Use CUPED to make a 2% effect detectable
Variance & ratio metricsHardProblem. You established that detecting a small lift on a team-randomized test needs an infeasible sample. Someone suggests using each team's behavior from before the experiment to reduce the noise. Explain how that works, how much it buys, and when it fails.
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.