Case Study: KeyBank achieves 70–80% reduction in non-value alerts with FICO Decision Tree Professional

FICO helps KeyBank "cut away the noise" to detect more wire fraud

KeyBank, a Cleveland‑based regional bank with branches in 12 states, was overwhelmed by an excessively noisy rules-based wire‑fraud detection system. A generic model and ad‑hoc rules produced as many as 7,000 alerts a month and a high false‑positive rate, straining about 30 front‑line analysts and making it hard to focus on real fraud.

In a two‑day engagement using FICO Analytic Modeler Decision Tree Professional, the team visualized risk by channel and customer type, redesigned more selective rules, and stratified strategies accordingly. The changes cut non‑value alert volume by 70–80%, improved analyst productivity, reduced false positives, increased fraud detection, and gave the fraud team greater confidence and clearer evidence for managers and auditors.

KeyBank

Michael Holloway

Senior Operations Analyst