The management said on the call that the macro is improving, evidenced by a special index score they defined by themselves.
CEO: While we continue to deliver model and product wins to support that growth, we also finally benefited from a macro tailwind, most clearly represented by the decline of the Upstart Macro Index in the latter part of 2024.
CFO: The macro did indeed remain largely steady over the back half of the year from a credit default perspective, even showing recent signs of improvement since attaining peak defaultiness sometime last spring, as reflected in our published macro index, the UMI. This welcome break from the consistently degrading environment of the prior two years allowed some of our more recent model improvements to see the light of day, giving us good momentum on conversion wins over the last two quarters of the year. Our conversion rate in Q4 was at its highest level in nearly three years.
And their guidance is based on an assumption of stable macro.
As we gear up for the year ahead, we will strive for the following objectives and plan against the following baseline assumptions. A relatively stable macroenvironment and a constant Upstart Macro Index. No assumption of any rate cuts. A historical pace of modeling wins and conversion gains that will drive the majority of our growth.
The CFO did mention seasonality when he gave the Q1 guidance. And what is the most impressive IMO is the full year guidance of $1B revenue, which will be a 57% growth! I know UPST had a track record of over-promising and under-delivering in the past. But they must be extremely confident at this time to give their guidance. If they deliver, this will be best-in-class growth in the current market.
With these items as context, and with a reminder that the first quarter is typically our seasonally slowest quarter, for Q1 of 2025, we are expecting total revenues of approximately $200 million, consisting of revenue from fees of $185 million and net interest income of approximately positive $15 million. Contribution margin of approximately 57%. Net income of approximately negative $20 million. Adjusted net income of approximately positive $16 million. Adjusted EBITDA of approximately $27 million. With a basic weighted average share count of approximately 95 million shares and a diluted weighted average share count of approximately 105 million shares.
For the full year of 2025, we are expecting total revenues of approximately $1 billion, consisting of revenue from fees of $920 million and net interest income of approximately positive $80 million. Adjusted EBITDA margin of approximately 18%. And we expect GAAP net income to be at least breakeven for the year.
For those who stopped following UPST long time ago same as me, I’m impressed to learn that UPST now has a much diverse product portfolio in addition to their sub-prime loan offering. And the CEO mentioned multiple improvements to their AI model, to which they attributed most of the growth. This made me believe that UPST could have reached an inflection point.
In Q4, overall, our origination volume grew 33%, and our revenue grew 35%, both on a sequential basis. On a year-over-year basis, this equates to 68% growth in originations and 56% growth in revenue. Originations for each of our new product categories grew at an incredible pace, with both auto and HELOC growing by about 60% sequentially, and our small-dollar relief product growing a stunning 115% quarter-on-quarter. None of this could have happened without insanely great work by Upstarters across the country. I want to thank each of them for believing in Upstart and achieving more than we thought possible just a year ago.
In our core personal loan product, we continue to deliver model innovations that separate us further from the crowd. Model wins that increase risk separation are the lifeblood of Upstart. They’re responsible for much of the improvements you’ve seen in our business lately, and our pipeline of potential future model wins is robust. If you recall with Model 18, the most impactful innovation was using the price of the loan, or APR, as an input to the model. This is what’s referred to as a feature of the model in ML speak, and it led to a giant leap forward in model accuracy. This model delivered much of the momentum we saw in the second half of 2024.
In Q4, we launched Model 19, which introduced a new capability called Payment Transition Model, or PTM. To explain this a bit, in all prior models, the underwriting model only considered the terminal state and timing of a loan in the training data set. In other words, the particular month when a loan was charged off or prepaid. PTM enables consideration of intermediate delinquency states that may have preceded the final status of the loan. This means delinquencies that recover to current are suddenly meaningful in the training data and can inform a more accurate model. It also means that our model can properly learn from loans that are delinquent but not charged off directly in the core model. We’re often surprised by the increase in accuracy these types of innovations deliver, but concepts like APR as a feature and PTM aren’t one-time boosts. They’re new model forms entirely. You can think of them as innovation vectors that offer our team many ways to refine and build on their advantages for a long time to come.
Moving on to our newer products. In Q4, we released new underwriting models for both our auto refinance and auto retail products, resulting in improved conversion rates and contributing to the roughly 60% sequential increase in origination volume that I mentioned earlier. Auto refi in particular has seen giant improvements in conversion rates, about a 7x improvement across all of 2024. Also, the modest reductions in base interest rates have begun to revive the auto refinance opportunity, and we hope to take full advantage of it in 2025. We’re increasingly focused on auto refinance as an excellent cross-sell opportunity for our millions of prior borrowers. Our HELOC product had a strong Q4, growing by approximately 60% sequentially, much like our auto business. This growth was driven by a combination of conversion improvements, cross-selling and expanding state eligibility. In Q4, we automated the counter-offer process, much as we did in personal loans years ago. This is an important conversion booster. In December, we launched a machine learning powered feature that increased instant income verification rates by 34%. We also ramped up our ability to cross-sell HELOCs to prior borrowers. We finished the year with our HELOC offered in 36 states representing 60% of the U.S. population.
There a lot of exciting news in the earning report and I recommend everyone to read the transcripts. I just started a 3% position in the pre-market today.
Cheers,
Luffy