Detecting unannounced restaurant closures on Yelp's reservations product

You are the product manager for the restaurant reservations product on Yelp. During the COVID-19 pandemic, many restaurants are shutting down without telling anyone, and Yelp needs to find out which ones have closed so they can be taken off the platform. The difficulty is that restaurants give the product very little data: a restaurant does not have to do anything to accept a booking, because reservations are confirmed automatically and only change if the diner or the restaurant cancels. How would you identify the restaurants that have closed? ```hint Silence is not a signal Because bookings confirm themselves, a closed restaurant looks exactly like a quiet open one inside the reservation system. List everyone who touches a reservation or a listing before, during and after the booked time, and ask what each of them could tell you, or could be asked. ``` ### Clarifying Questions - Should a restaurant that has paused service for a few weeks be treated the same as one that has closed for good? - Does "remove" mean deleting the listing, taking it off the reservations product and search, or labeling it closed? - Which mistake is worse right now: a diner arriving at a locked door, or an open restaurant that is struggling through the pandemic losing its bookings because it was wrongly marked closed? - Do restaurants on the reservations product use any Yelp tool we can observe, such as a host-stand app or a business owner account, or is the listing the only touchpoint? ### What a Strong Answer Covers - A goal and error-cost framing that weighs failed diner visits against wrongly removed open restaurants, and separates temporary from permanent closures. - An inventory of closure signals from diners, restaurants, Yelp's own activity data and outside sources, with the strengths and weaknesses of each given that restaurants do not confirm bookings. - A way to combine signals into a confidence level, with verification and graduated actions before removal and a path for restaurants to reopen. - Product changes that close the data gap in the booking flow itself, weighed against the friction they add for restaurants. - Success metrics for detection quality and speed, with guardrails for restaurants and diners. ### Follow-up Questions - What would you gain and lose by requiring every restaurant to confirm each reservation, and how would you roll that out during the pandemic? - An owner says your system marked their open restaurant closed and they lost a week of bookings. What failed, and what would you change? - How would you stop competitors from reporting rival restaurants as closed?