Your question is A/B Test for Fraud Rule. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
How would you design an A/B test to evaluate a new fraud-detection rule without severely impacting the user experience of legitimate members?
Explain the randomization unit, treatment exposure, primary success metric, guardrail metrics, sample-size and power approach, analysis method, and rollout or stopping criteria. Address how you would protect legitimate members, account for delayed fraud outcomes, and avoid invalid conclusions from repeated monitoring.