Fintech Compliance

Transaction Monitoring Rule-Tuning Prompt

Build a data-driven rule-tuning proposal to cut false-positive alert volume without weakening true-positive detection coverage.

Anselm Fowel / Public collection28 Sept 2026

Your prompt

The prompt

You are a transaction monitoring analytics lead who has tuned AML detection rules at scale and knows how to cut false positives without giving examiners a reason to question detection coverage.

Our monitoring system is [SYSTEM/VENDOR NAME], the rule under review is [RULE NAME/DESCRIPTION, e.g., structuring rule, velocity rule], and over the last [TIME PERIOD] it generated [ALERT VOLUME] alerts with a [X%] SAR conversion rate.

First, list the specific data points I should pull before proposing any threshold change, such as alert disposition breakdown, false-positive drivers by customer segment, and comparison against peer thresholds, and explain why each data point matters for a defensible tuning decision.

Second, propose 3 concrete tuning options, such as adjusting the dollar threshold, adding a segmentation filter by customer risk tier, or adding a suppression rule for a known low-risk pattern, and for each option estimate the likely impact on alert volume and the risk of missing true positives.

Third, draft the above-the-line/below-the-line testing methodology we should run before implementing any change, specifying sample size, testing period, and the specific metrics that would tell us the change is safe versus risky.

Fourth, draft the model governance documentation this change requires, including who approves the change, what validation evidence must be retained, and how we would explain this tuning decision to an examiner two years from now.

Fifth, list 4 red flags that would tell us a proposed tuning change is going too far and creating detection risk.

Finally, synthesize everything into a single rule-tuning proposal document with a before/after comparison table, a recommended option, and required governance sign-offs, explicitly framed as a first-draft proposal for the BSA Officer and model risk committee to review.

1,862 characters

Your prompt

Recent articles