Ask most fraud teams how their detection system is performing and the first number offered is alert volume — thousands reviewed this month, a rising trend line, a dashboard full of activity. It feels like evidence of a system working hard. It is not, on its own, evidence of a system working well.
The Metric Trap
Alert volume is popular precisely because it is easy to report: it goes up, it goes down, it fits on a slide. It says nothing, however, about whether the alerts generated are the right ones. A system tuned to flag everything remotely unusual will produce an impressive alert count and a mediocre fraud outcome, because the volume itself becomes the investigator's problem rather than the fraudster's.
What Actually Matters
Four numbers tell the real story, and none of them is alert count: the false-positive rate (how many flagged transactions turn out to be genuine), investigator productivity (how many cases a team can actually work through with confidence), and — the number that matters most to the business — actual prevented loss. A detection system that generates fewer, better-targeted alerts and prevents more loss is unambiguously the stronger system, even if its dashboard looks quieter.
A high number of alerts does not necessarily indicate an effective FRM operation.
- Alert volume measures activity, not effectiveness.
- False-positive rate and investigator throughput are better operational signals.
- Prevented loss — not alerts raised — is the metric that should reach the board.
- Detection effectiveness should ultimately be measured through risk outcomes, not alert counts. If your monthly fraud review leads with a volume number, ask what it's standing in for.