Upstart Q4 expectations/guesstimations

I’ll be the first to admit that trying to forecast Upstart’s revenue a quarter out seems daunting or foolish, given that others have tried and missed last quarter.

However, I’m going to give it a bash. Albeit with a generous sprinkling of assumptions, which may turn out completely wrong, of course. Still, I think we may have most of the pieces to make an educated guess. I’m going to discount auto entirely, and assume zero revenue contribution from that. Also, I’m only going to try to get to an estimate of revenue.

Let’s start out with what we know.


Rev$m	Q1	Q2	Q3	Q4
2018				24
2019	20	33	49	62.6
2020	64	17	65.4	86.7
2021	121	194	228.4

Number of loans originated

#Loans	Q1	Q2	Q3	Q4
2019	30,998	41,211	64,259	78,654
2020	84,214	11,873	80,893	123,396
2021	169,750	286,864	362,780

$Value of loans originated

$m	Q1	Q2	Q3	Q4
2019	369	486	825	1,044
2020	1,123	164	909	1,249
2021	1,729	2,795	3,130

That’s what they gave us and we know these numbers probably backward by now.

To get a better picture of what’s cooking, I calculate the average loan size, the average revenue Upstart generated per loan, and Upstart’s revenue as a % of the average loan size, per quarter. This is what that looks like:

Average loan size

Avln	Q1	Q2	Q3	Q4
2019	11,904	11,793	12,839	13,273
2020	13,335	13,813	11,237	10,122
2021	10,186	9,743	8,628

QoQ growth average loan size

	Q1	Q2	Q3	Q4
2019		-1%	9%	3%
2020	0%	4%	-19%	-10%
2021	1%	-4%	**-11%**

Upstart revenue per loan

Rev/ln	Q1	Q2	Q3	Q4
2019	645	801	763	796
2020	760	1432	808	703
2021	713	676	630

Revenue per loan as % of average loan

Rev%ln	Q1	Q2	Q3	Q4
2019	5%	7%	6%	6%
2020	6%	10%	**7%	7%**
**2021	7%	7%	7%**

My take-outs

A couple of things stand out for me here. First is that Upstart’s revenue per loan has stayed pretty stable at around 7% of the average loan for the last several quarters, with only 1 exception and that was the Covid quarter. I think that for purposes of Q4 we can therefore safely assume a similar level.

Secondly, and more importantly, the average loan size has decreased from $10,186 in Q1 and $9,743 in Q2 to its lowest point yet in all quarters of the past 3 years, to $8,628 in Q3.

Because revenue per loan stayed stable at around 7%, the decline in average loan size is therefore the key reason behind Upstart’s revenue growing by less that the loan volume growth in Q3. Loan volume growth was 26% qoq, but revenue growth was “only” 18% qoq, caused by them originating loans that were on average 11% smaller that the quarter prior to that.

Put another way, had the average loan size not decreased, but stayed stable from Q2, revenue for Q3 would have been around $245m for Q3 in stead of $228m.


Only one analyst asked about this - Michael Nicholas of Goldman. He asked:

“I was just wondering if you could talk a little bit about the loan sizes - average loan sizes in the quarter, and how we should think about that going forward into 4Q and then into 2022.”

The CFO’s answer was very thorough - he must have been expecting this question. His answer was, as others on the board have pointed out, basically that there were two things going on. He called one “optical” and the other “real” (both are real imo!) - so I guess he means the one is a bigger factor than the other - the “real” one being the biggie. These are the two things he called out:

  1. A mix shift of their total loans towards smaller loans as banks allow them to increasingly target the lower end of the credit spectrum by loosening their FICA guardrails. I.e. they are indeed generating smaller loans as a % of the total, because banks are increasingly allowing them to move to underwriting the lower end of the credit spectrum
  2. Actual loan size demand across the board coming down since the end of “last spring” ie from Q2 2020.

Indeed, the ave loan size in Q2 2020 was $13,813 vs $8,628 in Q3 2021. That’s a 37.5% reduction in average loan size in the last year.

In Q2 an analyst also asked about loan sizes and the answer was pretty much the same, but much shorter. In Q2 the CFO said:

“what that means is, we’re expanding downwards on the prime spectrum in terms of our capabilities. And of course, that will have an effect on loan size mix as well. And then we think that there’s probably some macro effects continuing in the economy, and those are a little bit harder to measure. But we have seen since certainly the onset of COVID, a pretty clear trend that I think is moderating, but it’s still sort of visible, which is that loan sizes and demand for loans in terms of what’s being requested has moderated downwards in the last 18 months.”

If there was one thing that they probably got wrong in Q2, it is the part I highlighted above. In Q2 they thought the decrease in loan size demand was moderating but in the end it didn’t moderate in Q3. Which is why his answer to a very short analyst question in Q3 was so thorough - it probably surprised them a bit (had it indeed moderated as he expected in Q2, then revenue would have been substantially higher in Q3).

Given how important this is for getting to an expectation for Q4, it’s worth looking at his response in Q3 about the “real” impact and what he expects going forward, in full:

”The real effect is that loan sizes in real terms have been coming down pretty consistently since pretty much the end of last spring. And that has been through – pretty much through the end of this past quarter. And we attribute that or I guess we call that the suppressed loan demand, and we originally attributed to the stimulus in the economy. And so we’re on a crystal ball as how this is going to propagate going forward, but I do think that as I called out in my remarks, I do think we’re starting to see the signs in terms of the macro economy of normalization in the economy with respect to savings rates, with respect to credit balances. And so if that is really accurate, what we would expect to see is a stabilization of loan size and then ultimately a gradual return to larger loan sizes as savings rates go down and credit bonds has come back up. So that is – I wouldn’t call that a prediction that’s maybe more of a thesis. That’s where we stand right now with respect to loan size.

Great. I’m going to put that into my expectations for Q4 then. A stabilisation of loan sizes. I don’t think that he will get this one wrong again (if he did indeed get it wrong in Q2, as I suspect).

So the assumptions for Q4 up to now are as follows:

  • Average loan size: $8,628 (stable - same as Q3)
  • Revenue per loan $630 (7% of average loan size, same as Q3)

All that’s left is coming up with an estimate of loan volumes, and then Bob’s our uncle to estimate revenue.


I’m starting with fish13’s analysis, where he splits out loan originations between Credit Karma and non-Credit Karma as part of his thoughts on diminishing concentration risk (here:…) , and going a bit further from there.

If we add the full-year split that Upstart give for originations from Credit Karma in their annual report of 58% for the full year 2020 to fish’s analysis, we can complete the quarterly picture, as follows:

**CK**	Q1	Q2	Q3	Q4
2020	40,423	6,660	44,947	82,188
2021	89,968	133,773	136,793
QoQ	Q1	Q2	Q3	Q4
2020		-84%	575%	83%
2021	9%	49%	**2%**

**nonCK**	Q1	Q2	Q3	Q4
2020	43,791	5,213	35,946	41,208
2021	79,783	153,091	225,988

QoQ	Q1	Q2	Q3	Q4
2020		-88%	590%	15%
2021	94%	92%	**48%**

From this I see an acceleration of non-CK originated loans at 48%qoq growth and a stabilisation of CK originated loans in Q3 at only 2% qoq.

I’m going to assume therefore that CK originations for Q4 will come in flat vs Q3 - zero growth, or 136,793, same as Q3.

And here is where I’m going to go out on a limb and make a key assumption, which may be true to a greater or lesser extent:

I’m going to assume that the number of bank partners of two quarters prior to the reporting quarter drive, or at least correlate with, to a large extent - the bulk of the non-CK loan originations.

Upstart had the following number of bank partners, from what I could gather:

2020 Q3: 10
2020 Q4: 12
2021 Q1: 18
2021 Q2: 25
2021 Q3: 31

Upstart further tells us that they expect their bank partners to incrementally contribute to loan origination as well as that getting a new parter up and running takes about 6-15 months. I’m going to go with 6 months (2 quarters), assume and calculate the correlation over the last 2 reported quarters (not before as the business was a very different animal back in 2020 vs 2021), and use the number of bank partners in Q2 of this year to forecast the non-CK loan originations for Q4:

Estimated average originations per bank partner per quarter:

Q2 2021: 153,091 non-CK loans from 12 bank partners in Q4 2020 = 12,758 per bank partner
Q3 2021: 225,988 non-CK loans from 18 bank partners in Q1 2021 = 12,555 per bank partner

→ Average for the two quarters of 12,656.

So for Q4: number of bank partners in Q2 2021: 25 x 12,656 per partner =
Q4 non-CK originations est: 316,405
then add
Q4 CK originations est: 136,793


Putting it all together gives my guesstimates for Q4:

Total # of loans: 453,198
Loan volume $m: $3,910
Revenue per loan: $630
Revenue: $285.5m

If Q4 Revenue came in at $285.5m that would constitute a 7% beat to their guide of $265m, a 25% qoq revenue growth vs 18% in Q3 and a 229% yoy Q4 growth.

Show me another one of “our” companies with anything similar to that…

Also, I see upside to this guesstimate. From auto, which I have not included at all, from the macro environment improving for credit origination and potentially leading to an increase in loan sizes (vs being flat as I’ve assumed), and an improvement in their models targeting higher FICO customers leading to higher loan sizes being included in the total mix, thereby potentially improving the average loan size.

Saul gave an estimate of $281m Revenue based on a 6% beat which is not far from my guesstimate above.

Would welcome any other takes on where they may end in Q4.

(Still pretty long UPST)



Great analysis!

I will add my current Q4 projection, although we know how poorly my Q3 forecast went!!

The only alt data measurement not invalidated by Q3 results is the Trustpilot review count.

As mentioned before, Trustpilot review mechanisms changed starting July 1.

So Q4 will be the first proving grounds for this metric.

In Q3
July: 2208
Aug: 1985
Sept: 1886

That’s 362780 loans divided by 6079 = 59.6776 loans per review

It means if we assume 100% of borrowers are getting invited for Trustpilot review, there is about 1.6757% response rate (including organic counts)

For Q4
Oct: 2193
Nov (to date): 1706

We are halfway through November. If we just double it as the run rate…we estimate 7798 total reviews for Q4.

That’s QoQ increase of 28%, at 465365 loans.

If we have loan sizes stay unchanged at $8628, we get $4.015B in dollar loan volume origination

and if take rate of dollar loan volume remains 6.7%, then total fee revenue becomes $269M

If we add interest income identical to Q3, we get $18M in interest income.

So total revenue in Q4 could be $287M, implying 8.3% beat on high end of guidance.


And here is where I’m going to go out on a limb and make a key assumption, which may be true to a greater or lesser extent:

I’m going to assume that the number of bank partners of two quarters prior to the reporting quarter drive, or at least correlate with, to a large extent - the bulk of the non-CK loan originations.

Great analysis, and thank you. I wonder if, rather than focusing on the number of bank partners, it might produce a better Upstart revenue estimate to focus on some measure of the size of those bank partners, such as total assets or some measure of each of their number of quarterly loans prior to and straddling their partnerships with Upstart. For instance, if newer bank partners were significantly larger than previous bank partners, and likely to produce a bigger bang for the bank partner buck, so to speak, that could affect Upstart revenues more than the simpler number-of-bank-partners metric. Just a thought.


Great Analysis.
I would like to add one comment regarding your most important assumption: I do not think it is possible to correlate non-CK revenue with the number of bank partners.
As we know the majority of loans is originated by CRB and Finwise. My understanding is that all traffic that Upstart can generate towards will end up with one of those two banks. And this traffic is growing with their AI based marketing campaigns. In a nutshell: Smaller bank partners only play a minor role in Upstarts revenue contribution. If Upstart is successful in gernerating traffic towards this will be much more meaningful.


“In October,<2021> our average used list price is $31,676, which is up 29 percent compared to a year earlier.

… current used car prices are 42 percent higher than they were in the first quarter of 2020, before the pandemic caused widespread closures and market disruptions. The price increase has helped drive an overall rise in the nation’s inflation rate.

… consumer searches for used cars are double what they were a year ago. That means the cars are being sold a lot faster… "


OK, so we can all agree that auto loans will add to this really fine analysis. The question is just how important will it be? I think it’s a mistake to focus on new and used car sales. Let me quote the following statement from Girouard during the CC:

“On the auto refinance front, we continue to make fast progress to eliminate the time and effort required to refinance a car loan. while the complexities of liens and titles, as well as a bewildering array of state-by-state processes, fees, and regulations, conspired to keep Americans trapped in their mis-priced car loans. We’re on track to repeat the final gains we experienced over the years in our personal loan product.”

Apparently, they are keenly focused on existing auto loans. Exactly how they are marketing their service with respect to refinancing auto loans is not clear to me, but then it doesn’t have to be clear to me in order to understand that there is a substantial market that is not dependent on selling a car. And if most of the extant car loans are “mis-priced” (meaning over-priced as compared to an Upstart auto loan) it’s an absolute no-brainer for just about everyone with a car loan to refinance. Especially considering that some 70% of Upstart loans are completed nearly instantaneously online without endless forms and requests for information. Quite possibly, more than 70% of refinanced auto loans would be fully automated in that the loan amount is already established and the loan is secured.


It may be worth noting that many growth oriented financials are sliding right along with UPST, including SQ, PYPL, V, MA, etc. The entire consumer financial sector was down near 3% the past week, and near 3.5% this past month.

This seems like a momentum shift. We may be witnessing another institutional rotation, this time out of finance and energy, which were big winners through most of the year.

It could mean a couple of things; maybe the COVID relief cash sloshing around on Main street is drying up, or consumers may be starting to pull back due to inflation, or I don’t know, maybe it is a coincidence.