When A/B testing is not feasible, what causal inference techniques do you use to estimate the business impact of changes? Describe methods like difference-in-differences, propensity score matching, or regression discontinuity.

Sample answer preview

In many real business situations, randomized A/B tests are impractical or impossible. You cannot randomly assign some stores to receive a renovation and others not. You cannot randomly determine which customers see a price change when pricing varies by region or contract.