Decide and test Groupon program incrementality
Overview
This question evaluates experimental design, causal inference, uplift and profitability modeling, operational risk assessment, and KPI-driven decision rules within the Analytics & Experimentation domain for data scientists, testing both conceptual understanding and practical application of A/B testing, sample-size considerations, and guardrail metrics. It is commonly asked to determine whether a candidate can translate business constraints into a rigorous experiment and profit-oriented analysis, balancing incremental impact estimation, cannibalization concerns, and operational trade-offs to support go/no-go decisions.
Before signing, list the concrete factors and metrics you would evaluate to decide whether to participate in the coupon program, including but not limited to: capacity utilization vs 20 tables/day, expected incrementality (new vs cannibalized tables), average check lift/erosion, mix by daypart, repeat purchase rate of coupon users, service/ops risks, and any guardrail metrics. Then design a practical experiment to estimate the incremental profit impact: choose the unit of randomization (tables, customers, days, or geos), define treatment/control, primary KPI (daily profit), key secondary metrics, sample size/duration approach, and a quantitative decision rule. Finally, using the following numbers, derive the minimum share of coupon-using tables that must be incremental (not cannibalized from the 20-table baseline) for the program to be at least profit-neutral: baseline 20 tables/day at 100/day; with the program you observe 25 tables/day, 10 coupon tables, overall average check = $36/table, and a 40% commission applied only to coupon tables. State and justify any additional assumptions you make.