet. We sort of overbuilt in order to build this business and ramp it quickly and make sure we have safety margin. And when we’re operating at scale as we are in personal lending, those will be much more efficient, much more finely tuned.
So the combination of the fee model, the acquisition costs, which are still immature, the operations, which is still early stage are such that as the volume of auto gets bigger, the mix between the two, the auto will pull down the overall contribution margin. And in rough terms, I think we’ll run the personal loan business at a contribution margin that’s close to 50%. But as I said, I think auto as it scales if we get to the numbers that we’re talking about, it should pull down the full year numbers by on the order of 5%.
Nat Schindler
Makes sense. And then on a separate question on auto. Over the last 18 months or so, there’s been such absolutely absurd appreciation in the used car market. I think it’s actually the single highest appreciating category I saw. And — at least in last year. And with — that has made the gain on sales of loans very high because basically the risk on an auto loan went to zero, because you could just sell — if you repo the car, you could sell it for more than the bucket force of the loan was paid off.
So what’s going to happen if that normalizes? And how if suddenly stimulus, there’s a lot of belief stimulus added to that price appreciation. If stimulus goes away and price appreciation starts falling off, how is that going to affect auto loans over the next year, 18 months as you’re comparing certainly to the last 18?
Dave Girouard
Well, I don’t — there are recent phenomenon of used cars. Gaining value is, obviously, any of us who grew up — we all got advice that spend as little as you can on a used car because they only go down in value. And we’re in a unique situation right now. That situation is certainly not baked into our model, an assumption that the price of used cars is going to continue to go up. That’s I think safe to say nowhere in our model. So we aren’t banking on the unnatural situation we’re seeing in the market with respect to auto pricing either for new or used cars. So that’s — we don’t see that as something necessarily impactful to us.
Nat Schindler
If there’s any mean reversion though and prices actually go down to go back to renormalize, not that this will happen. But if it does happen, does that make the loan more risky? And just if the price goes up faster, is the loans less risky, if the price comes back fast, shouldn’t the loans be much riskier?
Sanjay Datta
Yes. Hey, Nat. I guess maybe another way of putting it is that we’re not — we’re not – maybe there’s an analogy to where are in sort of general lending on defaults. We’re in new normal situation but we’re not – we’re not diluting ourselves into believing that that’s sort of a new normal and we’re not pricing accordingly.
And so put another way, these auto loans that we’re writing — if the world did not normalize, it probably overperformed because we’re not sort of baking in assumptions that reflect the current reality. And so the current reality, what it’s doing is inflating values. And so if the world does normalize, presumably, our returns would go back to what you’d expect in a normal environment. So we’re not sort of baking in the current world in that respect into how we’re pricing.
Nat Schindler
Okay. Great. Thanks, guys.
Dave Girouard
Thanks, Nat.
Operator
Thank you. And that does conclude today’s question-and-answer session. I’d like to turn the conference back over to Dave Girouard, for any additional or closing remarks.
Dave Girouard
All right. Just going to wrap it up. Thanks, everybody. We’re really happy with how 2021 turned out. And obviously, we’re feeling pretty bullish and optimistic about 2022. So thanks for listening today. Thanks for all who have stuck with us through all this market turmoil, and we’re looking forward to a great year. And we will be in touch with you all very soon.
Operator
Thank you. That does conclude today’s teleconference. We do appreciate your participation. You may now disconnect.
Be the first to comment