Over-escalation rate
The share of participants who escalated alerts on customers whose explanation was documented in the account notes. There is no current industry figure for this.
AML Benchmark Annual Study
What AML teams actually do under pressure
We aggregate anonymised assessment results to measure what practitioners detect, what they miss, and when they escalate unnecessarily.
What the report covers
Account notes opened against escalation decisions
How often an innocent customer was escalated by a participant who never opened that customer's file. Recorded but not penalised in the score.
Detection rates by finding
Which planted findings are caught and which are consistently missed, across every case, broken down by sector, by discipline and by difficulty.
Score against certification status
Sector-assessment performance cross-tabulated with self-declared certification, reported whichever way the data falls.
Time taken and drop-off
Where participants slow down and where they stop, which indicates which judgments take longest under time pressure.
Data and privacy
What we record
- Your decision on each alert, including any you went back and changed
- Which account files you opened, and for how long
- Your written response and the time you took
- Role, sector, seniority, jurisdiction and certification status, which are optional and requested only after you have finished, so results can be compared by peer group
What we do not do
- Attach your name to a score in anything published
- Report your result to your employer, or sell it to anyone
- Publish any breakdown narrow enough to identify an individual
- Present a result as a professional qualification or a regulatory attestation
Individual transcripts are read only by the review panel, and only for scoring quality review. Percentile comparisons are withheld until the participant pool is large enough for them to mean anything.
The headline measure
Escalating every alert in the queue scores 15 out of 100, which is below the 25 scored by a participant who dismissed the whole queue without opening a file.
Over-escalation goes largely unexamined because it does not present as an error. Nobody is criticised for filing a disclosure that turned out not to be needed. The cost falls on analyst capacity spent on customers who were never a concern, on reporting volume that makes genuine disclosures harder to act on, and on customers exited over a risk profile their bank accepted at onboarding.
We have not found a published industry benchmark for over-escalation under controlled conditions. The annual study is intended to produce one.
The weekly case
One documented case each week. One decision. The answer and the source afterwards.