Claims Fraud Detection for Auto Insurer

3AI August 15, 2020

Leading Auto Insurer in India

Problem Statement

• A scenario of an estimated 3% fraudulent/improper claims against an industry average of 1%.
• Objective was to enhance the upstream fraud detection capabilities not only at the claims level but also at the customer level

Analytics Led Approach

• Broad process steps which were followed :
  • Sensitive data encryption
  • Load the data from
  • multiple sources Logistic Regression Model creation
  • Data cleansing ; Data transformation; Business Significant variables deduction
  • Distribution analysis
  • Correlation analysis
  • Multi- dimensional Analysis
  • Fraud analysis through data profiling
  • Identify key suspect indicators which indicated possible provider frauds

Business Impact

• A proof of concept wherein Customers’ auto claims data for specific calendar year were analyzed and trends and patterns from analysis were presented to the business team
• Claims fraud detection helped to identify the potential losses due to fraud or suspect claims and thereby helped to reduce claims leakage to a great extent

Critical Success Factors

• Significant reduction in false positive
• 3.5% of 2010 claims identified as potential fraud
• 26 Crores of potential losses due to fraud