Measure Super Bowl ad impact with causal design

Suppose your app aired a 30-second national Super Bowl commercial. Create a measurement plan to estimate the incremental effect on installs and revenue. Be specific: (a) Define the primary and guardrail KPIs along with exact measurement windows (pre-period, game day broken down by quarter, post-period). (b) Propose at least two identification strategies since there is no clean randomized control—for example, geo-level difference-in-differences across DMAs with varying ad viewership, synthetic control built from similar apps or markets, matched markets, or marketing mix modeling with a spike regressor. (c) Write out the DiD estimating equation, state its assumptions (parallel trends, SUTVA, no differential shocks), and explain how you would test those assumptions. (d) Describe how you would address confounders such as concurrent promotions, outages, app store featuring, or seasonal traffic spikes during the game and halftime. (e) Show how you would estimate and report uncertainty, including confidence intervals from robust or cluster-robust standard errors, placebo tests, and randomization inference. (f) Explain how you would validate the result using falsification checks (pre-trend, 'placebo Super Bowl' dates) and heterogeneity analyses (new versus existing users, platforms). (g) If the true lift is a 4% increase in daily installs on game day, outline the required data and compute the approximate minimum sample size (markets and days) needed to detect it at 5% alpha and 80% power.

Overview: This question tests a data scientist's skill in causal inference and experiment design for marketing measurement, including identification strategies (such as DiD and synthetic controls), confounder adjustment, uncertainty quantification, validation checks, and power analysis.