Every person I truly respect takes accountability when they make mistakes, strives to learn from them, and works to do better. I endeavor to embody that ideal myself and this article is part of that process. The title above applies both to me and to Upstart (NASDAQ:UPST) the company. Two weeks ago I wrote about Upstart, calling it everything you want in a growth company. Last week's earnings report came with a significant downgrade in guidance and a nasty surprise that they are using their own cash to fund some of their loans. The stock price completely dropped off a cliff and now here we are. I will cover what I got right, what I got wrong, what Upstart got wrong, what Upstart got right, and where we go from here.
I also strive to always add unique analysis in my articles rather than regurgitating the same information that has previously been discussed or can be easily understood from an earnings presentation. There are a few nuances about the earnings, machine learning performance, and future outlook of the company that I have not seen covered anywhere that I will explore.
My previous article used LendingClub's (LC) recent Q1 results to forecast what Upstart would do on their Q1 earnings. I speculated that LendingClub's earnings beat on revenue and EPS was a positive sign for Upstart's chances of likewise posting a double beat in Q1.
Turns out I was in fact correct. Upstart's Q1 numbers looked great, with $310M in revenue compared to $300M expected (a 3.3% beat) and $0.34/share of profit compared to $0.24/share expected (a 41.7% beat). The first quarter was an extension of Upstart's consistent record of beating consensus estimates. Also, as predicted, Upstart originations grew QoQ, continuing to outpace origination growth from LendingClub. This is the same graph that appeared in my previous article but it now includes Upstart's 1Q 22 numbers. The trajectory of originations is exactly what I expected it to be.
I was, unfortunately, also right in my Risks section that highlighted that "The main risk for Upstart is that they have yet to experience an environment with rising interest rates. [...] A rising rate environment will be a stress test of the AI model that Upstart has yet to experience." That is a great segue into my next section.
I neglected to take into account the full extent of the risks involved to Upstart in a rising rate environment. I also put too much emphasis on the results of Q1 and not enough on the possible effects of a change in guidance or capital structure of the company. One of the highlights of Upstart's business has been that they are not a capital-intensive company. They had trimmed their balance sheet of loans and reduced their capital risks significantly since coming public.
In Q1, because of the rising rate environment and increased research and development (R&D), this trend reversed. They increased the amount of loans they hold from $261M at the end of Q4 to $604M at the end of Q1. This was a surprise to analysts and investors and was a big reason behind the precipitous drop in share price after hours.
There is one last thing I got completely wrong. I said this was everything you want in a growth stock. Considering the updated guidance, I might have to amend that statement to everything you want in a growth stock except, well, growth (at least as far as 2022 quarterly growth is concerned). 1Q22 revenue was $310M. Total guided revenue for 2022 is now $1.25B, compared with their previous guide of $1.4B. If you take their first quarter revenue and multiply it by four you get $1.24B. That means they expect every other quarter for the entirety of 2022 to come in with basically the same revenue numbers as the first quarter. While $1.25B of revenue for 2022 still represents a 47% increase in revenue when compared with 2021, no quarterly growth for an entire year is not what you want to see.
At this point, you might be scrolling back up to the top of the page to double-check that I did in fact rate this as a strong buy. I did, and I still think this stock is a great long-term investment. There are significant and important mitigating circumstances surrounding their revenue growth numbers that have to be considered and will be discussed below.
Upstart did not anticipate how adversely the market would react to the increase in loans that they are taking onto their balance sheet. There were a few passing questions about the guidance adjustment in the Q&A section, but every single analyst continued to circle back around to the issue of taking loans on the balance sheet. They also did not have a very clear message or guidance for how much they will be taking onto the balance sheet for the remainder of the year. This was asked in a variety of ways by the different analysts and the answers from management were guarded and vague.
They did provide a little clarity on the loans that is informative. Three quarters of the loans taken onto the balance sheet are for R&D, and most of those are for auto loan refinancing. The other quarter was personal loans. The reason stated for taking on the personal loans is that Upstart "started to selectively use our capital as a funding buffer for core personal loans in periods of interest fluctuation where the market-clearing price is in flux." In other words, their loan buyers, whether banks, credit unions, or others, were unwilling to underwrite a portion of Upstart's approved loans because of the fluctuations and uncertainty with the current rate environment. Upstart is using their own funds to bridge the gap until the uncertainty settles down. This is a red flag because it's a failure of management to adequately prepare for a scenario that should have been anticipated.
The loans taken on for R&D make sense and are not particularly troublesome in my opinion. This is part of their business. Banks, and other organizations that originate loans and take on debt securitizations, are very risk-averse. They are understandably wary of taking on the risk of a new loan class until Upstart has enough data to convince them that the algorithms work in the real world. While it isn't ideal for Upstart to use their cash to fund loans, this is a necessary R&D cost that cannot be avoided if they are committed to expanding their product offerings.
However, if Upstart was aware of this as a step in the R&D process, this should have been communicated prior to this call. The very last thing Wall Street wants in the current environment is an unpleasant surprise, and you get the feeling that this caught the analysts completely off guard. These analysts are in regular contact with Upstart management, and you could hear how confused and concerned they were at the change in direction.
*Start of edit*
After I submitted this article, some news from the Wall Street Journal came to my attention. Upstart CFO Sanjay Datta has clarified that they will no longer park loans on the balance sheet that financial institutions are not interested in buying. They will continue to hold loans for R&D purposes for vetting new products. They also announced that they began buying back shares with their buyback program.
*End of edit*
Finally, the guidance downgrade means they are less resilient to rate hike cycles than previously thought. Analysts were expecting $335M of revenue in Q2 and $1.4B for the year, and Upstart's new guide is for $295M-$305M for Q2 and $1.25B for the year. The downward guidance was a result of four factors:
Rising rates mean rising costs for the borrowers. If the cost of the loan goes up, some will choose to delay or forgo the purchase or debt consolidation that was the reason for them to take the loan in the first place.
Partner banks are unwilling to take on as much risk because of the uncertainty of the economy and the uncertainty surrounding rate hikes. Lower risk tolerance means less desire for them to originate loans.
Their new auto loan segment is for refinancing existing loans. People usually refinance into a lower rate to save on their monthly payment. Rates rising so quickly means less incentive to refinance because the monthly savings decreases over time.
The effects of rising rates on fair values of loans held by Upstart (more on this in a minute).
Let's talk frankly about those loans on the balance sheet. They increased the loans they hold by $343M during Q1. The upside to holding loans on their books is that their interest income should increase. This is added revenue. However, there are three disadvantages that need to be explored.
The first is the risk involved with defaults. Rising rates, a slowing economy, and an end to stimulus have already resulted in rising delinquencies. Upstart has already lent that money out, and if they see a high rate of default, they will never see that money again. Every loan that is on the books exposes them to potential losses through default. Their own data even show rising defaults in the last few quarters.
Management said that they have seen stabilization in the last 60 days around the 0% mark, meaning that their loans are now performing as expected. They attributed the previous overperformance to government stimulus. That makes sense as people are more able to make payments when they get a windfall from an unexpected source such as a stimulus check.
However, should their models begin to underperform, they are now on the hook for those loans. If you believe that their models will perform (and I will take a deep look at this point in the next section), then their risk is low. However, if their models underperform, this decision will prove costly moving forward. This is somewhat mitigated by the fact that most of the loans they financed from their own cash reserves in Q1 are auto loan refi loans. That means that there is a physical asset that can be seized if the borrowers default on the loan. This decreases Upstart's risk relative to holding unsecured personal loans.
The second disadvantage of increasing their own loan portfolio is that it becomes a loss of revenue in a rising rate environment. This, as far as I can tell, has not been discussed by other analysts, and it is a very important point to understand. Without going into too much detail, the loans held on the balance sheet are assigned a fair value of what they are worth. As rates rise, the "fair value" of the loans on the books decreases because entities buying the loans now have a higher cost of capital.
This is best understood with a hypothetical scenario. Let's assume that 3 months ago Upstart originated a $10,000 personal loan with a 10% interest rate. I am a bank who buys loans and my cost of the capital three months ago was only 0.5% because rates were so low. At that point in time, I would be making 9.5% on the spread between my cost and the interest rate of the loan, so I am willing to buy that loan for $11000 knowing I'll make that $1000 back on the spread. Today, after a 0.25% hike and a 0.5% hike, my cost of capital is now 1.25%, and the spread is only 8.75%, so I'm only willing to pay $10,700 so I can still get a good return on that investment. So the "fair value" of that loan that Upstart has on their balance sheet decreased by $300 between last quarter and this quarter.
The combination of the net interest they earn off the loans on their balance sheet and the change in the fair value of those loans makes up Upstart's non-fee revenue. If rates rise slowly, the increasing interest income from the loans held makes up for the change in the fair value. However, when rates rise fast, the decrease in fair value can overwhelm the net interest revenue. Fair value adjustments were a tailwind for all of 2021, contributing positive revenue to every quarter last year. However, the winds have turned, and these adjustments represent an $18M revenue loss in 1Q22 (that is over 5% of the revenues from fees). This loss was enough to completely offset all interest earned and become a drag on quarterly revenues.
This is actually a big reason why 1Q22 revenue numbers were weaker than originations would suggest. Comparing 1Q22 to 4Q21, originations grew 10.7% and revenue from fees grew 9%, but total revenue only grew 2%. That discrepancy results from the fair values adjustments. They changed from a $9.3M gain in the fourth quarter to an $18.0M loss in the first quarter, a $27M swing. This is also a partial explanation for the decreased guidance for 2Q22 and the full year. Interest rates are expected to move faster in the second, third, and fourth quarters of 2022 than they did in the first quarter. So the fair value adjustments are a significant headwind moving forward.
Holding loans on the balance sheet means that you are using up valuable capital that you could spend elsewhere. That cash could be used on marketing, additional product development, or even as part of the share buyback program. The cash deployed in this way is necessary for de-risking new products, but because of rising rates, I expect it to be a drain on revenue throughout 2022 through fair value adjustments. Cash used to finance loans decreases both the top and bottom line results and keeps them from being able to use that cash to contribute to the business in other valuable ways.
That was a whole lot of bearish analysis. In the face of the significant headwinds that Upstart is facing and several managerial miscues in execution, why in the world would anyone suggest that this company is a worthy investment? I invest with a minimum holding time of 3 years in mind. Short-term price drops, headwinds, or macro changes do not deter me if the underlying thesis remains strong, and the potential for future earnings remains bright. We got a lot of data in this earnings report that confirm that the underlying thesis is as strong as ever. What is more, even though there will be short-term pain from funding loans, it is still the right thing to do for the business and for investors as long as it is for product development.
Upstart's entire business proposition comes down to one simple question: Can their algorithm predict delinquencies better than other methods? If it can do that demonstrably better than other methods and that performance is repeatable, they will continue to grow and investors will be well rewarded. So the most important question we can answer is whether or not their AI is worth the referral fee that banks pay. Right now the industry standard is the FICO score. Let's see how they compare.
As a chemical engineer, I am used to seeing really important data put into really cumbersome graphics and charts that are difficult to parse. This, for example, is a diagram I have used to design cooling systems. Yes, there are 6 different axes on that 2D chart. I was convinced that an engineer came up with the following table that appeared in the Upstart first quarter earnings presentation. That suspicion was confirmed when I sent it to my sister who is a graphic designer and she sent me about 5 different facepalm GIFs as a response.
This table contains, in my opinion, the most important data that Upstart has ever published, even though it is not immediately apparent what any of it means. The table shows default rates for loans that Upstart lending partners have made from 1Q18 through 3Q21, broken down by Upstart Risk Grade and FICO Score. This data includes both Upstart referrals and other loans that partners have originated over that timeframe. That is almost 4 years of data and represents a significant enough sample size to draw conclusions.
Let's walk through how to interpret this chart and then discuss its ramifications. The first important thing to realize is that Upstart breaks potential borrowers into 5 "Risk Grades" which they label A+, B, C, D, and E-. If you ignore all the middle numbers and just look at the average of each Risk Grade, you'll see that A+ grades have a default rate of 0.7%, E- grades have a default rate of 9%, and the other grades fall in between. I've included the same graphic again with all other data muted so you can see what I am referring to.
Similarly, you can ignore all the middle numbers and see that as FICO score decreases, average defaults increase. The table shows a 3.4% default rate for borrowers with scores of 700 or above all the way to 7.7% for borrowers with a score of 639 or below. Again, this is easier to see in the graphic below.
Just looking at the averages already gives you a picture of why banks are willing to pay a premium for an Upstart referral. A 0.7% default rate is unheard of. This does not guarantee that default rates will remain at 0.7% moving forward, as this timeframe overlaps with the overperformance during COVID times as seen above. However, the FICO score loans have the same benefit from stimulus and, nevertheless, they wildly underperform relative to Upstart Risk Grades. During these four years, Upstart's Risk Grade A was less than one quarter as likely to default as someone with a FICO score above 700. Upstart Risk Grades are more predictive and repeatable than FICO scores.
That is great in and of itself, but the truly impressive performance of the Upstart model is in the corner cases, as highlighted in this version of the table below.
If I were a bank, this is the data that would have me salivating. Experian defines super-prime borrowers as anyone with a FICO score of 740+, prime borrowers fall between 670-740, and subprime is anyone with a score of 669 and below. If you look at the top right corner, you'll see that people with a FICO score of 700+ but who have an E- Upstart Risk Grade default an incredibly high 9.2% of the time! That means that Upstart's algorithm can accurately identify groups of prime and super-prime borrowers who default on their loans more than 8% of the time. Likewise, the algorithm can identify groups of subprime borrowers whose default rate is around 1% even though their FICO score would suggest they default at a rate between 5.4% and 7.7%.
Let that sink in for a minute. The Upstart algorithm identifies borrowers whose FICO scores are 639 or below that only have a delinquency rate of ~1%. These are people who under normal circumstances would be laughed to scorn by banks for even considering taking a personal loan but who are one-third as likely to default as the entire cohort of borrowers whose FICO score is 700+. A bank can easily charge these people a premium because they can't get credit anywhere else and rest assured that the loans are in fact extremely low risk.
Just as important are the people who banks are giving loans to at prime or super-prime interest rates, but who are about three times as likely to default on their payments than the average borrower with their same FICO score. These are people that banks believe are low risk but are costing them billions of dollars in defaulted loans. And Upstart can identify them and keep those billions in losses safe in the bank's coffers.
These published default rates suggest that Upstart Risk Grades of C perform just as well as FICO scores of 700+. Both Risk Grades A and B significantly outperform prime and super-prime borrowers. The model is working, it's working extremely well, and the more data they get the better it can perform.
In December 2020, when Upstart released their IPO, they had a grand total of 10 lending partners. They are now up to 57 and added 15 during 1Q22, or a little over one per week. Growing their portfolio of banks and credit unions lowers their risk by decreasing their reliance on any one lender to take a loan. It also expands their user base as these partners use Upstart technology to qualify their customers who come to them for credit approval.
Even more exciting, and further verification that the algorithm is working, is that 11 of their partners have removed all minimum FICO score requirements from their credit policies, up from 7 in 4Q21 and zero this time last year. This means lending partners have seen firsthand that Upstart's model is significantly outperforming FICO scores. The performance is good enough and dependable enough that they are willing to completely abandon the conventional FICO score method for approving loans.
Finally, the Salesforce CRM integration I mentioned in my last article should only accelerate the acquisition of lending partners moving forward. That integration reduces almost all the friction for any bank or credit union that uses Salesforce to manage customer relations. It would make switching to using Upstart to approve loans completely seamlessly.
The final thing that Upstart is getting right is that they are not afraid to put the long-term needs of the business first. Auto loans are only the first new vertical they are exploring. During the Q1 earnings call they mentioned two other new products that are in the pipeline. According to CEO Dave Girouard:
I told you in November that we aim to launch the small dollar loan before the end of 2022. Our small dollar team set an aggressive goal to launch the product by the end of Q1, and I'm pleased to report that they achieved this ambitious goal.
Additionally, I'm happy to share that our small business lending team is likewise making impressive progress and is aiming to have their product in the market within a few months.
Unfortunately for 2022 results, this points to even more originations hitting the Upstart balance sheet in the coming quarters as they begin vetting their small dollar loan and small business lending models in the real world. I would also expect to see an increased R&D budget as they work out the intricacies of these new markets.
Because we find ourselves in a rising rate environment, taking loans on the books will probably result in more fair value losses that continue to be discounted and will decrease revenues in the short term. This is just speculation on my part, but this might be another reason they lowered the full year guidance and why they were cagey with analysts about the balance sheet questions. They know they are ramping up these activities and they know that investors and analysts are not going to view that move favorably in the current environment.
Despite the fact that this will be a drain on resources and results for 2022, it is unequivocally the right thing for the business. Up until now, Upstart has relied heavily on increasing originations of unsecured personal loans to drive growth. However, there is a glass ceiling with personal loans that they are bound to hit sooner rather than later.
The entire personal loan TAM is ~$100B annually. Upstart originated $4B loans in 4Q21 and $4.5B of loans in 1Q22 and the overwhelming majority of those were personal loans. That means Upstart already has a 15-20% market share in unsecured personal loans. That is already very high penetration and leaves little room for growth. Now is absolutely the right time to be expanding their product offerings.
On the earnings call, Dave Girouard said "We see a clear path to building a company with more than $10 billion in revenue in the coming years and are maniacally focused on achieving that goal." If they are getting ~$300M in revenue fees on $4B of quarterly originations, it means they would need $33B in quarterly originations to meet their goal of $10B in annual revenue. That is impossible if personal loans are their only product. To achieve the growth they've targeted and that investors require to get the returns they want, Upstart has to give itself optionality across the entire lending spectrum. Investors and analysts may not like the increased R&D spending and increased loans on the balance sheet, but it is absolutely the right move for the company.
The market did not like Upstart's earnings report and shares plummeted 56% the day after earnings. They plunged all the way to $25.43, but have since recovered to $51.58 at the time of writing. In fact, if you bought at the bottom, you would have already doubled your money on that investment. That is because the selloff was unjustified given the results. Yes, 2022 might be a harrowing year for the company as they spend on R&D for new products and take on perceived risk by funding those loans themselves to vet their models. However, this is a company with a P/E ratio of 31 that is still in the early stages of its growth story. The long-term thesis remains unchanged, and Upstart remains a buy.
This article was written by
Disclosure: I/we have a beneficial long position in the shares of UPST, LC either through stock ownership, options, or other derivatives. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.
Additional disclosure: The information contained in this article is for informational purposes only. You should not construe any such information as legal, tax, investment, financial, or other advice. None of the information in this article constitutes a solicitation, recommendation, endorsement, or offer by the author, its affiliates or any related third party provider to buy or sell any securities or other financial instruments in any jurisdiction in which such solicitation, recommendation, endorsement, or offer would be unlawful under the securities laws of such jurisdiction.