Transaction Fraud Detection
COVID-19 pandemic restrictions have shifted consumer buying behaviours to favour digital marketplaces and e-commerce, resulting in a significant increase in digital online payments.
The result of the increased online spending is a jump of online fraudulent transactions.
With the cost of fraud rising and cardholder trust declining, financial institutions need to take steps to ensure the protection of businesses and, in particular, their cardholders given the rise in online shopping.
Detecting and preventing fraud are paramount to meet consumer expectations and particularly relevant for online businesses and financial institutions, and they increasingly require AI and data-driven solutions to solve this problem.
Challenge
We were asked by a large financial service provider to find new ways of improving credit card fraud detection and providing insights of typical profiles related to fraudulent transactions.
Solution
AugustAï processed millions of transactions, and thousands of variables to automatically create a machine learning model in a matter of hours that picked up fraudulent transactions not detected by conventional industry tools.
We worked in a cross functional team that combined key stakeholders in the credit risk, fraud, product, technology, architecture and data teams. The project had visibility to the Board and Executive team.
Outcome
Fraud model was created and deployed in ~2 hours.
Provided the ability to understand the profiles and key drivers for fraud by:
Transaction volume
Location of transaction
Transaction type
Value
A fraud detection uplift of 10%, over and above the conventional detection system.
Complemented the conventional detection system with an additional 17% of credit card transaction value detected as fraud over 3 months.
Low volume of false positives, resulting in a minimal impact to OPEX regarding additional transaction volumes to process.
Significant reduction in losses incurred, saving OPEX.