Thoughts on UPST Q3


                    Q4-20      Q1-21      Q2-21      Q3-21
Loans transacted    123,396    169,750    286,864    362780
QoQ%                +52.5%     +37.6%     +69%       **+26.5%**

Avg loan size       $10121     $10185     $9743      $8628
QoQ%                           +0.63%     -4.34%     **-11.45%**
 
Origination vol     1249M      1729M      2795M      3130M
QoQ%                           +38.4%     +61.7%     **+12%** 

Total fee rev       84.4M      116.2M     187.3M     210.4M
QoQ%                +34.3%     +37.7%     +61.2%     **+12.35%** [Huge disconnect from loans transacted growth!]

Fee take rate%      6.76%      6.72%      6.70%      6.72%
 
Conversion rate     17.4%      22%        24.4%      23%
QoQ%                +14.5%     +26.4%     +10.9%     -5.74%

Inquiries           709,172    771,590    1,175,672  1,577,304***
QoQ%                +33.2%     +8.8%      +52.4%     +34.16%

Automated loan%     71%        71%        71%        **67%**

Contrib margin      49%        48%        52%        45.6%

***this figure is not including fraudulent inquiries

This was NOT a good quarter based upon my lofty expectations!
Loan sizes are down considerably which directly impacts fee revenue (loan transacted number grew 26.5% QoQ but fee revenue only grew 12.35%!). Although the new influx of lower FICO score borrowers means they tend to borrower smaller amounts, it’s clear the macro environment is not picking up faster than I thought, and that their marketing campaigns brought in less borrowers than I anticipated - as the absolute number of borrowers also came in far lower than I wanted.

My two immediate thoughts:

  1. My high expectations for Q3 of greater than 30% QoQ loan origination volume growth were heavily based upon my use of alternative/high frequency data from reviews/google trends/site traffic estimates. My modeling was flat out, 100%, completely wrong – blindsided by fraud attempts - something I was not smart enough to have thought about as a possibility at all!
    For example, the google search term “upstart login” increased by over 110% QoQ. But it turns out, fraud inquiries were nearly equal to the number of actual true inquiries this quarter, so thatcould account for much of the increase in popularity of that search term.

  2. Long term thesis for this company is intact. Upstart continues to be a fast growth first mover AI company in the credit industry with massive TAM and impressive cofounder led management.
    Clearly, insiders truly believe in the mission, as they have not done any additional selling of shares outside of automatic trading plans adopted in May and their largest shareholder (hedge fund Third Point) didn’t sell any since August.

BUT the question for me is: does UPST continue to qualify as a hypergrowth company that deserves an overweight spot in my portfolio?

Conference call highlights:

Auto refi: 2000 loans were completed in first half of 2021, although they only finished expanding it to 47 states by end of Q2. Now, another 2000 were done in Q3 for a total of 4000 YTD.
Ok, this is faster than 100% QoQ growth (off of a very small base) for auto refi, but still very small numbers. I am personally let down by that, although they might still be smoothing out the kinks?
Let’s suppose they grow 100% QoQ for auto refi throughout the next 5 quarters. Then they’ll have 4000 + 8000 + 16000 + 32000 + 64000. Which is 120000 total auto refi loans in 2022. Can they keep triple digit growth on this front?? It will probably be necessary, since average personal loan sequential growth may slow down further in 2022.
Of note, on the Q/A, it is mentioned that auto refi is very analogous to personal loans in terms of margins/unit economics.

Upstart Auto Retail: “tripled the number of dealers on our platform from a year ago…in Q3, added an average of more than 1 rooftop a day…just last week, the first Upstart powered loan was originated through our auto retail software…”
Ok, they look like they are moving as fast as possible with auto retail at the dealership.
On the Q/A, it is also mentioned that this is expected to have higher margins than auto refi due to less borrower acquisition costs involved.

Potential TAM expansion across FOUR fronts
1.) “Working on small dollar loan product, such as those who need a few hundred dollars and expected to be repaid in just a few months, to help financial inclusion for the underbanked…we aim to bring millions of marginalized consumers to our platform in the coming years so their lending partners can serve them with a host of affordable financial products over time…the interest in this small dollar product from our bank and credit union partners is off the chart, and we hope to bring it to market before end of 2022”

2.) “…there is an unmet need to provide…affordable installment loans to business owners…we aim to deliver the zero latency affordable credit solution that modern businesses require. This is in high demand from our bank and credit union partners and we hope to bring it to market during 2022 as well.”

3.) “The home mortgage market. It’s by the far the largest consumer lending category. We’ll begin to invest significantly throughout 2022.”

4.) Expanding their TAM to prime/above prime consumers. They’re no longer just limiting to below prime/near prime.
“Borrowers who use Upstart often see rapid improvements in their credit score and suddenly become very interesting to the entire consumer finance industry. Our bank partners want to serve these customers over their lifetime…by serving consumers across the entire credit spectrum…we’re beginning to deliver instant loan offers with no origination fees and rates as attractive as any on the market…”
And the CFO also said “we are expanding our footprint to serve more of the traditionally prime borrowers…as the proliferation of bank deposit funding on our platform enable us to be more rate competitive”

Conversion rate and fraud comments
Conversion rate fell QoQ from 24.4% to 23% and automation 71% to 67% for three reasons:

  1. “borrower segments newer to our models will initially tend to convert at lower rate than those segments which we have longer history…more conservative rates of instantaneous approval until greater loan volume for our models to train on”
  2. Volume of applications at top of funnel increased, so naturally conversion will drop
  3. A large coordinated effort to obtain loans fraudulently on platform… ”expect that episodes of this type will become increasingly common”. This impacted automation rate but not conversion rate because they removed fraud attempts from the calculation. However the fraud attempts were apparently immaterial to company performance.

Comments on macro
“We expect macro dynamics to ultimately lead to an increase in borrower loan demand, although that has yet to manifest in our results and remains upside to our forecast, as the exact timing is unclear”

Repeat borrowers
The call mentioned they had 315000 new borrowers, implying that about 48000 borrowers in the quarter were repeat customers. That’s a 13% proportion. My comment is that this proportion is higher than I would like at this stage; I wanted their personal loan growth to be expanding way faster than this (there should be way more new borrowers).

Bank partners
4 partners have now dropped FICO requirement. That’s very positive.
There are now 31 partners total, up from 25 in Q2. Fewer than I was hoping for Q3, but on Q/A, they reiterated confidence that bank partners “will continue to sign up at a rapid clip and there continues to be large demand”.

Summary of my thoughts from the call:
Long term, the company appears poised to do well and grow fast. TAM remains massive and they are working on more than just personal/auto with possibly business lending and small dollar products in 2022, then home lending in the future, and more immediately have expanded to above prime segments.

However, a sustained HYPERgrowth story is cloudier. While I believe they can grow, at minimum, ‘non-hypergrowth-fast’ for a long time, there are already many other great companies that can do that too.
As we know, UPST is not SaaS. Growth is expected to be lumpy, as management repeatedly states. I knew that, and I definitely would have been fine with sequential growth decelerating from the monster 60% QoQ Q2 results. But I was looking for at least 30% QoQ for Q3, instead of 12% QoQ!
Growth simply decelerated more than I expected. And for a non-SaaS company, a higher growth bar needs to be set, to justify overweighting.

I think the answer for me is I am definitely no longer overweighting heavily to UPST. I just needed to see core personal loan growth faster than today’s report. Since it is not, there appears a greater risk of heavier revenue growth slowdown in 2022.

So, I think keeping a huge position in UPST is more investing based on hope that:

  1. macro will improve faster than expected (this just can’t be forecasted),
  2. auto will grow quick enough to counter personal loan growth slow down (auto is still too unknown, and also adjacent LPRO earnings were disappointing today***),
  3. the new lines of TAM expansions will contribute meaningfully next year (they probably won’t, it’s just going to be too early and unknown unit-economics)

I did sell my entire position the instant I saw the loan volume numbers reported in after hours. While it’s disappointing to be now way off my peak YTD portfolio gains, UPST still contributed the vast majority of my current portfolio YTD growth of 228%, so I have no regrets for my >90% position before Q2 and Q3 reports as I truly believe I made my decisions based upon the data best available to me - even with the mistake of not thinking of fraud accounting for alt data.
It’s still very possible UPST will surprise on the upside again next couple quarters! Their guidance is interestingly positive (all things considered) for Q4. I will likely buy back some shares tomorrow to continue to keep an eye on the business.

***Adjacent competitor in auto lending is LPRO (Open Lending): their results today after hours were disappointing. They completed 49332 certs from 46408 certs which was only QoQ growth of 6.3%, and lowered guidance. The auto environment is not looking great, although yes, UPST is starting out from a negligible base.

250 Likes

Thank you John, some great thoughts, and I have massively appreciated your input on UPST.

For me, it’s quite astonishing the curve ball this earnings has brought. I had to read it 5 times when I saw the revenue number, I just couldn’t believe it would be that low. It’s caught all of us off guard considering the gushing positive news flow over the past three months.

But, the guide for Q4 for me is what will potentially keep my UPST position. That is some serious acceleration from Q3, which makes the possibility of a Zoom like slow down less likely, and adds weight to the train wreck of Q3 being predominantly down to this fraud issue. And if they are back on track with circa 20% QoQ growth, then 2022 100%+ revenue growth looks possible.

Management certainly tried to cover up this quarter with a lot of potentials, business loans, mortgages etc but that is all a long way off, even Auto feels like it’s going to take a fair bit of time to start moving the needle. So for me I need to decipher whether their core market still has enough serious gas in the tank to hit revenue around the $1.7bn mark next year. Some pondering to do.

This is their first big bump in the road, but can they recover ala Datadog?

32 Likes

I agree the QoQ fee revenue growth numbers are low, but i’m not as sure that UPST is still not in hypergrowth mode

Given the limitation for UPST’s growth currently is borrower demand, the fact that they increased the # inquiry volume by +34% QoQ would tend to indicate that they are still seeing substantial growth at the top of the sales funnel.

And if conversion rates and avg loan sizes improve, then we will see that reflected in UPST’s revenues.

For example:
(1) On conversion rate, i expect the current level to be temporary and would expect that to go higher again (which they have done since dropping in Q3toQ4 2018 and increasing until Q2 2021). As an upper bound for where UPST’s conversion rate can go, using QFIN US as an example (given the China e-lending/fintech space is more developed), in their 2019 F1 they had indicated a conversion rate in the mid 50% range.

(2) On avg loan sizes, the drop in avg loan size is concerning in that it seems due to the majority of the new customers being in a higher risk category and this may represent UPST having maximized their initial target market of “future prime consumers currently left out of the financial system” but these avg loan sizes approved should rise as UPST obtains more data and ‘proves’ out their AI algorithm (assuming we believe in UPST’s AI credit scoring engine)

Some separate points/questions:

  1. I am concerned that their growth in bank volumes has been muted as their % of loans originated held by banks ($ loans originated minus value of loans purchased for immediate resale) remains in the 20-25% range, with Q3 dropping to 20% from 25% in Q2. This doesn’t seem to fit their thesis that banks are increasingly using UPST’s originations to grow their loan books.

  2. On the auto loan refi growth, can the growth be alot more than 100% QoQ, given the main distribution channel (using CK, Nerdwallet, LendingTree etc) is the same as for their personal loan product so UPST would just be adding an additional product to their existing borrower ‘audience’?

These are just my thoughts and would love to hear your (and others) feedback on it.

15 Likes

I appreciate all of the info and posts, I believe the call was clear and most of your concerns were addressed in the call by the CEO/CFO and analysts appear to have been pleased with the answers provided.

I don’t see any meaningful weight on the fraud concern - that was very clearly communicated and addressed by Upstart’s management team and frankly to be expected with all of the noise surrounding the company. I also don’t think you can judge using Google trends on what searches were fraud and what legitimate, I think organize crime is more sophisticated than looking up Upstart login links using Google :slight_smile:

I will address two of your other concerns:

  1. Auto loans:

In the USA - there are NO cars to be sold by the dealerships! I’ve tried to buy a car in June/July, no chance. I’ve tried earlier in October - no chance. Every dealer has limited number of cars and they are marking them up like crazy. So whoever can afford to buy a car now is either super loaded and does not care about loans or rates, or desperate because their car broke down.

Please check this article confirming that new auto car sales are DECLINING:

https://www.edmunds.com/car-news/new-car-sales-are-plummetin…

In addition there is pent up demand building for cars, for people that are either extending their leases or delaying buying a car until the chip shortage gets under control. So I am guessing right now is good time for dealers to improve their processes and address potentially much higher demand in 2022/2023 for new/used cars with better margins and more buyers to sell to using the Upstart platform.

Upstart is going initially after the biggest dealerships so we can expect the loan demand to increase significantly/exponentially, and I don’t think we can use the model you are using given the above.

  1. Loan Sizes + Number of loans: With stimulus money and government benefits coming until last month or so - why would you need a BIG personal loan? This is a comment from the CFO in the call - I thought he speaks clearly on the subject

“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 true pretty much through the end of this past quarter. And we attribute that – or I guess we call that sort of suppressed loan demand that we largely attributed to the stimulus in the economy.”

So obviously we can expect things to change with no more government money, no more stimulus and people spending for the upcoming Holiday season. In addition what impacts the loan size is that if their models are adjusting due to high demand from new borrowers from new cohorts - I can expect that to change with time as well, especially when the new borrowers become repeat clients.

Another thought here is, if people are unable to buy a car, perhaps they can spend their saved down payment money for other things and that will offset their immediate demand to borrow money through personal loans or even credit cards.

67 Likes

With Upstart the train has left the station. It will be hard to enter into this business unless it is one of the FAANG.

Why ? Data. AI “models” become smarter as they have more data across.
A positive feedback cycle builds. More data → better models → more clients → more data…

It becomes no brainer for banks and other clients to sign up because it it hard to replicate what Upstart offers and is a readymade day 1 offering.

It helps banks grow their business (expand client base) and lower costs (minimal inhouse risk resources) and lower risks (proven AI models).
What’s not to like for the client ?

The only question is if the growth is lumpy or linear.
My guess is that it is and will remain lumpy. A few bank signings and the metrics will swing on the other side. These numbers need to be smoothed over several quarters.

  • I have a few Upstart shares but going to jump in today and accumulate some more.
17 Likes

My final thoughts on UPST’s Q3

What I learned/what’s changed/what’s new information to me
1.) Use of alternative data to track core business growth was a horrible idea. I was treating UPST like a SaaS stock by assuming I had visibility into business performance with alternative data. Before Q2 results, I saw alt data looked like it correlated very well with past UPST quarters all the way back to 2018.
So, when UPST Q2’s alt data looked awesome, I kept 90% allocation before August Q2 results, and that worked out great!!
I thought alt data indicated Q3 would also be a blow out. Obviously, was totally dead wrong, now my portfolio is 30% below its mid-October peak.

2.) Fraud attacks, something I never considered. It CAN happen again. This is a new uncertainty and future risk. UPST did a great job deflecting this as there was no ‘material impact’, but I think it DID divert resources from their AI/ML/engineering teams that could have gone towards positive updates to marketing/underwriting AI - I suspect now these resources are freed up, it’s allowing them to guide upwards for Q4.

3.) Repeat borrower proportion in Q3 was 13%. Too high, this early on in my opinion. Again, due to inability to acquire the number of new borrowers that I wanted – their personal loan growth decelerated way too much in one quarter – but this is offset by the fact they guided quite well for Q4…was this due to the fraud attack diverting their resources…? I think so, because their Sales/marketing expense only grew 22% QoQ (previously, it grew 43% to 54% QoQ for the past year!). I don’t think they spent as much as they wanted last quarter.

4.) Average loan balances fell significantly. This can happen if: decrease in proportion of high income borrowers (they tend to borrower larger sizes), increase in proportion of lower FICO prime borrowers (they tend to borrow smaller amounts).
So a drop with avg loan balances not necessarily a bad thing, I don’t think there’s anything wrong with acquiring more lower FICO ‘hidden prime’ borrowers, BUT if you don’t acquire enough of them, then of course the total revenue just won’t grow as fast as expected.

5.) Macro environment was said to be ‘not yet favorable’. A surprise to me!…was this a legit excuse by management? I think it would have been a definite cop-out reasoning had they failed to meet their guidance, but they didn’t actually miss Q3 guide…and they guided well for Q4

6.) Autorefi did another 2000 in Q3 alone, versus 2000 in first six months of 2021. This is likely growth of above 100% QoQ. They only finished rolling it out to 47 states by end of Q2. Is this fast enough?? I do think we need Q4 to see how this progresses.
(Also when you look at the 10Q filing, they state they carried $70.3M of auto loans as of Sept 30. Correlates with the 4000 YTD refi loan number).

7.) The first Prodigy originated UPST loan began last week. This is ahead of schedule.
8.) Expansion to prime/above prime personal loan market. Also four banks total have dropped FICO requirement. These are immediate positive developments.
9.) Expansion into small dollar loans and small business loans in 2022, and beginning to invest in mortgage category. Positive, but not going to be material for 2022.

What hasn’t changed:
-They are definitely not SaaS! Lack of recurring revenue and lumpy revenue growth. Exposed to macro effects of the world.
-Fast growth
-Asset light with minimal loans carried on balance sheet
-Already profitable
-Cofounder led, top tier management team, robust company culture.
-Secret sauce remains intact. Their AI produces default rates leagues ahead of all existing competitors.
-First mover advantage in AI lending, with first mover regulatory advantage (CFPB NAL)
-Massive potential TAM
-Disrupter status. With increasing network effects, as banks slowly sign on to partner. banks are conservative and I believe future partners will keep gravitating to the established AI player in the credit industry rather than other competitors.

What hasn’t ‘changed’, BUT we should now shine a spotlight upon.
Personal loan was thought to still grow at a torrid pace for quarters to come, but clearly, Q3 has injected fear that this may not be the case. And, if personal loan growth decelerates in 2022, can their entry into auto supplement it?

1.) Is Upstart quickly approaching their personal loan market share ceiling? My personal opinion, I don’t think so. Not yet. Like I said, I think Q3 was a fluke with the fraud attack diverting their resources.

We can also compare to LendingClub, who can’t expand outside of the prime/above prime market due to their inferior underwriting AI.
LendingClub’s record personal loan origination volume was around 2.9B in Q3. They have never before went above 3.0B in a quarter, and the company has existed for 15 years.
Meanwhile Upstart has already surpassed them in quarterly new borrower counts AND personal loan origination dollar volume (3.13B), despite a fraud attack and being 10 years younger…and UPST is guiding higher for Q4 while LendingClub is guiding flat next quarter.

To me this is evidence that it remains feasible that UPST can eventually claim up to 50% of the entire personal loan origination market (they currently have 15%).

2.) Supply chain issues could cause Prodigy to not grow as much as desired in 2022, even when coming from a negligible base.

A direct to competitor to UPST’s entry into auto market is Openlending. They had disappointing revenue/ guidance on their Q3 total auto loan certs at 6% QoQ growth (when you do the math, the number of purchase auto-loan certs DROPPED quarter-on-quarter. They were only ‘saved’ by growth of the autorefi certs).
On their call, they blamed supply chain woes causing fewer cars to be sold overall.
The saving grace was 30% of their certs comprised of autorefi (14800), up from 9282 in Q2, so it actually seems autorefi grew 59% QoQ (indicating that supply chain problems, of course, shouldn’t impact autorefi).

3.) So, it seems, supply chain problems shouldn’t cause near term problems with auto-refi growth…but why haven’t other competitors taken bigger share of the autorefi market in the past?

Let’s look again at LendingClub: they did $3.1B total loan originations in Q3 (composed of personal, auto, patient finance, education loans). If you calculate based on what they said in their Q3 call (100K new personal loan borrowers, 100K repeat personal loan borrowers at avg $14.5K loan size) then $2.9B is personal lending.
That means, $200M originated for everything else.
So, they did less than $200M autorefi originations in Q3, despite rolling this product out FIVE years ago in 2016. Assume average loan balance of $20K, and that’s at most 10,000 autorefi loans in Q3.
However, I believe much of this horrible growth over 5 years is due to their inability to expand past the prime/above prime market. The average auto loan rate for scores of 700 up is less than 5%, so the margin for above prime is thin for refinancing.
Since becoming a bank, with cheap depository capital, they DID grow autorefi 85% QoQ, per their conf call, which supports this idea that it was a likely this prime borrower/margin problem for them.

Going back to UPST: they don’t have the issue of being stuck with prime borrowers. By targeting below prime folks, there is a much fatter margin to entice customers for autorefi. They also have, most importantly, AI marketing advantages over its competitors and AI automated processes to smooth over the enormous friction involved (liens, titles, power of attorney, 50 different state requirements).
It’s very possible for UPST to grow autorefi next year in a meaningful manner, but no guarantees.

Plus, it’s expected that the macro tailwind for autorefi will continue (https://asreport.americanbanker.com/creditunions/news/whats-…)

Summary:
On Tuesday I sold everything upon seeing Q3 results, immediately sensing something went terribly wrong in the quarter and that a sustained hypergrowth thesis was broken. When I saw the loan origination volume was 3.1B and transaction numbers of 362K, there was a problem! $3.1B over second quarter’s 2.8B is like 10% growth. 10% sequentially for a year is a mere 40% YoY growth for a non-SaaS stock. That’s why I sold instantly.

Then after thinking about the above, going into Wednesday morning, I decided UPST is definitely still worthy of at least ‘regular’ 15% weight in my portfolio.

I think the fraud attack is mostly to blame for its less than expected Q3 growth (again, as evidenced by sales/marketing expense growing much slower than previous quarters), and that there remains large upside to personal loan hypergrowth to continue well into 2022, bolstered by the strong Q4 guide, plus a decent chance of meaningful auto-refi growth in 2022 to supplement this. Anything else (Prodigy revenue, small dollar loan/small business lending revenue etc) is icing on the cake.

Finally… for whatever its worth, Form 4 filings from yesterday showed that cofounder Paul Gu acted on the stock price falling this week.
He scooped up 25000 shares through RSU exercise on 11/9 at $313, and another 25000 shares at $256 on 11/10, and as usual, sold not a single share, despite the taxes he’ll have to pay.

And to avoid cluttering the board with a separate post:
Here’s from a NAFCU talk with Upstart’s Jeff Keltner, uploaded this week to their website but was recorded before Q3 results:
https://www.nafcu.org/education-certificationeducational-res…

[Host]: You talked earlier about you guys starting your technology in the unsecured lending spot, and now you guys have gone and started into the auto lending side or at least some engagement. How’s that been going?
[Keltner]: It took us years to get really good at personal loans. And I think we’ve dramatically accelerated that in auto loans, but we certainly see a tremendous opportunity. About an order of magnitude bigger than personal loans. Refi is a big play in personal loans, often credit card refi. Obviously, many lenders have had a lot of success with mortgage refi over the last number of years too. Auto refis however are like one percent of the market. And so we think there’s just a tremendous opportunity to go out there and improve the rates the way we do in other product categories.

[Host]: So what’s next after auto?
[Keltner]: Well, you know, I can’t say exactly, but we have we haven’t officially announced any products. But obviously, if you look at our technologies and where they make a lot of sense, obviously mortgage is a huge, largest category of consumer lending, so we’d be crazy not to be looking at how we could work there. And then I look at categories like small medium business or smaller dollar loans, shorter duration loans as places where the ability to really well understand risk, to automate the process, to remove costs.

If you think about smaller dollar loans, either SMB being smaller than commercial real estate or mortgage or really small dollar consumer loans, just you don’t make enough interest to pay for manual processing. And so we see an opportunity that if you can really take what we do to understand risk and increase approvals and lower the cost to originate through automation and kind of AI based processing, then you can take products that today banks often can’t issue in very large quantities because they’re just not profitable, and you can make them profitable. And I think those are the kinds of products that I hear a lot of demand for from from the credit unions.

171 Likes

10% sequentially for a year is a mere 40%

Hi Jon, it’s actually 46%. Just multiply it out. :grinning: You knew that!

Besides, 3.1 is actually up 10.7% from 2.8. That gives you a run rate of a little over 50%.

Which is all academic anyway because it was artificially suppressed this quarter anyway.

Best,

Saul

51 Likes

I think the fraud attack is mostly to blame for its less than expected Q3 growth (again, as evidenced by sales/marketing expense growing much slower than previous quarters),

I am not able to relate resources used for working on fraud prevention affecting the work that needs to be done on sales/marketing side. These are two different departments with different skill sets. So, what could cause slow down in sales/marketing expenses? Less sales is what I can think of.

6 Likes

I am not able to relate resources used for working on fraud prevention affecting the work that needs to be done on sales/marketing side. These are two different departments with different skill sets. So, what could cause slow down in sales/marketing expenses? Less sales is what I can think of.

While UPST acquires many borrowers via aggregators like Credit Karma, they do obtain half of their borrowers through other channels like direct mail or digital advertising. Sales/marketing expense includes spending on these aggregators and other channels.

The AI/ML team is intimately related to Sales/marketing and borrower acquisition.

My take is that in Q3, with fewer marketing campaigns being recommended by the AI/ML team due to their preoccupation with defending fraud, the sales and marketing team won’t know how to efficiently spend the next extra dollar in sales/advertising. This company is not going to indiscriminately advertise and waste money.

For example. How would they know who to send their next wave of direct mail offerings, without direction from the AI/ML team?

On a separate front, we already know their AI/ML team has a backlog of projects that provide positive changes to their AI underwriting/marketing etc models. If they are busy with fraud projects all quarter, they can’t push as many of those updates out.

Recall from what the CFO said:
https://discussion.fool.com/upst-cfo-fireside-chat-34929453.aspx…

Sanjay Datta: Sometimes I wish I were a CFO of predictable SaaS business. What we try to do is forecast innovation. Unfortunately that happens lumpy and not smoothly. The forecasting process fundamentally is we have a backlog or list of projects we think we want to work on, each with estimated impact and estimated timing. There’s obviously uncertainty around execution and the impact, so we apply very conservative discounted probabilities of execution and add it up. It results in guidance framework that errs on the side of being conservative. The execution is lumpy. We’ll have quarters where we’ll blow it out when a model single handedly jumps revenue 15%. Then quarters where we’ll meet our numbers if we didn’t get as big impact as we thought we would.

53 Likes