Assessing the impact of the coronavirus pandemic on a customer’s credit standing is similar to a dean of admissions judging prospective college students: the more information they have, the faster a decision can be made.
Likewise, auto lenders’ decisioning processes ascribe values to certain components of a consumer’s credit history, aiming to efficiently and effectively evaluate who is in the best position to pay off an auto debt.
But the impact of the coronavirus pandemic on the consumer credit market has further complicated decision-making processes for lenders, who in recent years have been looking to reduce funding times through increased automation. Recession conditions triggered by the pandemic, however, have prompted lenders to pull back on borrowers with lower FICO scores and switch to more manual decisioning.
For lenders that might tighten up and buy fewer deals — shrinking auto portfolios in the process — experts say there are still ways to stay competitive. Lenders that have grown their portfolios and kept delinquencies down in this topsy-turvy economy have done so by increasing their focus on modeling possible scenarios and incorporating alternative data and artificial intelligence.
Flexibility in origination strategies is a key component of why lenders’ portfolios remain in positive territory during the pandemic. Modeling played a crucial role in Ford Motor Credit Co.’s historic third-quarter earnings results, CFO Brian Schaaf said on a Nov. 30 investor call with Bank of America.
“It’s a further testament to our underwriting and the performance of our portfolio,” Schaaf said. “From that standpoint, we’re encouraged that under any scenario, we planned appropriately.”
Modeling potential outcomes for auto portfolios takes substantial investment, leveraging staff and cash toward R&D. Nevertheless, frequent modeling is essential for running an auto lender in all business conditions, said Tom Schneider, Ford Credit chief risk officer. Stress-testing for a variety of scenarios is necessary to keep an auto portfolio healthy, but 2020 posed a unique challenge, he said.
Modeling generally is based on historical information, and “there’s not a lot of historical information for a pandemic,” Schneider told Automotive News via email. “In any type of situation where you don’t have a lot of historical data, you have to stress-test your models more.”
Lenders are spending more time stress-testing their decisioning processes during the pandemic than they would during a typical year — or even a typical recession, said Vladimir Kovacevic, co-founder and managing partner of Inovatec, a leading software provider to U.S. and Canadian financial institutions. Auto lenders begin to see trends in loan portfolio performance in 12- to 18-month cycles, but lenders should be adjusting their programs far more often.
Waiting that long could expose auto lenders to undue risk, or box-out business opportunities. But fast-paced changes brought on by the pandemic could mean that data incorporated into lenders’ decisioning processes will be useless in a few months.
“During a crisis like this, who’s to say that what was erratic or unpredictable behavior a year ago can be classified as a negative today?” Kovacevic said. “It doesn’t mean you’re all of a sudden a bigger risk. It’s just a function of what’s happening.”
Layering consumer data and utilizing artificial intelligence engines also are helping lenders. Partnerships with technology and software companies can augment these processes, Kovacevic said.
BMW Group Financial Services forged one such partnership with software provider Yellowbrick. The company’s data warehouse allows the lender to leverage historical data and run modeling scenarios on a much larger scale, according to Justin Kestelyn, vice president of product marketing at Yellowbrick.
Such data warehouse platforms act as an engine for business analytics, Kestelyn said, designed to help companies make these decisions much faster.
“There’s pretty much no limit to the kind of data you can bring into your data warehouse,” he said.
Other partnerships allow lenders to maximize existing data. Zest AI, formerly known as ZestFinance, works with about 10 auto lenders through its artificial intelligence software.
Mike de Vere, CEO of Zest AI, said lenders have much more data at their disposal than they actually use for decisioning. Zest’s tool allows banks to leverage more variables in the process, making the predictions based on that data more robust.
“It’s with data they already have, but applying better math, that you’re able to get these really significant results,” he said. “There is a way beyond just shrinking your credit box when the market starts having difficulties.”
Investing in alternative data is part of Westlake Financial Services’ strategy, said Kyle Dietrich, senior vice president of originations. The lender’s $11.5 billion auto portfolio is heavy in subprime loans, a group for which additional data is needed to lend with confidence, said Dietrich.
“This business is beyond the FICO,” he said. “Our alternative data spend is significant — in excess of $5 to $6 million a year — on data analysis outside of what the bureau does.”
Arivo Acceptance, a small auto lender, relies on trended credit data it gets through a partnership with credit bureau TransUnion. The tool, CreditVision, is a subscription-based service that provides consumer credit data that goes back farther than a traditional credit pull. Since signing up with the bureau’s service in June 2018, the lender has experienced a 40 percent capture rate increase without dropping the average interest rate on loans. The company also experienced a 30 percent decrease in decisioning turnaround.
Landon Starr, chief risk officer for Arivo, said using the tool to see a borrower’s credit history across several months shows patterns that help the company make quicker, more confident decisions on auto loans. A customer with a higher credit score, for example, could be more of a risk if there were negative trends in their credit background that a simple credit pull wouldn’t reveal.
“A lending environment with heightened unemployment rates and credit deterioration potential is something every risk officer is acutely aware of,” Starr said. “We are finding better-risk consumers in lower FICO bands. Our loss rates are less than half of our competitors.’ ”
Arivo, founded in 2017, has a $300 million auto portfolio across 13 states. At first, Starr said, the lender was more uptight about its loans. One of its rules? Only extending an auto loan to consumers who had already had one. Machine-learning models and layering consumer data over longer periods are changing that.
“We don’t have to use these antiquated rules and antiquated boxes. With all of the data possible, we try to determine if this person is going to be qualified for the loans,” he said. “We have the opportunity to bring more loans to people who deserve them.”