Case Study: Nubank reduces fraud losses and improves detection accuracy with TigerGraph

Nubank cuts fraud losses by $50M with TigerGraph

Nubank, a major digital financial services platform, was facing significant challenges with its existing fraud detection system. Their legacy infrastructure, primarily built on Google BigQuery, was failing to accurately identify fraud, resulting in approximately $1.8 million in monthly losses. The system had low precision and recall rates, generating excessive false positives and missing the majority of actual fraud cases across its 65 million accounts, all while needing to process over 50 million daily transactions in under 80 milliseconds.

To address this, Nubank implemented TigerGraph's high-speed graph database platform, TigerGraph Cloud. The solution integrated graph-native intelligence into their machine learning pipeline, calculating 30 powerful graph-based features to identify complex relational patterns and suspicious money flows. This implementation by TigerGraph resulted in $50 million in operational savings, protected 60 million households, and dramatically enhanced fraud detection accuracy with sub-80ms response times for real-time analysis.

Nubank