Upstart with Jeff Keltner | E234

Making credit more accessible through better credit modeling.

On today's Fintech Impact episode, Jason Pereira will talk to Jeff Keltner, Senior Vice President of Business Development at Upstart. It is an online lender that utilizes artificial intelligence throughout various parts of the underwriting life cycle in order to issue policies as quickly and effectively and with a better risk profile than conventional.

Episode Highlights

  • 0.32: Upstart is an AI lending marketplace where we power bank and credit union lending programs across a variety of products including unsecured consumer loans, auto refinance and purchase loans, says Jeff.

  • 1.03: The other key thing we work with our partners on is helping them acquire new customers. We not only serving their current customer base but being able to find new customers effectively and efficiently for their institution, says Jeff.

  • 1.39: Jeff did some research with Trans Union on the report that many more Americans are creditworthy and it really showed that more than 80% of the American public had never defaulted on a credit obligation.

  • 3.16: We run a marketplace for consumers at upstart.com where they could come and we will pair them with one of our bank or credit union partners depending on the risk profile, the geographic footprint, and things like that, says Jeff.

  • 4.02: The credit file has hundreds if not over 1000 pieces of information about a consumer and most lenders reduce that credit file down to four or five things.

  • 6.09: Jeff answers what are vendors incentive? It is to simply access to more market, or is it just tastes to be being able to increase the wheels on underwriting all the other?

  • 7.40: Jeff thinks that the degree of control one has over the underwriting model is really important. 

  • 8.02: Jeff says that they have got banks that are regulated by every major regulator and credit union. They have been through exams with these loans on the portfolio and most of them are driven by the desire to serve customers better.

  • 8.52: Before we started working with any of our lending partners, we went to the CFPB and asked the question like how you think about fair lending and how you think about testing and monitoring for that in the context of a model of non-traditional variables, says Jeff.

  • 10.11: Every consumer-oriented bank Jeff has talked to is mainly focused on "how I better serve the consumers that I am working with and that drives more than expected what they want and their willingness to take a little bit of risk."

  • 12.10: We do a lot of feature engineering where somebody list tells you that's interesting is how expensive things they have and compared to their income how much free cash flow might be left at the end of the month and what kind of payment they can afford to make, says Jeff. 

  • 13.18: In the era of online lending, sometimes people are applying for loans at multiple places at once, and you may not have the information about the latest loan taken out on the Bureau you got because they took it out the day after they applied for you.

  • 16.01: At the end of the day, it always comes down to cost distribution in almost any business, and that determines your potential market size for any product, says Jason. 

  • 19.49: We find that the model that public sector workers often are more creditworthy of their credit score, says Jeff.

  • 22.13: Jeff says that the hardest challenge to deal with was our macroeconomic disruptions. We saw one at start of COVID where you feel like your business is doing everything right and things are going well but there is a slowdown among capital markets or securitizations that ends up impacting you.

  • 23.05: One of the things we do to stay connected is we do our reviews every week and seeing the impact that these things may be on our consumers lives is really the thing that keeps us going, says Jeff.

3 Key Points

  1. Jeff talks about the entire consumer lifecycle compared to vendor life cycle and how they go about getting approved and how does it work differently with you than at elsewhere. 

  2. When someone submits the online form, and we are going to assess the risk and we are going to assume at this point that everything they told us is true. That may not be a good assumption, but for the risk model, we kind of says given these inputs, what's the risk, says Jeff.

  3. Jeff explains the difference between getting digital and really optimizing these processes in the productivity enhancements.

Tweetable Quotes

  • "The most online lending was kind of the digitization of the process, but not fundamentally changing how we thought about risk." - Jeff

  • "The way our technology works with our partners is each of our lenders specifies a credit policy, so they do have a minimum credit score to maximum debt to income." - Jeff 

  • "There are three things I hear when I talk to our lending partners. One is we want to get into the unsecured consumer loans. Second is even if I did this myself and could make the costs work, I know I am turning away a lot of good customers and third is I need a good asset to put my deposits to work." – Jeff

  • "Avoiding the downside at all costs is hugely a concern for most people because this is an industry that almost rewards people for saying no." – Jason

  • "We also use connectivity to the bank account to look at transaction history and that actually helps you identify three things at least." – Jeff

  • "When I applied for a loan in one of our first test subjects our bank partners asked me for a video selfie holding my license and telling why I wanted the money, because I triggered a whole bunch of fraud signals." - Jeff

Resources Mentioned

Fintech ImpactJason PereiraFintech