Case Study: Large US Financial Institution achieves 30% higher fraud detection and $4M+ annual savings with DataVisor's Unsupervised ML

Financial Institution Implements Unsupervised Machine Learning to Stop Application Fraud

Large US Financial Institution faced rising online application fraud—synthetic identities, stolen identities, and coordinated bust-out attacks—while rapidly scaling new account openings. Traditional rules, supervised models and manual review could not keep up, resulting in thousands of fraudulent accounts per month and many losses misclassified as credit losses. The bank engaged DataVisor, deploying its Unsupervised Machine Learning (UML) Engine alongside the Global Intelligence Network (GIN).

DataVisor’s UML Engine and GIN analyzed applications holistically to surface subtle, coordinated fraud patterns in real time, bypassing manual review. As a result, DataVisor increased detection by 30%, achieved 90% detection accuracy with a 1.3% false positive rate, captured fraud at least one day earlier, and delivered over $4M in annual fraud loss savings while identifying large fraud rings (200+ and 300+ accounts).