Design a geo-holdout to measure a channel's true incremental revenue

Experiment designHard

Problem. You want the true incremental revenue from a paid-media channel that can't be cleanly randomized at the user level (no reliable identity). Design a geo-holdout experiment to measure it.

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.

MMM, attribution, or incrementality — which do you use when? →