Design a Ranked Home News Feed for a Large Social Network

Design the home news feed for a large social network. When users open the app, they see a ranked, infinitely scrollable list of recent posts from the friends and pages they follow. When they publish a post, it should show up in their followers' feeds. Design the backend that accepts new posts and serves each user's feed. ```hint Where to pay the cost Decide whether a feed is assembled when a post is written or when the feed is read, then test that choice against an account with an enormous follower count. ``` ```hint Narrow before you rank Scoring every post a user could possibly see on each request is too slow; think about what produces a small candidate set first. ``` ### Constraints and Clarifications - Media storage and delivery can be treated as an existing service; focus on post metadata and feed assembly. - Scale, freshness targets, and ranking signals are not given; ask for them or state assumptions. ### Clarifying Questions - How many daily active users are there, how many accounts does a typical user follow, and what is the largest follower count to support? - Is the feed ranked or chronological, and which signals may ranking use? - How quickly must a new post appear in followers' feeds? - Are ads or recommended posts from accounts the user does not follow in scope? - How must deletions, edits, and audience restrictions on a post be reflected in feeds that already contain it? ### What a Strong Answer Covers - Fan-out on write versus fan-out on read, and a hybrid for accounts with very large follower counts - Precomputed per-user feed storage and bounds on its size - Candidate generation followed by ranking within a latency budget - Cursor-based pagination that stays stable while new posts arrive - Correct handling of deletes, edits, privacy changes, and unfollows - Caching, sharding, and hot-key handling - Failure handling, graceful degradation, and freshness metrics ### Follow-up Questions - An account with a very large follower count publishes a post. Walk through what happens on the write path and on a follower's next read. - How do you make sure a deleted post, or a post whose audience was narrowed, disappears from feeds quickly? - How would you serve the feed of a user returning after months of inactivity? - How would you test a new ranking model without hurting feed latency?