Credit & Risk

AI Can Help Auto Lenders Weather A Rocky 2019

Zest AI team

January 22, 2019

The auto industry is a conundrum right now. Ride-sharing is ubiquitous, yet vehicle sales are at record levels. APRs are rising — experts predict 6% might be the new normal — and that usually causes downward pressure on prices. Yet to consumers, prices are going up. The share of car payments of $600 or more is increasing, according to Cox Automotive. The average payment hit $533 in 2018. We have cheap gas, but cheap cars would be nicer.

In some parts of the industry, they’re already using the “r” word. Fitch’s index of subprime auto loan delinquencies of 60 days or more continue to hover at 5.5%, highs not seen since the Great Recession and up from 1.8% in April 2011. A passel of specialized lenders has already gone bust, torching the banks and private equity firms that funded their business. The common playbook in these situations is to pull up stakes and move to higher credit scores. That reinforces the cycle as more people can’t get the loans they need to buy cars. Perhaps this explains why dealers are pessimistic for the first time in a long time.

One option is to find growth in new ways, starting with better credit signals to unearth worthy buyers you might have overlooked. Machine learning-powered credit models can do that, finding more good borrowers and fewer bad ones. How? By using more data and better math to re-score risk across the entire credit spectrum. A lot of 620 FICO drivers perform like 690, a jump that could make the difference between affordable and not.

Prestige Financial Services, a $1.1 billion (assets) auto lender, used ZestFinance tools to build and deploy a machine-learning model that changed the way it evaluated loan applicants. The new model looks at 2,700 unique borrower characteristics, more than 100 times the 23 indicators Prestige had traditionally used to underwrite loans. Suddenly, it was picking up business everywhere.

Prestige more than doubled its lending to $55 million over the course of 2018, while reducing credit losses by one-third. By digging deeper into applicant data for credit signals, Prestige’s new machine learning models tripled the approval rates for thin-file applicants and improved the approval rate for millennials by 25%.

This is a path other auto lenders can follow. In an enigma of a market, a better flashlight is better.

Thank you for subscribing!
Something went wrong while submitting the form.