Adverse-media screening for a private bank

The client is a Swiss private bank serving high-net-worth clients, where responsiveness in onboarding is part of the service promise and financial-crime obligations are strict. The two pull in opposite directions, and the bank has to satisfy both.

Their compliance analysts were highly skilled but overwhelmed by the volume and noise of manual adverse-media checks, which slowed onboarding and risked burying genuine red flags.

Manual adverse-media checks could not keep pace with onboarding, creating a backlog that slowed new client relationships in a service business built on responsiveness. Every delayed onboarding was a dent in the client experience the bank sells.

Crude keyword matching buried analysts in false positives: every common name returned pages of irrelevant hits to wade through, so most of the effort went into clearing noise rather than assessing real risk.

The bank needed far greater coverage without diluting the evidential quality its financial-crime obligations demand. More screening that produced weaker evidence would have traded one problem for a worse one.

We built an entity-resolution and relevance-scoring pipeline that distinguishes the real subject from look-alikes before scoring how genuinely adverse each match is. Resolving identity first is what kills the bulk of the false positives that crude matching produces.

Analysts see ranked, evidenced hits with the source article attached, rather than an undifferentiated keyword dump, so their time goes to judgement rather than triage. Suppressed look-alikes are logged too, so the screening remains fully defensible under review: nothing is silently dropped.

We calibrated thresholds against the bank's own historical alerts to hit the coverage-versus-noise balance their risk appetite required. Tuning to their actual case history, rather than a generic default, is what made the results trustworthy to their compliance function.