When transaction monitoring rules outgrow their documentation

Fintech monitoring programmes in Taiwan often grow through weekend rule tweaks after a fraud spike or a peer incident. Six months later, the playbook still describes last year’s thresholds, while production runs a different set. That drift is one of the first things a thematic review exposes.

Inventory with owners

List every active scenario, its owner, last calibration date, and known false-positive pattern. If nobody can name an owner, treat the rule as orphaned until someone accepts it.

Re-perform a handful of scenarios

Pick rules that generate volume and rules that rarely fire. Re-running them against known historical cases shows whether documentation still predicts behaviour. Surprises here usually mean informal threshold edits that never made it into change control.

Fix the change log before the thresholds

Teams sometimes rush to retune numbers. Supervisors care as much about how changes are approved. Restoring a simple change log — date, rationale, approver — often reduces findings more than another round of silent recalibration.