Facebook Metric Design: What Counts as a Meaningful Interaction and How to Track It

You are a product manager working on Facebook, the core social app. Explain how you would define a "meaningful social interaction" on the platform, concretely enough that the team could compute it from product data, and which metrics you would use to evaluate whether Facebook is delivering it. ```hint Not every interaction counts Before naming any metric, sort a few example interactions into "meaningful" and "not meaningful", such as a thoughtful reply from a cousin versus a tap on a brand's post. Then ask what separates them and which logged signals could capture that difference. ``` ### Clarifying Questions - Should the scope be only the main Facebook app (Feed, Groups, and messaging that starts from it), or the whole family of apps including Instagram and WhatsApp? - Will this definition be used to steer Feed ranking, to serve as a company goal metric, or both? A metric that drives ranking needs to be much harder to game. - Should interactions with Pages, public figures and Groups of strangers count, or only interactions between people who already know each other? - Is success judged on short-term engagement, or on longer-term outcomes such as retention and how people feel about their time on the app? ### What a Strong Answer Covers - A definition precise enough to compute: who is on each side of the interaction, which actions qualify, and what effort or reciprocity threshold separates meaningful from passive engagement, tied to why it matters for Facebook's users. - A metric hierarchy: one north-star measure, supporting metrics for content creation, consumption and responses, and the reasoning behind the choice. - A way to check the behavioral proxy against what people themselves say is meaningful. - Guardrails for gaming, divisive or low-quality engagement, and side effects on other parts of the business. - How the metrics would be segmented and used to judge experiments. ### Follow-up Questions - After a ranking change, comments per user go up, but survey respondents say their Feed feels more negative. How do you reconcile those signals? - How would your definition treat a long comment argument between two strangers in a large public Group? - A user with only a handful of friends will score low on your metric no matter what. How do you avoid building a product that ignores them?