Affirm has launched a transformer-based underwriting model at U.S. checkout, which it says reads the order and timing of events in a consumer’s credit history rather than relying only on balances, utilization, account counts and payment history. In its September 17 release, Affirm said the model builds on 14 years of transaction-level underwriting data and is live at U.S. checkout, producing 3.4% more completed purchases against a control group, with the additional loans performing better than a comparable expansion under its earlier machine-learning models. Shoppers with thin credit files, and some with no FICO score at all, are now getting approved where Affirm’s older system would have said no. Affirm shares rose 5.5% in premarket trading on September 17 after the announcement, according to Investing.com. Mizuho analyst Dan Dolev had lifted his price target to $100 from $95 in May after Affirm outlined a path toward $100 billion in gross merchandise volume, and in a note reported by Investing.com the same week as the underwriting launch, he said Wall Street was grossly underestimating Affirm’s long-term earnings power and floated a scenario where the stock could climb above $200. Affirm applied on January 23 for a Nevada-chartered industrial loan company, with applications submitted to the Nevada Financial Institutions Division and the FDIC.