By David LaRoche
Credit card issuers – whether they’re lending to a broad consumer base or to the customers of affinity partners – are facing challenges ranging from rising delinquencies, pressure from partners that don’t want to alienate their fans and customers with low approval rates, and a desire to mitigate risk.
For the past decade, the card lending community has faced increasingly tighter regulation, restrictive lending criteria, and continued economic challenges. More recently, they’ve felt pandemic and rising inflationary pressures that some believe could lead to a near-term recession. As a result, lenders are increasingly looking at revisiting their underwriting strategies.
Are you putting updates on hold?
Many of them, however, are so understaffed that they’ve had to put updates to their scoring models, segmentation strategies, and model documentation requirements on the back burner.
In some cases, they’ve turned to external data scientists to help them address their concerns before they become full-blown issues with regulators and internal finance teams. You may find that you can use what we have done for clients to assess your own situation.
Case Study: Top 10 Card Issuer
PAG worked with a top 10 card issuer with more than 30 co-brand partner portfolios and several branded portfolios. The client asked its small staff of risk-management analysts to revise the custom update to its portfolios and then asked us to take a broader look at its strategy.
The team took a two-stage approach, starting with developing a new custom score for underwriting. We then used the approved score to develop custom underwriting strategies for several of the client’s larger co-brand portfolios and two of its branded portfolios. Our analysts and data modelers worked with the bank to gather the needed data requirements from both internal bank data and external data that PAG supplied. Our team then developed a new custom underwriting score with six unique segments. That led to an increase in the Kolmogorov-Smirnov (KS) Statistic from 34 to 48 for the overall scorecard when rolled up. The client quickly approved the score and quickly coded it into the system and refreshed its underwriting strategies using the new custom score and the Vantage Score.
In the case outlined above, the new model drove higher approval rates across the targeted portfolios by 22%, while targeted loss rates decreased by 6% and profitability increased by 26% across those portfolios.
Having an adequate number of data sources is also a concern of many issuers we talk to. We worked with a client whose underwriting strategy for line assignments, risk-based pricing, and decisioning used generic data sources that had been in place for 24 months. The portfolio’s performance had been acceptable, but the client wanted to integrate more alternative data sources and use a retro study with several data providers. The client also wanted to reduce loss rates by 15% without impacting approval rates.
Try some different approaches
If you’re having similar issues, you could obtain retro data on previously reviewed applications (both approved and declined). The PAG team used two quarters’ worth of data and asked our data scientists to create a reject inference model to project the performance of declined applications to allow both swap-in and swap-out capabilities of the new strategy. We then used actual performance on the approved accounts and reject inferencing on the declined applications to create a new underwriting strategy, line assignment, and risk-based pricing approach.
As a result, our analysts and data modelers were able to identify and implement key areas of automation for the company with three new data sources for both credit and fraud risk targeting and create a model that was trending toward a 28% decrease in Lifetime Loss Rate within the first nine months – about 25% better than original projections.
In addition, approval rates came in significantly above the client’s approval rate and manual referral volume objectives.
Many issuers are so busy playing whack-a-mole with current application volume that they don’t have time to take the longer view. In other cases, they’re so used to “doing things the way they’ve always been done” that they don’t assign a fresh set of eyes to the problem. And that may be a mistake.
Dave LaRoche is director of business development for Predictive Analytics Group, which helps clients make better decisions, reduce regulatory risk, and optimize business performance through data storage and consolidation, creation of account-management strategies, actionable reporting, new marketing segmentation opportunities, scorecard builds/validation, and short-term staffing augmentation.
