Fraud rings sit in both arms. Is a user-level test still valid?
InterferenceHardProblem. A test randomized individual users between the current rules and a stricter new rule. It shows fraud that got through falling 32% in the treatment arm. A colleague points out that many frauds come from rings that share devices and bank accounts, and ring members ended up in both arms. Is the 32% trustworthy? If not, how would you redesign?
Before you reveal: say your answer out loud, as if you were in the real interview — get your reasoning across clearly first. There is no single correct answer: reading what the interviewer is really after and defending your own thinking is what makes an answer strong.