You work on a streaming product where a new feature is being evaluated through an experiment, but the team also wants to compare feedback from a customer survey tied to the same launch. Some users may generate multiple sessions or responses, and leadership is unsure whether the right analysis unit is the user, session, or household.
How would you choose the right unit of analysis for the survey or experiment, and how would that choice affect the metric definition, statistical test, and interpretation of results?
