Merchant Fraud Risk Score Development

3AI August 17, 2020

A leading payment processor

Problem Statement

• Modify existing fraud rules in the system to improve the performance
• Reduce false positives (Detection Noises) in merchant fraud detection model without compromising on adverse action rate

Analytics Led Approach

• Developed a score to identify potentially fraud merchants from the rule-based detections
• Identify a cut off for direct adverse action and another for risk based due diligence
• Implemented the scoring model and score-based rules in the rules engine, Blaze Advisor

Business Impact

• 15% reduction in review cost due to false positive reduction
• Another 5% review cost savings for direct adverse action recommendations
• A web-based monitoring system developed to track benefits of the strategy through monthly reports

Critical Success Factors

• Knowledge of fraud rules
• Expertise in various analytics techniques like decision tree, logistic regression etc
• Ability to identify fraudulent patterns based on historical data analysis
• Continuous monitoring and refinements of fraud rules for improved fraud detection over time