Objective

Validate a US Bank's model for detecting employees committing consumer-account and credit-card fraud for CCAR submission.

CRISIL's Solution

The CRISIL team performed a comprehensive validation of the model, a simple formula-based scoring model, to identify anomalies based on data grouping using expert judgment. We used the following process:
 

• Data Validation: Validated data sources for their relevance and timeliness, and assessed the quality checks made by the developer;
  
• Conceptual Soundness Check: The model development team was queried on various tested methodologies and assumptions, and assessed the responses qualitatively;
  
• Analysis of Tables and Results: The CRISIL team validated the tables and results provided in the model development document and performed backward computation to check the model's conceptual soundness assessment.

Client Impact

• The model was validated and the Internal Fraud Validation Document was completed within the stringent deadline for CCAR submission.
• The model was awarded a patent by the US Patents and Trademark Office.

Questions

Looking for high-end research and risk services? Reach out to us at: