How is the auto loan industry leveraging alternate data to underwrite borrowers and lower the rate of interest?

There's a ton of companies using alternative data, like Upgrade and Upstart, who pride themselves to be using 1,500+ data points. Lendbuzz is also heavily using AI to target subprime customers, and they seem to be doing very well. So a similar thing is happening for auto fintech.

On the auto insurance side, credit scores are also being used, funnily enough. It was surprising for me, as an immigrant coming here, that credit is perceived as a judgment of character to a certain extent in the US. Even in insurance, in most states, maybe California, Hawaii, Michigan, and Massachusetts, they don’t require it. But as far as credit score, if we’re the same person with the same driving record, the same age, everything the same, but I have bad credit, and you have a great credit, I might be paying 70% more on car insurance than you. On top of the credit score, there’s the driving record and a lot of data providers that aggregate from all of that from the departments of motor vehicles across states and from the courts—everything from speeding tickets impacts your rate (Verisk, LexisNexis). Lately, the most sophisticated thing is telematics, which is heavily used in the UK now, and is also being increasingly used in the US.

But surprisingly, telematics, and all these additional advanced vehicle technologies—like the camera in the back and the bells and whistles you can add to the car to decrease probability of accidents, and track the customer's way of driving—they aren’t heavily decreasing the premium that consumers would be paying. It's merely shaving maybe 3% to 4% off the premium that the customer pays, and that's because these technologies increase the repair cost per accident.