Upstart (UPST) just had a stellar earnings report.
Q1 revenues: $121M versus $116M expected (about 5% beat)
Q2 revenue guidance raised from $117M to $155M
Full year 2021 guidance raised from $500M to $600M (!!)
The stock looks cheap for a company growing revenues at more than 100% YoY. Glad I loaded up on this, and bought more on the dip yesterday.
-Ron
Revenue of $121, up from $64 million was up 90% and up 39% sequentially!!!
Guidance for the year was raised from $500 million to $600 million!!! Or from up 115% to up 158%, from $233 million last year.
Guidance for next quarter was $155 million at the midpoint, up 28% sequentially, which they are sure to beat.
Contribution Margin, which I think is something like gross margin, was 48%, up from 38% last year.
Adjusted EBITDA was $21 million up from about $4 million same quarter last year.
Adj EPS was 22 cents. Their adj EPS for ALL of last year was just 23 cents.
Wow, what a report. You can’t say I didn’t tell you in advance! 
Saul
The stock looks cheap for a company growing revenues at more than 100% YoY. Glad I loaded up on this, and bought more on the dip yesterday.
Congratulations on your timing . I notice that UPST is up 17% in after hours. This is great. I also loaded up on UPST but that was May 3. So I am about even on the transaction. I thought about adding today but was so heavily committed I couldn’t do it. C’est la vie.
cheers
draj
UPST earnings was very impressive. One thing that jumped out at me and would like other’s opinion. In slide 6 they report $800M of vehicles sold through Prodigy in Q1. Given $626B in auto loan originations seems Prodigy has only 0.5% of the market. UPST did $1.7B of loan transaction volume and should do about $9B of loan transaction volume this year. So, it seems Prodigy’s existing loan business does not add a lot of loan volume.
Long UPST, 2ish%
Because of a slowdown in lending in the 2nd quarter of 2020, Upstart posted 17.4 million of total revenue Q2 21 guidance of $155 million will mean +790% YoY! How about that as a comp!
-HugoStocklitz (long UPST 16%)
A really great earnings. They are clearly going to beat $600m, so we could be looking at a company doing about $700m in 2021. A multiple of 20x would be fair if they can keep their growth north of 50% for 2022.
What I find really exciting is that the CFO/CEO have clearly communicated that they are still operating against headwinds, mainly the stimulus checks. From Q3/Q4 this should turn into a strong tailwind as more small loans are needed and there is more demand to spend in travel, leisure, home etc.
I am also becoming to see the CEO, Dave Girouard, as a visionary but also a great story teller and salesman ala Jeff Green, George Kurtz. I recommend everyone having a read of the earnings call, its littered with positivity and statements about how a lot of future growth lies ahead, which after all, is what we are all about…
“Despite softer loan demand from consumers due to government stimulus programs, almost 170,000 loans were transacted by our bank partners in Q1, more than double the volume from just two quarters ago. Not unusual to see fast growth in a fintech company, but it’s quite rare to see that growth paired with real profits”
“Our optimism for the future comes from two facts: first, we continue to have a likely backlog of projects that will improve our funnel throughput and lead to more growth in the future; and second, we don’t see competitors on the same AI-centric journey that Upstart is on”
“As I said last quarter, we believe Prodigy (Auto loans) will enable Upstart to tap into one of the world’s largest buy-now, pay-later market opportunities.”
I was one of the lucky ones to top up yesterday and made Upstart my biggest position but as I said at the start, I think end of year we’re looking at a lot bigger company so I don’t think anyone is late to the party here.
On the last question of the CC, the CFO seem to imply they were expecting just under 30% QoQ growth through the end of the year.
If they report 28% QoQ growth through the end of the year, they will end up with ~ $730 million in revenue or 213% growth vs the 114% they initially guided for.
It would look something like this:
03 06 09 12
Q1 Q2 Q3 Q4 FY
121,345.00 155,000.00 199,000.00 255,000.00 730,345.00
YOY growth 90% 793% 204% 194% **213%**
Seq growth 40% 28% 28% 28%
And it includes nothing from the Prodgiy acquisition. And 28% is the midpoint of next Q’s guidance.
Here’s the Q&A
Ron Josey – JMP Securities – Analyst
Thanks for the questions. Just we’re getting a lot of questions here on just the guidance. And so Sanjay, can you unpack a little bit more of the increase in the annual guidance and wondering if auto is just maybe a larger part there given, I think, Dave, you mentioned making fast progress overall, not just at Prodigy, but now in 33 states in general. So just any more insights, Sanjay, on the increase in annual guidance from the $500 million to $600 million would be great.
Sanjay Datta – Chief Financial Officer
Sure, Ron. I mean maybe just to put a point on it, it still includes really nothing meaningful from the auto side. We continue to be, I guess, you might call it an incubation mode there, developing the operations, expanding the Prodigy dealer footprint. So we’re still not counting on any meaningful contribution.
A lot of it boils down to the kinds of things that Dave talked about every single week. We are watching improvements to the credit models, improvements to the verification models, improvements to the acquisition-targeting models. And as those improve, as those drive our daily and weekly numbers, we propagate them through with some statistical confidence interval, if you will. And so we’ve had a very good quarter, obviously.
We’re obviously leaving the quarter on a good front. Our guidance for the next quarter is a little bit under 30% quarter over quarter, and we’re propagating that strength through to the end of the year.
What is the opinion here on the cyclical nature of the business? They’re obviously very dependent on the demand for loans and interest rates…Which is very much against the nature of our preferred SaaS companies that have that recurring, recession resistant revenue.
I have a small position because of the hypergrowth and reasonable valuation (but maybe the valuation is reasonable due to my concerns hereabove?) but I am not confident yet to make this a big position.
Secondly, the 7 year head start to potential competitors is nonsense to me. The marginal value of each additional customer diminishes with the number of customers. What’s the difference between data points from 1 million customers versus 10 million customers? The latter model has more information but it doesn’t really add that much additional valuable information. Once you reach a certain “data threshold”, the additional data no longer offers significant extra value.
Any comments?
Hi Rubenslash,
The more I read about Upstart the more I see that that Upstart is Creating a vertically focused SaaS Market. They’re providing the bases on which others take financial risk. At Tickertarget.com, Bert Hotchfield weighs heavily the benefits of this first mover advantage in an AI driven business model. My rudimentary understanding is that the more information added makes a better product. Being the First (Bank Agnostic)Mover, Upstart will likely continue to accrue ever greater amounts of information to feed into their AI product. Being a first mover in an AI driven vertically oriented cloud company creates an enviable bottom line because of near zero marginal costs for continued dominance(if being a horizontally oriented cloud company means there is a requirement for building out infrastructure).
Despite Upstart having greater customers concentration than I would like, almost every other criteria needed for me to make Upstart a full position is met (My criteria is outline here in this post https://discussion.fool.com/jason8217s-jan-port-summary-34738748…. As far as I can tell, Upstart has no AnnualRecurringRecenue/DollarBasedNetRetention of any kind yet. This and the high level of the customer concentration may be explained with how young and small Upstart is ; so, I’m keeping this somewhat under a full position. I’ll be looking closely at customer concentration going forward. And if Revenue growth picks up at either ZoomInfo or Docusign I’ll be trimming Upstart and adding to those due to my valuing ARR more than just revenue. Although it’s not difficult to imagine Upstart developing up sells to their current customers, as the company learns to apply its technology beyond unsecured loans and auto loans, to include appliance loans, mortgages, and so much more.
And as far as customer count, how many underwriters does a company need as long as there is a large market for what those underwriters are selling (the vast majority are sold to institutional investors), I’m more interested in the market for the securities these underwriter are selling into.
Best,
Jason
What is the opinion here on the cyclical nature of the business? They’re obviously very dependent on the demand for loans and interest rates…Which is very much against the nature of our preferred SaaS companies that have that recurring, recession resistant revenue.
I agree that increasing interest rates are going to hit UPST harder than other SaaS stocks. However, increasing interest rates are going to affect the valuations of all SaaS stocks. If the CPI number keeps printing like it did this morning, look out below.
I suspect that a lot of posters here are going to be shocked at how reasonable the valuations of these once high-flying SaaS stocks become by the end of the year if inflation keeps on roaring like it did today.
What’s the difference between data points from 1 million customers versus 10 million customers?
Each customer over a longer period is important. As an extreme example, 1 million customers over 12 months is way better than 12 million customers for 1 month.
GR
So, it seems Prodigy’s existing loan business does not add a lot of loan volume.
There were a couple of questions about this during the CC. Short story is their 2021 projections do not include significant if any contribution from Prodigy. They stated that they did not expect meaningful contribution from the auto business in 2021, and they were projecting 30% Q/Q growth to the end of the year…
cheers
draj
I was mistaken when I stated in my previous post that upstart had no Annualized Recurring Revenue.
On Upstarts S1, they state that they have three areas of revenue collection.
1.Upstart charges 0.5% to 1% of the loan amount to the loan Holder (banks making up 22% and The 100+ institutional Investores that Regularly fund the loans make up 76%)as an ongoing annualized servicing fee
2.Upstart charges Banks a referral fee of 3-4% f the loan principle amount.
Thanks for this great thread,
Jason
Once you reach a certain “data threshold”, the additional data no longer offers significant extra value.
Any comments?
Is this statement a conclusion based upon the nature of AI models? Or is it just an assumption based upon the supposition that at some point you have all the data you need.
Is there in fact a theoretical point of diminishing returns for training an AI model in general or a loan underwriting model in particular or does each set of new conditions offer opportunities for further selectivity refinements.
How did you arrive at your conclusion?.
draj
draj,
Is there a point of diminishing return for AI? The answer is yes, but it is not that simple. Look at Google search queries vs. say Microsoft or Yahoo! Anecdotal, but time and time again I am forced back to Google (despite hating to use them) because their search results are just that much better. Why are they better? Better modeling, sure. But also they have that much more data to ingest and fine tune their models with. This despite the fact Microsoft and Yahoo! each have billions of queries as well. Seems the diminishing return is not so much a thing in search.
However, diminishing return is a real thing in deep learning models. It is something project managers in AI plan for so as to not overtrain their models. Will this be an issue in the market that UPST is pioneering? Good enough is good enough? Maybe, but I don’t think it is that simple of a question to answer. Here is a quick technical article on it: https://clear.ml/blog/quantifying-diminishing-returns/
Bank of America, as an example, may be able to do it in-house as they have large volumes of data one would think. However, that would be internal to BoA and BoA will hardly have the focus and talent in staff to create the algorithms and running the proper data to make it so. But yes, theoretically, diminishing return is a thing in deep learning models.
History tells us, however, that this concern will not materialize as a material threat to the first mover in a market like this. Thus, I have been a buyer and not a seller of UPST. I will follow the numbers because I can access the scientific data in regard, but I am an outsider and when I get myself in trouble is when I try to over think things.
Tinker
History tells us, however, that this concern will not materialize as a material threat to the first mover in a market like this. Thus, I have been a buyer and not a seller of UPST. I will follow the numbers because I can access the scientific data in regard, but I am an outsider and when I get myself in trouble is when I try to over think things.
Tinker
Thank you for sharing your insights on the subject of AI models. I was commenting on an earlier post by Rubenslash who in turn questioned the assumption that Upstart’s head start in AI for loan underwriting was really not as great as some were claiming. He argued it seemed that once a bank or any other model builder reached a certain level of data input nothing further could be obtained and so therefore Upstarts advantage would be dissipated . That seem to me to be not quite true at this time because my impression is we ,or Upstart, despite their experience of 7 years is still in early stages of ultimate model development.
As you suggest there is no conclusive answer to the question of how much data is enough to train a model to its ultimate capabilities. Any such answer would be stated in complex mathematical terms different for each case. An infinite amount of data I would assume is more than enough. But for this type of underwriting model my intuitive feel is that we’ve not come near the upper limit of useful data. So Upstart’s head start and access to multiple bank sources makes it the the go-to purveyor of underwriting intelligence for now and for the foreseeable future…
So I’ve taken an 8% position in UPST and am debating whether or not to increase my stake. I have several ways to do that but for the moment I’ll just observe matters while the market continues to rattle along.
Your comment that history tells us that effective competition as others catch up is not likely to materialize in a case like this was enlightening. For me that is new and reassuring information and it also provides a bit of an answer to the questions I was concerned about.
cheers
draj
On the issue of diminishing returns, the CEO spoke to this on a CNBC interview in December. He sees both a wide lead for their model and a lot of room for improvement.
“One way to quantify it is…if you thought of a system that was completely random as 0 and a system that was omniscient at 100. Our view is most lending systems in the consumer world are at about 2, and we believe our system reasonably is at about 10. So we’re 4 to 8 times more accurate than a traditional lending system but we’re only 10% towards what is theoretically possible. There’s always some randomness in the world so you can’t get to 100% predictive but we are literally just scratching the surface of what’s possible with AI.”
“One way to quantify it is…if you thought of a system that was completely random as 0 and a system that was omniscient at 100. Our view is most lending systems in the consumer world are at about 2, and we believe our system reasonably is at about 10. So we’re 4 to 8 times more accurate than a traditional lending system but we’re only 10% towards what is theoretically possible. There’s always some randomness in the world so you can’t get to 100% predictive but we are literally just scratching the surface of what’s possible with AI.”
wheelzofsteel
Thank you for posting the quote. It really is an eye-opener. Imagine ! He said most lending systems are at a 2. I noted his remark during the Conf call that he though FICO scores were poor underwriting tools but 2 out of 100 is worse than poor. If what he says is accurate , and it is likely so, we won’t be worrying about banks or other competitors to UPST for quite a while.I’m also amazed to learn that he feels he is only 10% of the way there. Wow.
It also occurs to me that things change. Today’s predictive data points gradually (rapidly?) go out of style. Habits and mores change and new kinds of data appear which become better indicators for decision making. So in view of this insight I fully expect Upstart to keep and even expand its lead and strengthen its moat for quite a while yet going forward. ( Does anyone ever go backward?)
Perhaps I will raise my stake
cheers
draj
As a wise man once said, if there is one thing that is for certain then it is change. While we can be sure things change, things, people follow certain patterns. This is where AI comes in. It tries to learn from the past and constantly adapt to change. That is why FICO is useless today and it is also why early mover advantage is huge. Look how incredibly difficult it has been for anyone… anyone with theoretically far better data (mapping companies, tech companies, car companies etc) … to catch up to Tesla self driving technology. It is all about data, their accuracy, and past patterns… including patterns of change. I have zero concern that UPST will be caught up to by big banks. If anybody does catch up to them it will be companies like Facebook or Google. And if I were one of them, I’d just buy UPST outright. So I’m not worried about FB or GOOG competition either really.
FYI I believe the lockup on Upstart expires on 6/14, which may partly explain the recent weakness in the stock price. Hope this isn’t OT for the thread, but thought people might want to know.