Metrics question

Uber is facing a challenge with the increase in cancelled rides especially during peak hours. As product manager, how would you solve this situation?

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Tests diagnosing a cancellation problem by separating rider-initiated from driver-initiated causes before proposing a fix.

Split cancellations by who initiates them, rider versus driver, since peak-hour causes and fixes differ substantially between the two.
Hypothesize rider-side causes: long ETAs from driver shortage during peak demand, or riders price-shopping and cancelling after seeing surge pricing.
Hypothesize driver-side causes: drivers cancelling low-fare or short trips in favor of waiting for a higher-value ride during high-demand periods.
Segment the data by city, time window and cancellation-initiator to confirm this is genuinely concentrated at peak hours and not evenly spread.
Propose targeted fixes: proactive driver incentives for short or low-fare trips during peak hours, and clearer ETA transparency for riders before they commit to a request.

Splits cancellations by initiator first, the most useful diagnostic bucket for this exact problem.
Names a specific, plausible driver-side behavior, cherry-picking higher-fare rides during peak demand, grounded in real driver incentive dynamics.
Proposes fixes tailored to each cause separately, rather than one generic blanket solution.

Treating all cancellations as one bucket without splitting by rider versus driver-initiated cause.
Proposing a fix, like a flat cancellation penalty, without first understanding why cancellations are actually happening.

How would you prevent a cancellation penalty from unfairly punishing riders who cancel for legitimate reasons?
How would you measure if your fix is working within the same peak-hour window?