Design an experiment with marketplace network effects
Company: Uber
Role: Data Scientist
Category: Analytics & Experimentation
Difficulty: hard
Interview Round: Technical Screen
Uber plans to launch a new networked product in a two‑sided marketplace (riders and drivers). How would you design a causal experiment that accounts for interference/network effects? Specify: (a) the unit of randomization (user, driver, city, or geo‑cluster) and why; (b) how you’ll limit and measure spillovers (e.g., cluster randomization, geographic lift tests, partial‑interference assumptions, graph cuts/holdouts); (c) primary and guardrail metrics and how you’ll compute them; (d) power/sample‑size calculations under clustering (ICC, design effect) and expected duration; (e) bias‑reduction techniques (e.g., CUPED, pre‑period stratification) and novelty/washout handling; (f) diagnostics you’ll run to detect SUTVA violations and noncompliance; and (g) how the design changes if driver supply is effectively unlimited. Provide a concrete rollout plan and analysis outline.
Quick Answer: This question evaluates a candidate's competency in causal experiment design for networked two‑sided marketplaces, focusing on interference and spillover management, cluster randomization choices, metric definition, power/sample‑size calculations, bias‑reduction approaches, and diagnostic checks.