UPST reported last night and announced a loan sale wherein the company “invested 40.2M in risk-sharing linked to credit performance.”

The biggest sign of distrust of a lender’s ability to project loss rates is the extent of the loss retention required by purchasers of its loans.

Management was, at best, unclear, and more likely, evasive about the accounting for the sales. (It looks to me like they took a 10% first loss position, which is ALL the loss, and for which UPST isn’t capitalized.)

But what is clear is that, given the number of vintages of loans that have underperformed loss expectations, UPST’s claim to having built a superior underwriting algorithm rings completely hollow.

The nonbank, noncredit analysts that cover this name don’t seem to understand that nothing matters more (not even the myriad other “headwinds” that management claims to be able to weather) than providing loan investors with yield reliability.

You don’t get to have vintages underperform and simply claim they were “R&D” and you fixed it. Superficial sell side analysts may give you a pass but not sophisticated loan buyers.