How do you reduce false positives without missing true matches?

Detailed Explanation

Mostly by fixing data, not by loosening thresholds. Name, address and identifier quality drive match accuracy more than the algorithm does — transliteration inconsistencies, missing dates of birth and free-text address fields generate far more noise than fuzzy matching itself. After that, thresholds are tuned to the institution's stated risk appetite rather than to a target alert count, and the tuning is documented. Lowering sensitivity to clear a backlog is a decision a regulator will ask about. Good-guy lists and whitelisting of previously cleared matches also help, provided each entry has a documented reason and is re-reviewed when the underlying list changes.