When subprime auto lender Prestige Financial Services tightened lending standards about two years ago to manage losses incurred from customers defaulting on their loans, its lending portfolio shrunk.
Tighter standards meant fewer applications passed the company's underwriting policies, which led to fewer approvals.
To increase business, Prestige, a unit of Larry H. Miller Group of Cos., which also owns dealerships, replaced its legacy scoring techniques with software from Los Angeles-based ZestFinance six months ago.
"Across the board in subprime, you were seeing escalating losses, not just isolated to Prestige," said Steve Warnick, chief credit and analytics officer for Prestige. "We were letting fewer applicants in the door, and we wanted to find a solution to not continue to see escalating losses but also not lose market share as a result."
Since deploying Zest's Automated Machine Learning software, the Salt Lake City lender says it was able to bump up the number of new applicants 36 percent, from approximately 39,000 applications in January to about 54,000 in July. The company's approval rate grew 14 percent. On average, Prestige booked 1,989 auto loans per month in 2017, but in the second quarter of 2018, it booked 2,135 per month, a 7.3 percent increase.
Previous underwriting standards filtered applicants based on 10 to 15 characteristics. Using data from Prestige's portfolios, Zest's technology examines credit applications against 2,700 attributes to determine whether a loan is worth the risk. And more importantly, it can explain why.
Though Prestige, which works with 3,800 dealerships nationwide, says its own underwriting techniques predict loan performance more accurately than ZestFinance's software, Warnick says the legacy techniques are infinitely harder to explain, which is why Prestige can't use them.