B.6 7 Pre-Experiment Checklist

1. Map out unit interactions and plausible spillover networks.
2. Collect high-resolution substitution logs or co-click data.
3. Generate clusters and validate using exposure metrics.
4. Design meta-experiment with ≥25% traffic allocated to Bernoulli randomization.
5. Write and register a pre-analysis plan (estimators, heterogeneity checks, etc.).
6. Monitor compliance; re-balance if significant imbalance arises.
7. Report bias estimates and effect sizes from both designs.
8. Discuss implications for external validity and platform-wide rollout.

---