Jeff discusses subprime lending as the number of consumers needing loans escalate exponentially in this post-COVID world.
Sramana Mitra: Let’s start by having you introduce yourself as well as the company.
Jeff Zhou: I am one of the founders of Fig. Fig builds risk models and lending software for non-profits. Our two core technologies are cashflow-based underwriting and AI operations.
Sramana Mitra: Let’s separate those two and double-click. Explain what risk models you’re building. How does it work? What’s the outcome? What’s the impact?
Jeff Zhou: Starting with the risk models, the idea is that the traditional credit score is built off of a 7-year history in most cases. People change a lot in a 7-year span.
The whole concept behind it is to say that someone’s most recent bank account activity is a potentially better indicator of their current credit quality than the traditional FICO score. We build the analysis that scores bank account activity. It creates a risk score for that person that can be used to supplement a more dated credit score.
Sramana Mitra: Who are your clients? Are these banks?
Jeff Zhou: Our clients are of two types. The first type is non-profit organizations. We help them set up lending programs for consumers who have generally lower credit scores that excludes them from traditional credit services. The second type is the end consumers. We’re helping them get an opportunity to get a loan and also build credit.
Sramana Mitra: Is that a big segment? Nonprofits who lend to subprime creditors?
Jeff Zhou: That segment is relatively small. Before our technology, it’s really hard to be a lender. There’s legal compliance. There’s the management of the loan. There’s the credit reporting piece. There’re a lot of nuts and bolts that go into being a lender that most nonprofits are not equipped to handle. In the US, 20% to 30% of Americans have subprime credit scores. That’s a huge percentage of the population that could be served.
Sramana Mitra: How many nonprofits are doing this?
Jeff Zhou: Today we are working with 70 nonprofits in the United States. In our most recent program, we have about 7 in beta pilot.
Sramana Mitra: What is the second one? Can you explain the second service a bit?
Jeff Zhou: The second technology involves AI operations, It’s the idea that for lending operations, the cost to serve a consumer becomes a lot more important when your loan sizes get smaller. 10 minutes on the phone with a customer service rep is totally fine for a $20,000 loan. That becomes a huge portion of your cost if you only have a $500 loan.
Can we take technology and use probabilistic modeling? Can we guess the state of the customer and then take natural language processing and apply it to the message that they sent us? Can we generate a response at a high level of efficacy?
Sramana Mitra: Is it working?
Jeff Zhou: It is working today. Customer service reps use our software to serve over 20,000 loans. I think of it like an Iron Man suit of lending. The idea is there are a bunch of routine requests that come in. I want to change the payment date. This is something that we can use big data to craft all the different ways that someone could ask that question.
There’s a bunch of those today that are handled fully automatically. For the ones that have a lower probability of confidence, you have someone training the model, looking at the different responses, and then taking the one that makes the most sense. That data is fed back in.