Sramana Mitra: Tell me a bit about what kind of heuristics you are using in doing the credit scoring for subprime.
Jeff Zhou: One area that people are interested in is alternative data sources like social media. This is an area that I think is very difficult to tie to credit behavior when you’re comparing it to actual credit behavior.
I’m taking it in another direction. Let’s go back to the basics. The way that you manage the cash total in your account, how you stagger your bills, how much money you keep in your account at any given moment, and the patterns that you use have been the most interesting in terms of recreating the credit score.
Then I would say beyond that, that’s where the alternative stuff comes in. Can we use that to better understand your character and get contextual information about you that doesn’t exist specifically in the cash flow itself?
Sramana Mitra: You said 70 nonprofits are using your technology to do their credit scoring. How many such organizations are out there? What percentage market adoption do you have now?
Jeff Zhou: There’re over 20,000 nonprofits in the US. We are just a drop in the bucket.
Sramana Mitra: Is anybody else doing something like this?
Jeff Zhou: There are other companies that are doing pieces of what we do. Some companies are focused on the lending software piece. There are companies that are focused on the risk model piece. One of the big differences is that we’re looking at vertical integration of these things and trying to say, “This is turnkey Lending-as-a-Service.” It’s almost like Shopify for lending.
Sramana Mitra: The same 70 that are using your risk models are good prospects for your lending operation software.
Jeff Zhou: Exactly.
Sramana Mitra: The 20,000 are TAM for both of the pieces?
Jeff Zhou: Yes. I don’t think it stops there. So any sort of member organization, whether it’s a church or any sort of affiliate group that have members, they understand well.
Sramana Mitra: Brainstorm with me for a moment. We are in an unprecedented time of hardship and the subprime category has ballooned completely at this point. What do you see is going to happen and how can you help?
Jeff Zhou: The first thing that comes to mind is this idea of digital services. How can we recreate all of the pieces of a person-to-person financial services experience through digital pictures?
One of the most well-known things is that there’s just more fraud in online lending than in person. Any sort of improvement in better understanding the applicant is going to help financial services providers to open up more products to subprime applicants.
Sramana Mitra: As I’m talking to you, I’m thinking about what you have. You said that there are a number of nonprofits that are doing subprime lending. One of the levers to address the current situation is to increase the number of nonprofits and/or the amount of money that is being circulated as subprime credit.
The tricky part of that is, will your heuristics hold up because the subprime consumers are really in bad shape. It’s going to take them a while to recover. Would your heuristics approve loans for this battered category?
Jeff Zhou: That’s a great question. The way that our model is trained is all on sub-600 credit score. We’re taking people who have the minimum credit score ever seen in the system. We’re going to be the person to put our money where our mouth is and give you the loan. That data is going to help us build the next generation of the model.
That is also the same data that we would take into any conversation with a nonprofit or a potential provider. We’ll say, “We’ve been operating for the last four years. Here are the results.” Very few people are interested in buying a risk model that hasn’t been tested.