RiskSpan launches new credit model for non-QM loans

Non-qualified mortgage loans, by their nature, have risks not commonly seen in conforming underwriting.

To meet those needs, RiskSpan launched Credit Model 7.1 last week. 

It works alongside RiskSpan's existing non-QM prepayment model, which predicts how fast borrowers pay off their loans. Together, the two models let clients estimate both credit risk and prepayment risk for non-QM loans in one place, instead of piecing together tools from different vendors. 

A lot of the non-QM growth is coming from debt service coverage ratio loans and from what is known as "fumbo loans" which are jumbo loans being put through non-QM instead of other channels. As the non-QM market grows, older risk models are struggling to keep up. Many of the models used by lenders today were built for more traditional mortgage types.

How the model works

Credit Model 7.1 separates borrowers into four groups based on how they document their income when they got the loan. For each group, the model tracks how likely a borrower is to move between loan statuses alongside three broader economic indicators. 

"Non-QM borrower behavior varies meaningfully by documentation type, and generic credit and prepay frameworks simply don't capture that," says Divas Sanwal, head of modeling at RiskSpan. 

RiskSpan's model was trained on $87 billion of non-QM loan balances originated between January 2018 and August 2025. 

There is also an artificial intelligence feature that can automatically read and organize loan data as well. Clients can add the model's output into their own systems and a dashboard for testing the model's accuracy is coming soon. 

The launch comes as some industry voices are urging caution about how fast non-QM is growing. Non-QM RMBS issuance was up 81% through the first 5 months of 2026. Christopher Whalen, chairman of Whalen Global Advisors, was concerned that fast growth in a lending market has historically been an early warning sign of credit problems down the road.

Additionally, DSCR loans make up nearly a third of all non-QM. If interest rates keep rising, loans that seem safe today could not hold up according to investor expectations, according to Whalen. 

Forecasts from ratings agencies all point to continued growth in non-QM issuance through 2026. Surveys of mortgage brokers show non-QM is increasingly becoming a first choice for borrowers. Rising volume moving through brokers and originators is why the accuracy of the models sitting behind non-QM lending and investing decisions is likely to keep drawing attention.  

"Credit Model 7.1 was built from the ground up on non-QM collateral, segmented by doc type, and validated with published backtesting — giving risk teams, auditors, and counterparties the transparency they need to stand behind the model," said Sanwal.