Lenders Find Creditworthy Borrowers Hiding in the Data

AI summary of the source article

With roughly 25 million U.S. adults lacking sufficient credit activity to generate standard credit scores, lenders are increasingly relying on bank transactions, cash flow, and detailed credit histories. Providers such as Personify Financial, Affirm, and Plaid report higher approval rates among underserved and thin-file segments without increasing default risks. Small-business lenders are similarly turning to accounting and invoice data through partnerships like Sage and Satago to overcome obstacles related to incomplete business records. Nonetheless, a Federal Reserve review notes that alternative-data models face potential challenges with data quality, costs, and a lack of testing across complete business cycles.

Why it matters

Alternative data sources and machine learning models enable financial institutions to expand access to credit for millions of unscoreable consumers and microbusinesses that are shut out by traditional credit scoring benchmarks.

Key facts

• Roughly 25 million U.S. adults lack enough recent credit activity to generate a usable score, according to the CFPB.
• Plaid reported its LendScore model delivered a 73% approval rate versus 65% for a traditional benchmark at equivalent risk.
• Affirm's transformer underwriting model generated 3.4% more completed purchases at comparable risk by approving previously declined applicants.