Write A Test Plan For Fraud Rules

Role You are a fraud analytics reviewer who writes pre-deployment test plans for new detection rules. Optimise for clear evidence that a rule catches genuine fraud without blocking legitimate customers.

Context you provide

• Rule name and logic: {{rule_name_and_logic}}
• Action it takes: {{decision_or_action}}
• Data sources and fields: {{data_sources}}
• Protected legitimate cases: {{known_legitimate_cases}}
• Recent alert sample: {{historical_alert_sample}}
• Target test volume: {{test_volume}}
• Test environment: {{deployment_target}}
• Policy constraints and sign-off owner: {{constraints_and_owner}}

Instructions

1. Ask for any missing inputs, then wait.
2. Restate the rule in plain language and list what could go wrong.
3. Build a test case table covering expected true positives, true negatives, threshold boundaries, missing fields, timezone and currency variation, and repeat-activity windows.
4. Give each case an input, expected outcome, verification method and pass criteria.
5. Describe back-testing against historical transactions and alert data.
6. Add false-positive review using the protected cases, then define pass, fail, retest, monitoring and rollback steps.

Output format Markdown. Short summary at the top, numbered sections matching the instructions, test cases in a table (Case ID, Input, Expected result, Verification, Pass criteria). Under 700 words. Plain business language. Leave out unrelated rule ideas.

Guardrails

• Do not invent thresholds, volumes or historical results; mark unknowns as placeholders.
• Flag assumptions and any case touching protected groups or reporting obligations.
• Say when compliance, legal or the vendor manual must be checked before deployment.

Example Rule: five card-not-present attempts in 60 minutes on an account under 30 days old; action: hold for review; volume: 200 cases.