Defining the Goal and Metric Tree for Facebook's Crisis Response Tools

You are the product manager for Facebook's Crisis Response feature. Assume it works as it does publicly: when a natural disaster or another emergency hits an area, people there can mark themselves safe and their friends and family see the update, the feature gathers information about the event, and people can ask for or offer help and start fundraisers. How would you set the goal for this feature, and how would you measure progress toward it? ```hint Picture the first hours Before you pick anything to count, describe what a good outcome looks like in the first hours of an emergency for each group of people involved. Then check whether each number you are considering would actually rise when things go well for them. ``` ### Clarifying Questions - Is the goal for Crisis Response as a whole or for one part of it, such as marking yourself safe? - Am I setting the goal for a new launch, or for the next planning period of the existing feature? - Is the company expecting this feature to add to engagement, or is it judged purely on user value and trust? - Should one goal cover every type of crisis and every region, given how much they differ in scale and network connectivity? ### What a Strong Answer Covers - Why the feature exists for users and for the company, and what success means for each group: people in the affected area, their friends and family, and people and organizations offering help - A goal stated as a user outcome, with a precisely defined primary metric justified against engagement-style alternatives - Supporting metrics across the course of a crisis, from activation through marking safe, help matching and fundraising - Guardrails such as notifying people who are not affected, misinformation, scams and privacy - How to set targets and compare results when crises vary widely in size and type ### Follow-up Questions - After the latest crisis, the share of notified people who marked themselves safe fell compared with earlier crises of a similar kind. How would you investigate? - How would you balance notifying everyone who might be affected against alarming people who are not? - Would you count likes and comments on safety updates toward the goal? Why or why not?