When does fan-out on write stop making sense?

When follower counts make the write amplification unpayable or the wasted work dominates. A post to 50 million followers means 50 million cache insertions, mostly for users who will never scroll far enough to see it. Most systems set a follower threshold: below it, push at write time; above it, store once and merge at read time for the followers who actually show up.

Why store post IDs in the feed cache instead of full post content?

Three reasons: memory (8 bytes versus kilobytes per entry, across hundreds of entries for millions of users), consistency (edits and deletes happen in one place instead of a thousand cached copies), and cheap fan-out writes. The cost is a hydration step at read time, which stays fast because it is one batched fetch against a hot post cache.

How do feeds stay correct when users unfollow someone?

Lazily. The unfollowed account's posts remain in the precomputed feed cache until they age out or a cleanup pass removes them, and the serving layer filters hydrated results against the current follow list so the user never sees them. Synchronously rewriting feed caches on every unfollow would be fan-out on write for an event nobody is waiting on.

Common Designschapter 3 of 8 · 4 lessons

News Feed / Timeline

Aggregate, rank, and serve personalized posts from accounts you follow.

← All chapters View in cheatsheet

Someone with 200 followers posts a photo. Someone with 50 million followers posts a photo. Both should appear in their followers' feeds within seconds.

Do the work when they post, and the first case is 200 quick writes and the second is 50 million, which will still be running an hour later while the queue behind it backs up.

Do the work when somebody opens the app instead, and posting is instant, and every feed open now has to gather and merge posts from the 400 accounts that person follows, at 20,000 feed opens a second.

Neither answer survives on its own, and the reason is that your follower counts are not evenly spread. That single fact shapes every feed system in production.

Lessons

4 in this chapter

Requirements and EstimationFeeds are read 100 times more than they are written, but fan-out multiplies every write.2 min
Fan-out on Write vs Fan-out on ReadPush for the many small accounts, pull for the few giant ones, merge at read time.3 min
The Serving ArchitectureStore IDs in the feed cache, hydrate the content at read time, and rebuild lazily for the inactive.3 min
Ranking and PaginationCursors keep infinite scroll stable; ranking is a candidate-scoring layer on top of the same plumbing.3 min