Waymo Robotaxi Estimation: Rider-Reported Safety and Service Issues in the Bay Area

Waymo operates a driverless ride-hailing service in the San Francisco Bay Area. Estimate how many incidents the people using that service could report there. Lay out your structure, state each assumption as a round number, do the arithmetic and check whether the result is plausible. ```hint Start from rides, not residents Reports come from trips, so anchor the estimate on how many rides happen. Then ask whether a dirty seat and a collision should really share one rate. ``` ### Clarifying Questions - What counts as an incident: only safety events such as collisions and near misses, or anything a rider flags, including cleanliness, pickup problems, items left behind and app trouble? - Over what period should the number be given: per day, per month or per year? - Who counts as a user: only riders, or also pedestrians, cyclists and other drivers who report a robotaxi's behavior? - Should the estimate cover only the part of the Bay Area the service runs in today, or the whole region? ### What a Strong Answer Covers - An agreed definition of an incident and a time period before any numbers appear. - A ride-volume estimate built from fleet or demand assumptions, ideally cross-checked by a second method. - Separate rates per incident type that distinguish how often something happens from how often riders bother to report it. - Correct arithmetic, one final number with a plausible range, and a sanity check. - A link to the decision the estimate serves, such as staffing rider support, and which assumption moves the answer most. ### Follow-up Questions - The fleet doubles next year. Do reports double too, and what could make them grow faster or slower than rides? - How many support staff would your estimate imply, and how would you handle the most urgent incident types differently? - Reports per ride drop by a third after an app update. How would you tell fewer problems apart from fewer people bothering to report?