2019 Edelman Finalist Microsoft

Prospective Dynamic Fraud Control for Optimal Profitability in e-Commerce

Businesses and consumers purchase a diverse portfolio of physical and digital products and services online. The dynamic nature of shopping patterns and the adversarial posture of fraudsters make it challenging to stop fraud without interrupting legitimate customer purchases. In the U.S. alone, retail fraud amounts to tens of billions of dollars lost. Microsoft has tackled this problem by developing an innovative Fraud Detection System based on state-of-the-art AI, operations research, and automation. This new system employs a multi-stage decision-making paradigm that applies progressive machine learning (ML) models at each stage and co-optimizes these models across the decision chain to maximize profitability. Microsoft has thus dramatically reduced its Fraud Loss Rate resulting in $75 million annual savings and improved both its False Positive Rate and the Bank Acceptance Rate of legitimate purchases generating over $1 billion in additional revenue. This innovation is highly portable and Microsoft is working to make it available to its enterprise customers.