UPST CEO interview

Upstart CEO discusses balance sheet loans, auto future, insider selling and many more with Tom Gardner

Might only be accessible for paying subscribers of TMF

https://www.fool.com/investing/2022/06/30/upstart-holdings-c…

28 Likes

This sounds good. I like that he is not going into bunker/survival mentality. He shows lots of future thinking and planning. Things that are great for a growing company that wants to get through this time of wall street hate.

8 Likes

Thanks for this! It’s a great interview. Several things stood out to me.

1. Upstart sees itself as a “market maker.” They are not, and are not trying to become, a lender. They are helping the lending marketplace by giving lenders a far superior tool for risk assessment. Because of this, their TAM is enormous–once their pipeline is built out, every institution that does lending is a potential customer.

2. They simply “don’t see others building models similar to theirs.” For every moment this is the case, their moat widens, as the Godzilla-bytes of data pour into the maw of their eager-to-learn model.

3. Their main risk (according to Girouard) is whether or not they can execute. Their key risk has nothing to do with whether the economic cycle is boom or bust.

4. Dave Girouard is perplexed by the misunderstanding of their business. This is a nuance I’ve heard in every conference call when there has been something troubling to investors. He doesn’t see it coming and, therefore, doesn’t answer the concerns well. Sometimes he can’t even hear the real question because the question’s premise is based in a concern he believes is unrelated to their business or performance.

A key example is the hand-wringing by analysts who are unsure of how Upstart will perform in a recession. Just yesterday, in a Fintech hour on MF Live, there was the same concern, with the MF analysts saying something like “Well, we know how FICO does in a recessionary environment, but the AI model from Upstart hasn’t been tested yet.”

When Tom asked something similar in this interview, Dave Girouard pointed him to the same slide in the investor deck that Matt Frankel highlighted in yesterday’s show. (You can see it in this article: https://www.fool.com/investing/2022/05/13/are-upstarts-credi…) That slide is (to be fair to those questioning) less than clear. Okay, a lot less than clear. But it’s the chart that compares the default rates of loans based on FICO score with the same loans graded on Upstart’s A+ to E- scale. That chart, in my view, makes the recession question irrelevant. And that’s what I heard Girouard saying in this interview.

On both ends of the spectrum, Upstart’s scale would have served banks and consumers better. FICO above 700 had an average 3.4% default rate. But if you get more granular with Upstart’s AI scale, a much clearer picture emerges. An A+ from Upstart had an average 0.7% default rate across the entire range of FICO scores, from 700+ to below 639. The worst default rate of those receiving an A+ rating from Upstart was 1.2% (from those in the under FICO 639 category). So every single A+ loan, across the entire FICO range, outperformed the average 700+ FICO average by at least 2.2%.

But having a 700-plus FICO doesn’t earn you an automatic A+ from Upstart.

Specifically for those with a FICO above 700, but who got an E- rating from Upstart, the default rate was actually 9.2%. In fact, banks are underwater from an Upstart grade of C onwards, even just considering the 700 FICOs. FICO-only banks charged interest based on 3.4% for that entire cohort of loans, while the Upstart defaults for those same loans moved from 0.6% to 9.2% as you march from A+ to E-. That is a meaningful difference to a bank’s bottom line. And the 700+ borrowers with an Upstart A+ should have gotten much better rates.

Want to go to low FICO scores? Below a score of 639, the average default rate determined by banks is 7.7%. But if you were using Upstart’s model, those 639 FICO borrowers who Upstart graded as A+ had only a 1.2% default rate. So banks are leaving money on the table by either not offering or pricing people out of loans based only on FICO. And if Upstart deemed the 639’s an E-, their 7.7% assumption is losing money because the actual default rate for Upstart’s bottom tier there is 10.2%.

The chart shows that Upstart’s AI model is–already in its current iteration–better at predicting risk than FICO, which is why we keep seeing lenders using Upstart dropping FICO altogether.

Does lending change during a recession? Yes, it does. But the need to better assess risk does not. If anything, it’s even more important. Banks have a long history of knowing the percentage increase in default rates during a recession or during high inflation or whatever, boom or bust. They can just apply that percentage shift to Upstart’s scale instead of to their inferior FICO scale and bingo–they have better risk assessment no matter the macro economic environment.

Upstart’s model will learn new things with new variables in a recession. But that new data will enhance, not detract from, the effectiveness of the model they already have.

Saying FICO feels safe because we know how it does in a recession makes no sense to me, when Upstart has the data to show that FICO is a far more risky measure of loan defaults than their AI model. There will be more defaults in a recession, but that’s exactly why you would want a better model for predicting them!

5. Girouard says in this interview that “consumer demand is still super strong.” That is heartening. Will it slow as the higher interest rates filter down to consumers? Probably. And that would be a concern if Upstart already had full market penetration. But they have barely scratched the surface. At the rate they’re adding partners and developing new products, I see plenty of greenfield ahead.

It’s a lonely time to be an Upstart bull. And I think Dave Girouard has accurately stated their key risk. They absolutely must execute. But until I see signs that they are not, or until someone can explain to me why the well-established push and pull of default rates in recessions can’t be applied to Upstart’s model, I remain in for the long haul.

JabbokRiver

136 Likes

On both ends of the spectrum, Upstart’s scale would have served banks and consumers better. FICO above 700 had an average 3.4% default rate. But if you get more granular with Upstart’s AI scale, a much clearer picture emerges. An A+ from Upstart had an average 0.7% default rate across the entire range of FICO scores, from 700+ to below 639. The worst default rate of those receiving an A+ rating from Upstart was 1.2% (from those in the under FICO 639 category). So every single A+ loan, across the entire FICO range, outperformed the average 700+ FICO average by at least 2.2%.

I don’t think this comparison is entirely fair to FICO. Upstart can pick the borrowers within each grade, FICO cannot. FICO has no control over how lenders use the FICO scores to select borrowers, whereas Upstart has a total control. Here is an example. Upstart can assign A+ to someone whose FICO has temporarily dropped by 50-100 points to, say, 650(happens from time to time when someone closes a credit card account but during the same month has increased spending due to traveling, medical emergencies, etc.) but has otherwise maintained a healthy 700-750 FICO in the past. Upstart can use this case and say “See, this person got a poor FICO but we gave this person A+ and the person had never defaulted. We have a superior algorithm.”

The claim is true, but FICO is just a score, FICO cannot control how it is being used.

You can take the above example and ask wouldn’t this prove Upstart’s superiority? But my concern is why Upstart cannot get a major bank to be its partner? Why major banks, with its own data scientists and own models, could not find any values to partner with Upstart? FICO is not free, so why still stick with FICO with its supposedly inferior model? To me, this suggests that there are obstacles beyond AI adoption, obstacles like regulatory and compliance concerns that are preventing Upstart from growing further.

But maybe it just takes time to uproot something as entrenched as FICO.

FG

2 Likes

I don’t think this comparison is entirely fair to FICO. Upstart can pick the borrowers within each grade, FICO cannot. FICO has no control over how lenders use the FICO scores to select borrowers, whereas Upstart has a total control. Here is an example. Upstart can assign A+ to someone whose FICO has temporarily dropped by 50-100 points to, say, 650(happens from time to time when someone closes a credit card account but during the same month has increased spending due to traveling, medical emergencies, etc.) but has otherwise maintained a healthy 700-750 FICO in the past. Upstart can use this case and say “See, this person got a poor FICO but we gave this person A+ and the person had never defaulted. We have a superior algorithm.”

That just doesn’t make sense. UPST allows lenders to pick what grade of borrowers or how much risk they want to take. UPST grades them but does a better job of managing the risk of the borrower. UPST doesn’t tell the lender what to do they allow the lender to pick and choose how they want to use their algorithm. So yes it is a better way of grading borrowers and is superior to FICO.

The claim is true, but FICO is just a score, FICO cannot control how it is being used.

Either can UPST. It’s just a superior score.

You can take the above example and ask wouldn’t this prove Upstart’s superiority? But my concern is why Upstart cannot get a major bank to be its partner? Why major banks, with its own data scientists and own models, could not find any values to partner with Upstart? FICO is not free, so why still stick with FICO with its supposedly inferior model? To me, this suggests that there are obstacles beyond AI adoption, obstacles like regulatory and compliance concerns that are preventing Upstart from growing further.

That is how disruption works. It starts out small and builds itself up. But when they get their first large bank, Fico will be done.

Andy

33 Likes

I could be wrong but I think in either this or separate interviews upstart had mentioned that their business/growth model was to target credit unions and not large banks bc most large banks don’t see value in pursuing personal loans. So that could explain why no large banks have signed on. And sounds like they likely will not sign (or even be sold to) until proof of concept has been achieved with their auto loans, smb loans, or mortgage loan segments.

17 Likes

On large banks…

Jamie Dimon does not wake up every morning thinking, “I sure hope I can find a new disruptive technology to install at JPMorgan today!”

All institutions are resistant to change, and the larger the institution, the longer it will take to even get the ear of the right person, let alone convince them that they should unsettle their entire workforce by doing things differently. Continuing with the example of JPMorgan, as of 2021 they had just over 271,000 employees. Not all of them would be in lending, of course, but just the project of introducing Upstart’s service to the people, who know the people, who know the people, who can make a decision can take years. And then the decision-maker might say…“Well, that’s nice, but we need more data.”

Another hurdle is that it’s very hard for large institutions to even imagine that a better approach (to pretty much anything) exists. There’s a reason that change and paradigm shifts almost always come from the margins. Those central to the existing model can’t see, and often don’t care to see, how any kind of change can be beneficial. In a very large bank, the change might put the very people you’re trying to sell to out of a job. Blacksmiths were not rooting for the success of the Ford Motor Company, even though the latter talked about “horsepower.”

I’ve often wondered about the potential conflict of interest, or at least inherent bias, in having disruptive companies in Fintech rated by analysts whose jobs rely on the same industry the new companies are trying to disrupt.

Add in that banks, in particular, are not known for their forward thinking. When we hear about banks taking major risks, it’s usually in service to corruption, not innovation. And those that truly want to serve their customers are conservative by definition. Rule #1: Just don’t lose the money. The more assets a bank has, the higher the bar for any technology that wants to replace their legacy systems. A “major bank” will spend as much installing a new system, training their employees, and updating policies and manuals as smaller banks have in total assets.

A small credit union with 20 employees can turn on a dime and are hungry for any advantage they can get. They don’t survive otherwise. The “Too big to fail” crowd has no reason to rock the boat. And with the small institutions, if you call or walk into the institution, you can often talk to the decision maker on the spot.

I see the credit unions and smaller banks as a sort of phase 2 of Upstart’s R&D. Phase 1 is holding the loans on their own books to gain proof of product. That will be enough to sell a trial to a budding CEO trying to give their lending products the competitive edge. That small bank or credit union then partners with Upstart and tries it out. You’ll notice the announcements of new partners usually are not made until the partner takes the step after that–when the lender joins Upstart’s Referral Network. That transition seems to take three to six months or so.

A big bank is not going to get past their inherent resistance to change until Upstart’s model is so obviously better that the bottom-line savings of making the switch are at a scale that could make a meaningful difference on their balance sheets. The balance sheets of a JPMorgan and other large banks are so huge that just the loan numbers and dollar amounts that Upstart is able to carry on their books, won’t be enough. They will have to have data from a lot of other smaller lenders–lenders who are willing to speak up and refer (still larger) others. Then, somewhere in that process, a lender who is sold on what Upstart has done for them and who also has a relationship with a decision maker at one of the behemoth banks will get on board. Checkmate.

I totally agree with Andy: That is how disruption works. It starts out small and builds itself up. But when they get their first large bank, Fico will be done.

JabbokRiver

74 Likes

A big bank is not going to get past their inherent resistance to change until Upstart’s model is so obviously better that the bottom-line savings of making the switch are at a scale that could make a meaningful difference on their balance sheets.

Yep, and to tie that in to the meaningful change that Upstart has begun working with partners on their automobile refinance loans, this change should crack open the door to adoption by major banks in a way that their prior personal loan product did not.

Personal loans are not widely offered by banks, as they are relatively small dollar, unsecured loans with limited payoff to the bank. But, many banks offer automobile loans: safer because the loan is secured by the car as collateral, and typically larger dollar value. With the average price of a new car $47,000 currently, there are plenty of people out there carrying $20,000+ balances on existing loans that they might be able to re-fi at lower rates through Upstart.

The rising interest rate environment doesn’t help with that, but if Upstart determines that an individual is lower risk than their FICO score indicates, they might be able to offer a lot better rate than they initially qualified for.

Or if someone has paid down their loan significantly more than the car has depreciated, that would lower the risk of the refinanced loan and open the door to lower rates despite a rising overall rate environment, and Upstart’s low friction online portal makes that easier to manage than existing avenues to re-fi. The CEO said it’s been a major effort by them to stream-line the process.

After Upstart’s auto loan package is more widely adopted and proven a year or two, I think we’ll see some larger banks coming on board.

22 Likes

Jamie Dimon does not wake up every morning thinking, “I sure hope I can find a new disruptive technology to install at JPMorgan today!”

I appreciate the insights from the two different camps. On the one hand, there are posters who are saying big banks aren’t interested in personal loans. Then we have people saying Jamie Dimon is trying to protect his consumer lending business and is keeping Upstart down. Still both camps came away with the conclusion that Upstart is challenging the current paradigm. Looks like Upstart has the conviction of many and should be looked upon favorably.

I’ve often wondered about the potential conflict of interest, or at least inherent bias, in having disruptive companies in Fintech rated by analysts whose jobs rely on the same industry the new companies are trying to disrupt.

Analysts who cover companies are investment bankers. Those who work in consumer lending are retail bankers. They don’t interact (at work) even if they work at the same company. Besides, I thought we have read many positive analyses on Upstart written by analysts?

Personal loans are not widely offered by banks, as they are relatively small dollar, unsecured loans with limited payoff to the bank. But, many banks offer automobile loans: safer because the loan is secured by the car as collateral, and typically larger dollar value. With the average price of a new car $47,000 currently, there are plenty of people out there carrying $20,000+ balances on existing loans that they might be able to re-fi at lower rates through Upstart.

You can get a personal loan from Goldman Sachs, Wells Fargo, or Citi (not to mention also from Discover, PNC Bank, Fifth Third, Navy Fed, US Bank, Truist). I would say personal loans are widely offered. Also, loan sizes are usually up to $30-50k (or up to $100k from Lightstream).

I am just trying to provide some industry insights on the personal loan market and hopefully dispel the suspicion of conflict of interests. Please don’t take it as an attack on the company.

9 Likes

You can get a personal loan from Goldman Sachs, Wells Fargo, or Citi … . I would say personal loans are widely offered.

Thanks for the correction, FoolsGrad. I see there are a lot of sizable banks that offer them, though the two largest US banks do not offer personal loans: JPMorgan Chase and Bank of America.

1 Like