Scenario

DoorDash uses third-party account partners to recruit and onboard new restaurants ("merchants"). Leadership wants to assess the program’s effectiveness and ROI using historical observational data (no randomized test).

Questions

(a) Success metrics: What metrics would you track to judge account-partner performance for merchant onboarding?

(b) Causal lift without an A/B test: How would you estimate incremental impact using observational data (e.g., difference-in-differences)? State assumptions and diagnostics.

(c) Selection bias and controls: If account partners mainly recruit seafood restaurants, how would you build a comparable control group (e.g., propensity-score matching, synthetic control)? Which covariates would you include and why?

(d) Spend–response modeling: How would you model the relationship between partner-acquisition spend and incremental revenue? Why might a logistic or saturation curve be more appropriate than a linear model?

Hints

• Use causal inference concepts: parallel trends, matching, covariate balance.
• Include geo, cuisine, merchant size/quality, seasonality, and local demand/supply proxies.
• Explain diminishing returns in spend–response curves and practical model estimation/validation.