Fraud detection

Identifying questionable transactions to protect banks and consumers

Financial Services, Payments Network

In financial services, the volume of fraudulent transactions is outpacing the ability of institutions to recognise them.

Download PDF

Goal

Thanks to electronic banking, such transactions occur so quickly that financial institutions struggle to prevent losses when reconciling transactions at the end of each day. A national payments network recognised the need to better protect banks and consumers from fraudulent transactions and activity. It was looking to invest significantly in a new data business with a focus on two goals: detect at least 80 percent of fraudulent transactions and flag no more than 10 false-positive transactions daily in the process.

Insight and Action

QuantumBlack began by aggregating and analysing two years’ worth of payments data. Through this process, we identified patterns of behaviour—both fraudulent and otherwise. We developed an algorithm and model,enabling us to draw conclusions about questionable transactions.

With this information, we discovered a new capability to detect fraudulent transactions with incredible accuracy. We worked alongside the payments company’s in-house analytics, engineering, and business leadership teams to accelerate the adoption of new techniques—better protecting banks and consumers as quickly possible. Together we also developed a business strategy for bringing new data products to market.

Results

  • 90Percentaccuracy in detecting fraudulent transactions
  • 97Percentof days achieved its goal of flagging no more than 10 false-positive transactions
  • 18Percentimprovement in fraud detection methods