26. You are testing a collaboration feature used by whole teams. Should you randomize individual users?
MediumStatistics & experiments
Technical
Show model answer
Individual assignment may contaminate treatment because teammates influence one another or share the same workspace. Clarify how the feature operates and consider randomizing teams or organizations so the unit matches the likely interference boundary.
Account for clustering when estimating sample size and uncertainty; many users in a few teams do not provide many independent units. Check assignment balance, eligibility and cross-team exposure before interpreting the result.
What interviewers listen for
- Match assignment to interference
- Cluster-aware analysis
- Check exposure contamination