They beat their forecast for Q2 by 21%.
They beat their forecast for Q3 by 6%.
They grew Q/Q 60% from Q1 to Q2.
They grew Q/Q 17.5% from Q2 to Q3.
But they are forecasting Q/Q growth from Q3 to Q4 of 16.2%.
However, we expect them to beat their forecast. If they only beat their forecast by 5%, they would finish Q4 with $278m. That would give them an actual Q3 to Q3 Q/Q growth of 21.9%.
So if my napkin math is correct, we expect to see Q/Q acceleration when they next report.
Anyone see that differently?
Barclays analyst asked this exact question, about how Q4 looks like a strong guide, and why that is the case, was it due to people holding off until later in the year or any other factors.
Sanjay’s response somewhat escapes me, he did mention something about fraud mitigation, and he also mentioned a somewhat lumpy market quarter to quarter.
If anyone can help with Sanjay’s response to Barclay’s guidance question, please add!
Sanjay replied to the Barclays analyst that their guidance is based on what they see currently and that as they have said before their growth is driven by model improvements that can be choppy. My interstation is that we should see an growth improvement next q.
Q/Q growth is interesting but also dangerous because it can be strongly influenced by potential seasonality (think champagne: Q4 is always much higher than Q3 because of the Holiday season. It doesn’t imply that Q1 will accelerate further.) I have no idea how much of a role seasonality plays for UPST’s business. Unfortunately we can’t see it in the available historic quarters because Covid muddies the waters too much. But in the absence of that knowledge I would be careful to put too much weight on a possible acceleration from Q3 to Q4.
It’s a S/W term, Machine learning is based on the model, the more data ML program consumes more intelligently it can predict, this is called training the model. So with wider customer base, model is improving every day.
So if my napkin math is correct, we expect to see Q/Q acceleration when they next report. Anyone see that differently?
The math seems right but the question of if it’s a “dip-buying opportunity” is complex.
It depends on how much the street was expecting this quarter to be up after the 1000% blowout. A dip from $400 is definitely happening. What is fair value now is much harder!
If you have a model that is completely autonomous, and the information for that profile was purchased on the dark web, does the model issue the proceeds and allow for a fraudulent loan to be executed?
Additional fraud checks would be needed if this is something that was out of hand (even marginally impacting numbers).
My understanding is that the model can have a higher loan acceptance rate as the growing portfolio base builds a deeper pool of information, metrics and trends which AI can work with. So growth accelerates as the model improves
GolfCaddy,
Regarding your question: Can anyone explain what exactly they mean by the models improving? I didn’t fully understand. Do they need to train the model for each banking partner?
My understanding is that the AI models rely on real-world data to:
“predict” possible future outcomes (for a company like Upstart)
to make decisions (for machines in factories, or for automated driving, for example)
The more real-world data that the models have, the “smarter” they get. That is the better they get at making decisions (or predicting who will default on a loan).
So in the case of a company like Upstart, the more loans that they make, then the more real-world data the have, and the better the models get at making decisions.
I hope this helps. If I am misunderstanding this at all, please correct me.
“Can anyone explain what exactly they mean by the models improving? I didn’t fully understand. Do they need to train the model for each banking partner?”
First post on this board, hope it is helpful.
I am an engineer and work with machine learning and AI, so I think I can comment on this.
When I heard him say that, I assumed he was talking about the AI models they use. I also assume that the conversion rate is tied to how well the models work. So if they improve the models the conversion rate may improve as well, which means more revenue. However, the conversion rates have been pretty stable for a few quarters now for what it is worth. Also as banks get more confident in their models, they may originate more loans that they would not otherwise.
I was also surprised that they said onboarding a bank in only 50 days is fast. I would assume there must be some custom training in that case. Unless there is some other regulatory or other reason why it would take that long.
Finally, AI models are all about the data you have. UPST is getting exponentially more data this year as it deals with many more loans and customers. So it is feasible that the models are improving. They touched on this in the call when talking about the small dollar loans and how it will enable them to get more data on different categories of customers than they have now. In fact they said the small dollar loans were more about the data and not necessarily the revenue they will generate.
Just started growth investing and I have small position in UPST -
If they only beat their forecast by 5%, they would finish Q4 with $278m. That would give them an actual Q3 to Q3 Q/Q growth of 21.9%.
I agree. However, some parts made me think before buying more because the guidance does not only include revenue but also earning numbers. And some numbers in the current reports are not very ideal.
First, margins and conversion rate are worse compared to last quarter (data from their presentation)
They said they had to remove fraudulent loan requests to get 23%; otherwise, the number would be 13.5%, significantly smaller than 24.4% from the last quarter. This number from previous quarters is immaterial. Does that mean the “fraudulent” problem just showed up significantly this quarter? Why?
Second, they guided for 53M in adj EBITDA and 50M in Adj Net Income, even smaller than this quarter while the revenue still increases. Does that mean the margin is gonna go down even more? Why?
I haven’t listened to the call yet and usually prefer to read the transcript. Does anybody have explanations for my two concerns?
I decided to do nothing until I understand what happened.
Fraudulent loans - as Upstart’s profile has gotten bigger, they’ve been more of a target for fraud. They said in Q3, the fraudulent loan applications were more organized. They had to spend resources dealing with it. But it’s been an issue since day 1.
They also said that 4 banks got rid of FICO scores, and as a result the loan amounts from those 4 banks were smaller. Again, it’s about collecting data, refining models.
Second, they guided for 53M in adj EBITDA and 50M in Adj Net Income, even smaller than this quarter while the revenue still increases. Does that mean the margin is gonna go down even more? Why?
Didn’t hear call. Will read transcript later.
But others mentioned a forward-looking view to investing in solutions for mortgages…that there would be an investment cost associated with ramping in that space.
To an extent, would we see that with auto as well.
Point being, increases in operating expenses/reinvesting into new avenues of growth, could weigh on profitability.
Not sure if that is the reason, but it could be “a” reason.
So if my napkin math is correct, we expect to see Q/Q acceleration when they next report. Anyone see that differently?
I do not see it differently and, in fact, modeled the exact same beat of 4.9% for next quarter or sequential growth of 21.7% which is annualized growth of 119%.
Just looking at historical numbers, Q4 does not seem to be a seasonally strong outlier either though my data only goes back two years. For instance, Q4 2019 sequential growth was 26.5% off of a much smaller base.
But the big question is how will they perform going forward. Investors can’t rely on recurring revenue with Upstart like we can with most SaaS companies. So that is a question mark and one that could keep the valuation lower than the best SaaS companies.
20%+ sequential growth is certainly nothing to sneeze at. It is a no brainer of an investment if one assumes Upstart can keep that up!
Newby here. 17% holding in UPST. After the earnings call I’m left wondering how everyone is going to react. I won’t try and draw a conclusion since I haven’t earned that yet, but here are some observations vs. Saul’s tenets:
Saul Tenet 1: “First, most of my stocks start with a recommendation and write-up by someone I have a lot of confidence in. This could be someone on the board, Bert, Motley Fool, or more rarely a write-up by someone else on Seeking Alpha.” Jamie: I think UPST is pretty universally liked by us on this board.
Saul Tenet 2: “Second, I would want rapid revenue growth. My ideas about that have become inflated in the last couple of years and where I once might have looked for 20% to 25% as very fast growth, I’m now looking for 35% growth, and usually more.” Jamie: Assuming high-end Target Revenue of $265m, my calculation shows that revenue growth is decelerating… or optimistically somewhere around the 20% min for Saul tenet #2. On the plus side, they had so much growth already and are launching 3 new products in 2022/2023 (Small Short-Term Loans, Business Loans, Home Mortgage Loans), but it’ll take awhile to realize those increases in revenue.
Saul Tenet 3a:“Third, I look for a stock in a special niche, with something special about it. I guess this could be considered a moat. It also could be considered a potential big future.” Jamie: Pretty universally agreed upon on this board that UPST has that x-factor.
Saul Tenet 3b: “Tied for Third, I look for recurrent revenue. I want my company to have last year’s revenue repeating this year and building from there, and not a company that has to go out and grow by selling the whole thing over again. God, this is important! It usually means software, and a SaaS model, and NOT selling things. You just can’t keep growing at 40% selling things. And when an economic slowdown hits, people will put off buying a new car, or a new house, but companies won’t tear out the software that keeps their company going. Software also usually means not capital intensive, and it also means high gross margins.” Jamie:
Recurrent Revenue: Unlike our SaaS companies, UPST is transactional-based so they need to keep doing more transactions every year to keep pace with a high growth rate.
Gross Margins: Contribution Margin declined from 52% (Q2) to 46% (Q3), which was less of a decline than expected. Seems like this will be a key metric to watch.
Thanks Saul and all for your knowledge! I hope to add to the community in the future.
Saul Tenet 2: “Second, I would want rapid revenue growth. My ideas about that have become inflated in the last couple of years and where I once might have looked for 20% to 25% as very fast growth, I’m now looking for 35% growth, and usually more.”
Jamie: Assuming high-end Target Revenue of $265m, my calculation shows that revenue growth is decelerating… or optimistically somewhere around the 20% min for Saul tenet #2. On the plus side, they had so much growth already and are launching 3 new products in 2022/2023 (Small Short-Term Loans, Business Loans, Home Mortgage Loans), but it’ll take awhile to realize those increases in revenue.
I think this has been done before in the forum, but seems relevant again…
From the earnings presentation, personal loans are about a 81B TAM - this year UPST is on pace to do $10-12B volume, and the prior year they did about $4B( with COVID qtrs, would have been more like $5B)
That volume translated into a FY20 revenue of $233M (or about 6% of GMV),and looking like it will be about 800M for FY21 (or about 6.6% of GMV)…
They have barely scratched the surface with auto loans (only refinancing) and that is a 672B TAM!!
so if FY 22 is the year of the auto loan for UPST, it could follow a path more like FY20 was for personal loans. Maybe UPST does half of what they can do in the personal loan space which was at 5% TAM penetration on FY20, and 12-15% on FY21.
FY21 they can add 16B in Auto loan transactional volume, and double that in FY22… and at 6% of GMV, revenue goes to 960M for the auto loan side of things alone…
JDR (long UPST, probably less tomorrow that it was today)