Upstart has repeatedly stated that they have zero interest in becoming a bank. Here’s what the dialogue has sounded like to my ears over the past year and a half or so:
Market: Oh, dear, it looks like we have bumpy economic waters ahead. Upstart has proven it has a superior model in boom times, but they are untested in a bad economy. Risk! Risk! People will default!
Upstart: Well, yes, more people will default in a bad economy; but a better model is still a better model. All defaults will go up for any company offering loans. Our model will still have a lower rate of defaults than those using FICO; because it’s a better predictive model of who will pay off their loans. Loan volume will go down because rates will have to be higher. But the loans we accept will still have a better return than loans judged by FICO. Because Our. Model. Is. Better. Moreover, our model is improving with every loan, while FICO’s model is static.
Market: Risk! Run! They are unproven! Don’t buy those loans!
Banks: (the industry whose system is being disrupted by Upstart’s model) Okay, we’ll go for Treasuries.
Upstart: They don’t trust us, but the people we serve really need credit and we know our model is reliable. We trust us. So we’ll keep loans on our balance sheet because we know they will perform better than FICO loans. That will help us both further our mission by approving more borrowers and cushion the financial impact of the downturn on our company as the loans mature at a profit for us.
Analysts: (who are mostly paid by and immersed in the industry Upstart’s model is disrupting) Gaaaah! You’ve filled your balance sheet with super, ultra-risky loans when the economy is tanking! Sell! Sell! Horrible management!
Upstart: Well, okay, if that’s what you really want, we’ll get those loans off the balance sheet. But we’ll be taking a loss to do it.
Analysts: Do it! Do it!
Upstart: Okay. Done.
Analysts: Gaaaaah! Look at the huge losses! This company is trash! Sell! Sell!
Credit Unions: (Peeking out from behind the bushes) We’re actually having pretty good luck with our Upstart loans. We don’t have massive reserves like big banks, and we need to make loans, even in this environment. These loans are doing well for us and right by the people we serve. We have to price them higher, which will decrease volume, but we’re in.
Upstart: The market hates us, and everyone is saying “sell.” We’re tied to the 2-year treasury, which is flying higher with Fed rate hikes, which is going to depress loan approvals. Still, we don’t yet have enough bank partners to originate the volume of loans that come through us, so bundling them for sale is still necessary. No one seems to understand that a better model is a better model in any climate, and they won’t until it’s proven through this downturn.
We have to do what is best for the company and, ultimately, for those we are trying to serve. Our reputation can hardly get worse, and we know the value of the loans we approve. So let’s go back to making more loans and holding them ourselves. That will both improve our bottom line over time and watching those loans more closely will allow us to keep improving and training our model, which keeps our first-mover moat in place and expanding.
Analysts: Gaaaah! They took big, scary loans on their balance sheet, then they said they wouldn’t, and now they’re doing it again! Sell! Sell!
Upstart: Whatever. The macro is really bad for all lenders. The loans we’re holding will pay off at maturity, but since others don’t yet trust us enough to buy them, we could face a liquidity problem. So let’s put off our work on SMB loans and mortgages. We’ll have to lay off the folks working in that area, but the alternative is to (again) sell off loans at a loss. The market will hate us either way, so let’s make the cuts.
And we’ll do more intense work with institutions who already understand the value of our loans (or can be convinced by our data) to get them to commit to buying our bundled loans until the day when we have enough lending partners to pass along at least most of the originations to them.
Upstart Q4 2022: Hey, even though you hate us, big banks, we’re still fighting for the people who need loans, which means helping you as well as ourselves. So here’s what we’ve been developing while you’ve been clutching your pearls:
(From the Q4 2022 call transcript. Dave Girouard speaking)
We also took advantage of the volatile economy to significantly upgrade our model’s ability to understand and react to macroeconomic conditions.
Last quarter, I announced our plan to productize the Upstart Macro Index or UMI. This new metric measures how changing economic conditions like inflation and unemployment are impacting credit performance. We continue to make breakthroughs in our methodology for calculating UMI, and we expect to launch this monthly metric to the public later this quarter. This is an exciting development from our machine learning team.
In an industry-first, Upstart will provide lenders with near real-time insights into the financial health of the American consumer, allowing them to adjust their lending programs accordingly. This is a big step toward providing banks and credit unions with lending infrastructure that autonomously, continuously, and rapidly adapts to changes in the economy.
Oh, and you know what else?
Despite the fact that our lending volume in 2022 was down 14% versus the prior year, our contribution profit was actually up 13% year on year. Optimizing our pricing represents a large surface area of opportunity, which we’ve only just begun to explore. In addition to these major improvements, we’ve also continued to innovate across our platform in support of future growth. In fact, I believe we made more progress with our technology in 2022 than in any year in our history.
And as capital markets and the overall economy normalizes, I expect this will become obvious to all of you. Some important areas of progress from last year include model accuracy. Our AI models continue to separate risk significantly better than a traditional cycle-based model, and we continue to increase our pace of model development. The increase in our model accuracy in the last seven months is more than what we delivered in the prior two and a half years.
Today, our small-dollar product includes loans from $200 to $2,500, with tenors from three months to 18 months. To date, we’ve originated more than 24,000 small-dollar loans to individuals who otherwise would not have been approved for our personal loans. More than 12,000 of these loans were originated in Q4 alone. This expansion of borrower coverage means we are dramatically increasing the pace at which our machine learning models are improving.
And just as importantly, in Q4, 88% of small-dollar loans were fully automated. Lending partners. In our earnings call a year ago, I told you we had 42 lenders on the Upstart platform. Today, that number is 92, representing growth of 130%.
Despite the hostile 2022 environment, banks and credit unions recognize and appreciate a fundamental secular change in technology when they see it. These partners are starting cautiously with us, but they represent a significant expansion of potential lending capacity on the Upstart platform once there’s a bit more clarity on the direction of the economy.
A satisfying line from the CFO, Sanjay Datta:
Net interest income came in above forecast, largely a result of choosing to retain more loans on our balance sheet than anticipated given the market conditions in Q4.
And the irony in the final question from a Stephens analyst made me laugh out loud.
On the balance sheet, if you could discuss how much capacity you do have for increasing the amount of loans that you have on balance sheet and if there’s maybe other ways to increase that capacity.
So yeah, now you can see Upstart loans are performing and suddenly want us to take MORE loans on our balance sheet. You can’t make it up.
Market: Boy, it’s fun buying puts on Upstart.
Cramer: AI is really going to change the world–every company. But not Upstart, they’re not a good company.
Upstart Q1 2023 call:
I’m pleased to tell you that we secured multiple long-term funding agreements together expected to deliver more than $2 billion to the Upstart platform over the next 12 months. This is a critical first step toward building resiliency and predictability into our business.
We pushed 2023 new and improved versions of our AI models into production during the first quarter alone about one every three days. We’re confident that our AI has never been as sophisticated or as accurate as it is today. But in order to deliver the best rates for all, we also need a diversity of bank and credit union partners, each with different priorities, business objectives, and balance sheet issues to solve. Today, we have almost 100 such partners in order of magnitude more than the 10 we had when we went public in December 2020.
We additionally need a strong presence and reputation in institutional and capital markets because the limited risk appetite of bank balance sheets will never serve the needs of the entire U.S. credit market. This quarter, we expanded our roster of institutional partners in ways that should help us deliver quality offers to consumers through all parts of the cycle. Second, more efficient borrowing and lending, every quarter we aim to make our platform more efficient for consumers and bank partners.
Some Analysts: Congrats on the quarter!
Other Analysts: You had big losses! This is terrible!
Castlelake: Hey, we’ll take $4B of those loans off your hands, either ones you’re holding now (we’d like those profits that are obviously coming) or new ones you’d like to make.