You work on a consumer or retailer experiment where the same shopper, household, or store may interact with the treatment more than once. The team is debating whether to analyze results at the impression, session, user, or store level, and there is concern that the wrong choice could bias the readout.