FinTech Lender Credit Risk Modeling Methodology

This piece explores credit risk modeling practices in the fintech lending sector, tracing how marketplace lenders have evolved since platforms like Prosper Marketplace launched over a decade ago, including non-bank lenders such as LendingClub, SoFi, and OnDeck that partner with hedge funds and asset managers instead of taking deposits.

Regulatory scrutiny from bodies including the CFPB and OCC, and litigation such as Madden v. Midland, continue to shape how bank-partnership lending models operate under the National Bank Act.

Several modeling pitfalls are identified: embedding third-party scores like FICO within a custom model obscures true predictive power, seemingly innocuous variables like renter-versus-homeowner status can inadvertently discriminate against protected classes, and age- or state-based variables carry legal risk despite being predictive.

MaxDecisions recommends substituting problematic variables with alternatives such as credit history length instead of age, and using broader geographic categories to preserve predictive power while reducing compliance risk.