Applying External Data in Consumer Credit Risk Management: Framework, Evaluation, and Joint Modeling

The presentation introduces the practical use of external data in consumer credit risk management, emphasizing three key points: understanding where data is applied, recognizing data characteristics and suitable scenarios, and evaluating and applying the data effectively.

It outlines the credit risk operating framework, describing business goals (scale, profit, risk, efficiency) and the four stages—acquisition, pre‑loan, in‑loan, and post‑loan—each with specific strategies such as channel selection, blacklist filtering, user segmentation, and pricing.

Data required for these stages is categorized into demand, performance risk, and performance capability, with examples ranging from demand frequency to credit history, income, assets, and liabilities.

The talk identifies three major challenges when introducing external data: selecting appropriate data sources, evaluating and applying the data, and enhancing the effectiveness of high‑quality data sources through joint modeling.

Evaluation of external data follows a six‑step process: defining the application scenario, preparing samples, obtaining historical data from third‑party providers, offline assessment, integration, and online experimentation. Key assessment criteria include compliance, stability, interpretability, effectiveness, and business benefit.

Joint modeling is presented as a solution to improve the impact of external data, especially when only scores are available or sample sizes are small. Techniques such as transfer learning (using conditional distribution alignment, MMD, EM) and clustering are discussed to address data drift, sample bias, and heterogeneous customer groups.

Finally, the importance of clustering is highlighted, explaining when and how to segment customers based on Y‑definition alignment, model performance gains, and differences in data richness across channels, concluding with a recommendation to balance model management costs against ROI improvements.