← all work

Watchlist alert filtering for a major national Spanish bank

sector
Banking / financial crime
outcome
-67% alert volume
PythonScikit-learnOracle

The problem

A major national Spanish bank’s watchlist filtering generated far more alerts than analysts could review, most of them false positives, burying genuine risk in noise.

What we shipped

A filtering layer that scores and suppresses low-risk alerts while preserving recall on true matches, integrated into the existing analyst workflow.

Outcome

Alert volume fell by 67 percent and analyst productivity tripled, with no loss of coverage on genuine matches.