Exploratory data analysis of accepted-loan datasets to evaluate default risk patterns across FICO bands, loan grade, and term. The goal is to replicate how a credit risk team monitors portfolio performance and identifies high-risk borrower segments for strategy adjustments.
- Default risk increased sharply when combining subprime FICO bands (<660) with 60-month terms, suggesting that restricting term length could mitigate portfolio losses in higher-risk segments.
- Loan grade analysis revealed overlapping default rates between certain adjacent grades (e.g., C and D), indicating potential misalignment between assigned grade and observed portfolio performance. This highlights a need for recalibration to ensure accurate pricing and capital allocation.
- Cohort analysis across vintages showed periods where default performance diverged from expectations, emphasizing the importance of ongoing portfolio monitoring and early-warning reviews.
- SQL and Python used for cohort segmentation, summary statistics, and visualizations.
- Trend charts and default curves developed to simulate portfolio monitoring dashboards.
This project demonstrates how early-warning risk signals can be quantified and communicated. Insights mirror how lending teams evaluate portfolio performance, diagnose deviations from expectations, and propose corrective actions to optimize risk-adjusted returns.