Estimation question

Estimate the number of incidents that could be reported by users of Waymo in the SF Bay area.

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

Estimation with a clear funnel and reasonable assumptions for a niche, real world autonomous vehicle scenario.

Define what counts as an incident: anything from a minor near-miss report to an actual collision or safety disengagement.
Break the estimate into a funnel: number of Waymo vehicles operating in the SF Bay Area, times average miles driven per vehicle per day, times an assumed incident rate per mile.
State assumptions explicitly, for example a fleet size in the hundreds of vehicles, each driving roughly 100 to 150 miles a day.
Assume an incident rate based on publicly discussed autonomous vehicle safety data, for example a rough range around 1 reportable incident per 10,000 to 100,000 miles driven.
Calculate the estimate and sanity check it against the order of magnitude of publicly reported autonomous vehicle disengagement or incident reports.
Segment by incident severity, since most will be minor and only a small fraction will be significant collisions.

States each assumption as a number out loud (fleet size, daily miles, incident rate) so the interviewer can push back on any single one.
Sanity checks the resulting estimate against the general order of magnitude discussed in public autonomous vehicle safety reporting.
Segments incidents by severity rather than treating all incidents as equally significant, which shows more realistic thinking.

Giving a single number without showing the funnel or assumptions behind it.
Ignoring incident severity and treating a minor near-miss the same as a serious collision.

Chapters of the AI PM course that teach what this question tests.