Uber Data Scientist Interview Experience — An In-Place Coding Question and a Deep ETA Case Study

Part 1: Coding

Given an integer array, move all zeros in the array to the front while keeping the relative order of all non-zero elements unchanged. Requirement: it must be done in-place.

Part 2: Case Study

Business scenario: Uber wants to evaluate the business value of lowering ETA (the estimated time for a driver to arrive at the pickup location).

• What metrics would you look at?
• From a business perspective, why is lowering ETA so important?
• Session conversion rate (similar to CTR — the conversion rate from a user opening the app to successfully placing a ride request): if you found a "positive correlation" between ETA and session conversion rate, how would you explain it? What confounding factors might be at play here?
• How would you design an experiment for this evaluation? (Switchback and synthetic control)
• After the experiment ends, if the confidence interval you get for the session conversion rate metric is [-5%, +1%], how would you interpret this result? And how would you improve the precision of the experiment?
• Suppose we're running a user-level A/B test, and now the PM asks you to do a deep-dive analysis on the user segment that "took >= 5 rides during the experiment period" — how would you respond and communicate about this request with the PM?