UPST: fireside chat with Abound CU partner

This post has my notes from UPST’s fireside chat with their recently new partner, Abound Credit Union, which was given to a credit union audience, and hosted by NAFCU.
Abound CU was announced as a partner in a press release on October 13, but has been a partner with UPST since July 2021.

https://www.nafcu.org/partnerlibrary/how-credit-union-leader…

BACKGROUND given by the Abound CFO and Chief Lending Officer:
Abound Credit Union was founded in 1950 by 10 civil service employees. Located in central Kentucky. In 2011 reached $1B assets. Today we have $2B assets. 18 branches. We are the largest CU in Kentucky. 115000 members.
Recently went through total rebranding from the name Fort Knox Federal CU to Abound CU in Feb 2020. Long standing tight affiliation and association of many members having a military background. Name change eliminates a perceptual barrier that you have to be military to join us.

Currently our loan portfolio is about 1.25B to 1.3B in size.
Loan portfolio breakdown: 45% in real estate. Indirect auto lending 25%. Branch based direct auto lending 11%. 17% other category (commercial, credit cards, personal loans etc).

WHY DID YOU DECIDE TO PARTNER WITH UPST?
It started from a discussion amongst our senior team in a weekly meeting. We saw our volume of unsecured personal loans wasn’t where we wanted it to be. We changed some things internally and tried targeted marketing and saw a nice increase but as we dug deeper, we saw companies like Upstart had a different model.

We could choose to develop things in house, or take a step back, and admit a third party could do it better. That’s where we landed with Upstart.
To be transparent, we talked to four different fintechs and each said they offered a similar product as Upstart. They all did personal loans, claimed to have AI tech, fast processes, good user experience.
But when we got to Upstart though, things clicked.
Upstart fit well with what we were looking to do: we want to help Kentuckians to really access more affordable personal loans.
As we dug through that, we had our risk, compliance, and legal teams get really deep into it. Compliance was so very impressed with the structure of Upstart! We really liked how we can tweak it to make the program our own and fit exactly with what we want to do and this helped push things through.
It was just a right mix of: right product, right delivery channels that uses tech that we can’t create in-house, and something else we don’t give enough attention to. It was a really really good cultural fit. Upstart’s team felt like an extension of the Abound team. When we know we have a cultural fit, we know we can make it a long term partnership in that we’re trying to achieve the exact same things.

QUESTION TO UPST’s Jeff Keltner:
I know you’ve onboarded a number of CUs this year. This wasn’t how Upstart began [your initial partnerships were only with banks].
What made you guys decide to focus on the CU industry?

We realized credit unions are under same financial pressure as other financial institutions: too much in deposits, lower loan demand, a hunger to implement technologies that drive loan demand, and CUs are noticing the consumer need and desire for personal loans.
I feel bad that we overlooked credit unions and this was a big miss when we began with personal lending. We underappreciated how much more consumer oriented credit unions are, and how much more consumer experience oriented they are – even more so than community banks.
A CU at the same asset size as a community bank will do way more consumer lending!
I noticed recently that whenever we post on linkedin about a new partnership, it’s usually a credit union. It’s exciting to see this momentum and uptick in CU partners.

We think our product brings a very sophisticated risk understanding – lenders fully control their risk appetite but we help them understand the risk of a given credit obligation to a given consumer very precisely; this is hard for most CUs to do the same for the asset category of unsecured loans.
We also help figure out the question of how do we also help find new members to your CU?
We get a pretty solid national flow of members looking for help for personal loans and increasingly, for the auto refi product.
Upstart can serve your current customers and at the same time find new members through a first rate interface, plus serve as many of your members as possible through a highly specific risk understanding.

QUESTION TO ABOUND: So, we had a good experience getting up and running together. We’re not the only fintech partners you’ve had. What’s your advice to CUs about working with fintechs to set themselves up to be ready to take advantage of opportunities that come up?
For us, we first had to make sure our members get a positive experience. We went to your website to see how the process is. We test drove it along with comparing to fintech competitors. I think Jake [Abound’s CFO] has 12 or 13 unsecured loans now!
We also think finding out that you can integrate your brand/logo in the process so members recognize it’s a credit union is important.
And administratively you want to make sure your board of directors is ok with AI partnership/risk analytics. Potential partners need a solid track record of portfolio performance that can quantify the risk.
Very tight net interest margins is a challenge for most CU right now, as mortgage rates, auto rates are low. The net yield of personal loans is much better. You need to understand what your ROI is.
We also prefer to service these loans as opportunities can arise for our marketing teams to offer more products to the new members.
You also want to be reasonable about expectations; setting policy limits that make sense to your overall portfolio/net worth for your CU, and that you can show regulators that you’re mindful of the risk, and limiting exposure until you get a track record going.

QUESTION TO ABOUND: What do you see as key priorities in lending/digital experience point of view to be competitive in next couple years?
We want to completely digitize and maximize convenience of lending for the do-it yourself members. Net interest margins are really tight right now. CUs need find other ways to differentiate since it’s hard to offer interest rates that really distinguish you from others. Need to find added value through increasing convenience and speed of services.

AUDIENCE QUESTIONS:
Can Upstart be placed on our website to serve existing customers?
We offer a completely white label all digital experience for any CU. There’s API integrations where we can take your customer info and preopulate a form.
We also have an Upstart referral network where we can channel nationwide demand for loans to target toward our CU partners that meet their membership criteria. It augments CUs by bringing non-members in while expanding your loan portfolio and your membership count.

HOW HAVE UPST LOANS PERFORMED through the pandemic? Does UPST use FICO scores to determine risk tolerances for AI lending?
UPST portfolio loans on all our partners have performed exceptionally well. Everyone in the industry did see much lower than expected levels of loss with government intervention in the pandemic, but if you just for example look at UPST hardships compared to other similar lenders, UPST saw much lower levels of impairments and higher levels of recovery despite having lower FICO averages than others. It turned out we were less riskier. The pandemic has been a great answer to the question of “you don’t know who’s swimming naked until the tide goes out" in that our models have held up really well.

We do use FICO, but minimally. We are data scientists in the purest sense for credit risk prediction. All data points are useful; we shouldn’t through anything away just because we don’t like it, but some data points are more useful than others. A FICO score is just a summary of info in a credit file. We find that each specific detailed piece of info in the credit file, is much much much more useful in determining credit risk than a three digit number that oversimplifies every variable in a summary.

Many of our lenders are actually moving to an environment where they do not have a specific credit score requirement in the credit policy anymore, where they allow Upstart AI models to leverage FICO data point all along with others to determine a risk level that no longer includes something like “FICO must be above 640 or whatever”, which is a smart move to allow UPST to serve more customers.

91 Likes

As always, John Wayne, thank you for your thorough research on UPST.

But YIKES I did not like this line:

To be transparent, we talked to four different Fintechs and each said they offered a similar product as Upstart.

And then: But when we got to Upstart, things clicked.

Nooo……what I wanted to hear was “When we got to Upstart, their numbers compared to the others blew us away.”

While I like cultural fit, I like numbers much more as a reason for choice. If there are serious contenders in this space, I want to know about them because my massive allocation to UPST is based on nobody seriously competing in this space yet.

So the very first thing I did when I read your post was go to Zest.ai because that’s the company I know is closest in AI/ML.

And unless I am mistaken, they have totally rebranded since the last time I went there. Gone is the dumb Bank of Turkey and other information that makes you think they have no cohesive strategy. It looks like they are now really going after credit unions, have a cohesive message, and have made a little traction. I did think they would ultimately do this but I thought it would come later.

When I say a little traction, I mean a little. They have a double banner on their screen showing their banks and CUs but if you look carefully, the second banner is just a repeat of the first banner, I imagine to make you think that they have more partners than they do.

But I think we need to keep an eye out for this company, as my initial research showed they really did have a similar offering re: AI/ML. And if there are other serious contenders, it does give me pause. I want to understand just how far ahead Upstart is from the pack, and I’m not sure I do anymore. I also want to understand Zest’s pricing to see if it is fundamentally more attractive, as well as their risk assessment numbers.

I would love it if some other people on this board gave me their thoughts. I know execution and the numbers are all that matters, so for now that’s what I’m waiting to see. But I really did think that the dominos would start to fall quickly once UPST reached critical mass, as it would be the one brand to draw the banks and CUs, and now I am not so sure.

I’ll be keeping an eye out for Zest. Also, if someone can contribute the other possible competitors, I’d like to keep an eye out for them, too.

31 Likes

I went to Zest.ai web site searching for Zest’s No Action Letter (NAL) from the CFPB. I did not see any mention of it. One would think that if Zest had a NAL, they would be heralding it.

This appears to be at least one advantage that UPST has over Zest.

FP

14 Likes

I want to understand just how far ahead Upstart is from the pack, and I’m not sure I do anymore. I also want to understand Zest’s pricing to see if it is fundamentally more attractive, as well as their risk assessment numbers.

Mizzmonika, I don’t know how the numbers stack up - no securitized loans from Zest AI exist to review.
But here is my comparison between Upstart and Zest AI:

UPSTART
-Credit underwriting AI to accurately price loan risk (payment flow prediction). This permits automated loan decisioning and reduces racial bias.
-Fraud detection AI.
-Prepayment prediction AI.
-Marketing/borrower acquisition AI
-Consumer facing/control of user experience. If the credit union chooses the white label service on their own website, UPST fully controls that application/loan processing experience. If credit union only wants to be in the UPST referral network, UPST again fully controls the borrower experience on upstart.com. This boosts the NPS scores for the credit union and UPST
-Branding power and reputation among consumers. It means future repeat borrowing and optionality (those who had positive interaction with UPST personal loans may also use UPST for autorefi, future home lending, etc)
-Provides servicing for borrowers, if credit union does not want to service. This means control over outreach. UPST uses AI to identify borrowers at-risk of delinquency and can talk to the at-risk borrowers to prevent defaults. It also increases UPST brand power.
-Has No Action Letter from CFPB
-Strong, clean culture and management could mean better fit for credit unions. Being a public company can also help.

ZEST AI
-Credit underwriting AI to accurately price loan risk (payment flow prediction or binary default prediction, like UPST??? Don’t know.) This permits automated loan decisioning and reduces racial bias.
-Fraud detection AI.
-Trustpilot reviews don’t exist for Zest AI for a reason. It is not consumer facing! Zero control over customer interaction with the partner credit union.
-Zero marketing or consumer branding power. Therefore does not actively try to bring in new credit union members from across the nation compared to UPST’s referral network.
-Does not provide servicing to borrowers. Cannot outreach to reduce delinquencies.
-Lacks CFPB No Action Letter.
-History of past predatory lending and fraud.

So which should a credit union choose?
Some have gone with Zest AI.
But,I suspect UPST’s national branding power/marketing prowess, ability to service loans, regulatory NAL, and superior culture/fit will give it continued advantages.
Either way the market is huge and there is plenty of room for competitors to grow unimpeded, for now.

73 Likes

But YIKES I did not like this line:

To be transparent, we talked to four different Fintechs and each said they offered a similar product as Upstart.

And then: But when we got to Upstart, things clicked.

Nooo……what I wanted to hear was “When we got to Upstart, their numbers compared to the others blew us away.”

I don’t think you need to be upset over that. Those other Fintechs said they offered a similar product. These were salesmen, they had to say that. Saying it didn’t make it so. They probably showed numbers that were similar, even if they needed to use smoke and mirrors to conjure them up. (Trying to sell an inferior product is a horrible way to make a living.)

7 Likes

Many of our lenders are actually moving to an environment where they do not have a specific credit score requirement in the credit policy anymore, where they allow Upstart AI models to leverage FICO data point all along with others to determine a risk level that no longer includes something like “FICO must be above 640 or whatever”, which is a smart move to allow UPST to serve more customers. [emphasis original]

I love where this is going. I moved from the left coast to a flyover state this year and signed up with a local CU-6 branches, $342M assets, 23.5k members. I am considering a used car loan and stopped in yesterday to ask how it’d work, what my rate might look like, and what info I’d need to provide. They went right to credit score, saying they only look at TransUnion, but that they don’t consider the FICO like a mortgage loan would. At the end of the short conversation, I asked whether they were looking at Upstart’s new AI-based program, and the answer I got was surprising. Basically not only no, but ‘heck no, we do things our own way in these parts’. I think a phrase she used was something like ‘if it’s not something we developed in-house, we’re probably not going very far with it.’ And this is a really tiny CU operation.

It doesn’t affect my short term view of that banking relationship or my likelihood of getting a loan with them (hypothetical terms were reasonable enough), in part because as a higher-earning, high-FICO type I’m not typical Upstart material. I don’t intend to price an Upstart loan for my own needs, especially after seeing the results posted in this forum from those who’ve done so. It also doesn’t affect my bull outlook for UPST, the stock. I just thought y’all might be interested in another ‘channel check’ data point. : )

-n8 (long)

20 Likes

Jon,

Excellent notes, as always.

Here are some extra research points comparing the two based off other info I could find:

Funding and Valuation
UPST was founded in 2013, raised $150M in private capital before going public in 2020 and today rocks a market cap of ~$25B.
Zest AI was founded in 2009, raised $250M in private capital since then, and according to Privco rocks a post-money valuation in the range of <$500M and an estimated 50-100 employees. This post-money valuation data could be wrong since they’re private, and I don’t want to start any discussions on UPST being over/undervalued, but let’s just summarize by saying that Zest AI today is essentially valued at a very, very small fraction of UPST despite having a few years of a head start.

Hiring
Zest AI has 10 job title openings listed on their website (with one or two titles like Account Executive which will hire multiple people under that one title).
Upstart has 110 job title openings listed on their website (with multiple titles that will hire multiple people).

Head Count & Growth
My LinkedIn Sales Navigator insights says that Zest AI currently has a headcount of 96 employees and has grown headcount by 14% in the last year, up only 5% in 2 years, with an average tenure of 1.8 years.
Upstart currently has a headcount of 1,112 employees, with 119% headcount growth in the last year and 249% growth in 2 years, with an average tenure of 0.8 years (likely due to lots of new hires, rather than turnover.)

While the securitized loan data would be helpful, I’m going to infer from this info that UPST just clearly has a superior product the market wants and firmly holds the first-mover and top-dog advantage between the two. UPST is sprinting away from Zest AI at an accelerating pace.

55 Likes

I’m going to infer from this info that UPST just clearly has a superior product the market wants and firmly holds the first-mover and top-dog advantage between the two. UPST is sprinting away from Zest AI at an accelerating pace.

Agreed. There is also another HUGE advantage that UPST has over Zest AI.

Management Team
Founders remain in charge at UPST. Listening to Dave Girouard and reading his GlassDoor reviews makes one want to work for this guy. We all love a Founder/CEO who has passion/vision to go along with significant skin in the game.

Zest’s CEO, Mike de Vere, was installed by Zest’s private equity investors. For me, listening to him on a podcast was totally uninspiring. Losing a founder can also have a devastating impact to a private company’s cap table. If that long-gone founder still has a significant ownership, that is a red flag for future fund raising. Plus the new CEO will command lots of stock options, which equals dilution. This could make Zest’s access to new capital that much more difficult.

–John
Long UPST

15 Likes