Amjad Ali

This chart shows the cumulative lift curve comparing the OriginRisk model against random selection. The gold line represents the model's performance: it plots what percentage of all delinquent loans are captured when reviewing a given percentage of the portfolio, ranked by predicted risk score. The red dashed line represents random guessing, where reviewing 10% of loans would catch roughly 10% of delinquencies. OriginRisk is a logistic regression model trained exclusively on origination time features such as credit score, loan to value ratio, debt to income ratio, interest rate, and loan age. No payment history or behavioral data is used, which ensures there is no data leakage from the target variable into the input features. The steeper the gold curve rises above the red baseline, the more effectively the model concentrates risk. In this portfolio, reviewing the top 20% of risk ranked loans captures approximately 59% of all delinquencies, nearly three times better than random selection.