Diagnose and reverse an adoption-rate decline

Over the past four calendar weeks, Google Meet’s enterprise adoption rate—defined as $\frac{\text{weekly active unique users}}{\text{total enabled users}}$—has dropped from 38% to 31%. You are the analyst on call.

(a) Give a precise definition of “adoption rate” and validate it: event definitions, de-duplication, bot filtering, identity across devices, time zones, and attribution windows. How would you backfill the metric and reconcile it with existing historical dashboards?

(b) Outline a 24-hour triage plan for separating instrumentation problems, seasonality, and genuine behavioral change; include concrete queries or checks, holdout benchmarks, and sanity ratios.

(c) Put forward at least three falsifiable hypotheses, broken down by customer tier, geography, client platform (web/iOS/Android), and feature usage (recording/meeting size). What evidence would support or refute each one?

(d) Design experiments aimed at reversing the decline (for example, an onboarding nudge, a performance fix, a reminder). Specify primary and secondary metrics, guardrails, MDE, power, sample size, duration, and a ramp plan; explain how you would select variants when traffic is limited.

(e) Show how you would distinguish causal impact from external shocks such as holidays or competitor launches using difference-in-differences, synthetic control, or interrupted time series. Which assumptions must hold, and how would you test them?

(f) Identify risks such as metric gaming, delayed conversions, and Simpson’s paradox. What monitoring and drill-downs would prevent false wins and ensure reproducibility?

Overview: This question assesses a data scientist’s ability in product analytics and experimentation, including precise metric definition and validation, instrumentation and event quality checks, rapid triage of adoption regressions, hypothesis generation, experiment design, and causal inference.