Design the sampling scheme for a 1-in-10,000 violation
Measurement & estimationHardProblem. You need a daily prevalence estimate for a policy where roughly 1 in 10,000 views is violating. You have a budget of 50,000 labels a day and a production risk model that scores content. A colleague proposes taking a simple random sample of 50,000 impressions. Walk me through the design.
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.