We deployed a suite of statistical models: constituting Home Price Index (HPI) prediction model; probability of success post modification model; and probability of prepayment model. This suite was setup to create cash flow models for each possible resolution option.

We then calculated key financial metrics such as NPV. This was followed by the deployment of nonlinear optimization to identify optimal reduction in monthly payment, as well as the timing and amount of step-up in interest rates (for example, principal reductions as a discount upfront or as a reward for remaining current or as a balloon, etc.). This nonlinear optimization was done at 2 levels to find the best option for borrower:

Level 1: For a particular level of monthly payment, optimal offer structure was identified that reduces re-defaults and maximizes NPVs impact on re-default probability, pre-payment probability, investor constraints / guidelines adherence and feasibility given borrower’s affordability. This step also lent visibility into impact on cash flows and NPV as well as the incremental NPV over a foreclosure sale option.

Level 2: The above procedure was repeated for different levels of monthly payment (sometimes lower than the borrower’s affordable level). At every point a monthly payment curve was generated through an optimized combination of principal, interest and term vectors that maximize NPV.


  • Increased conversion rate by 10 times the industry standard with the inclusion of HAMP* with Trial-to-permanent modification


  • Achieved global optima by evaluating multiple points of monthly payment curve
  • Realized the current highest roll rate in the industry for subprime loans in both FRM and ARM
  • Introduced Shared Appreciation Model to the industry