While many online lenders have turned to machine learning tools to help detect fraudulent applications, their reliance on internal data exclusively limits their efficiency. Our customer was using a rules-based approach for detecting potential application fraud, which works when the flow of applications is relatively low or stable but is easily overwhelmed. As a result, the company was losing almost $3 dollars for every $1 dollar tied to a fraudulent application, and nearly a quarter of its charge-offs were tied to application fraud.
Our customer connected their datasets to Explorium, and the results were immediate. Thanks to their new, enriched datasets, the lender generated new features for a model that could dynamically detect fraud, including:
By connecting to Explorium’s Enrichment Catalog, the lender was able to more precisely determine fraudulent applications and improve their risk assessment, lowering their lending costs and charge-offs by 15%, and boosting their fraud detection rate on applications to 92% of all incoming loan applications before moving on from the initial form submission.