Design an email flu-shot experiment

Plan a complete email campaign to raise confirmed flu vaccinations within 30 days for eligible members. Specify (a) targeting and exclusion rules, (b) the randomization unit (individual, household, or clinic) and how you will contain spillover/interference, (c) holdout design (A/A, A/B, multi-arm) and seeding, (d) primary and guardrail metrics, how you will source ground-truth outcomes (e.g., claims/EHR, pharmacy feeds), and how you will treat missing outcomes when vaccinations happen outside our network, (e) power analysis: estimate the required sample size if baseline vaccination is 2.0% over the window and you expect a 20% relative uplift; use $\alpha=0.05$ and $\text{power}=0.8$ with an option for unequal allocation, (f) cadence/frequency capping and multivariate subject-line/creative testing that does not contaminate the main test, (g) controls for seasonality/holidays and overlapping campaigns, (h) compliance/privacy constraints (opt-outs, HIPAA) and what they imply analytically, and (i) the analysis plan (intention-to-treat vs per-protocol), heterogeneity reads, and a fallback causal method if randomization is partially broken.

Overview: The question assesses a Data Scientist's end-to-end experiment design and causal inference skills, covering A/B testing and randomization decisions, interference control, metric definition and power/sample-size calculation, outcome sourcing from claims/EHR/pharmacy feeds, cadence/MVT design, seasonality controls, and compliance/privacy considerations. It is frequently used in analytics and experimentation interviews because it checks both practical execution and conceptual understanding of field experiment planning, real-world data challenges (missing outcomes, claim lag), and analysis decisions (ITT vs per-protocol, multiple testing) within the Analytics & Experimentation domain.