Quantify impact without an A/B test
Overview
This question evaluates a data scientist's competency in causal inference, quasi-experimental methods, time-series analysis, identification strategy selection, and rapid analytical design under tight time constraints.
Stripe-like scenario: You ship a change aimed at increasing payment success rate for APAC merchants, but randomization is infeasible. In ≤6 analyst‑hours over one week, design the analysis to estimate causal revenue lift. Specify: unit of analysis; data you need; identification strategy (choose one and justify: matched difference‑in‑differences, synthetic control, regression discontinuity, or interrupted time series); exact formulas for the estimator; how you will check parallel‑trends or model fit; how you will handle seasonality, merchant growth/selection, and holiday effects; power/MDES considerations given limited time; 2 robustness checks; and a clear decision rule (e.g., ship, roll back, or iterate). Outline 4 slide titles you would present onsite.