Practice/Apple/Design a Social Media Feed Generation System

Design a Social Media Feed Generation System

System DesignMust

Problem Statement

Design the system behind a social media home feed — like Twitter, Instagram, or Facebook. When a user opens the app, they see a personalized timeline of posts from people they follow, ranked by relevance, with new content appearing in near real time.

The challenge is deciding how and when to assemble each user's feed. Do you pre-compute feeds when posts are published (fanout-on-write), or assemble them on demand when the user opens the app (fanout-on-read)? Each approach has trade-offs around latency, cost, and freshness — especially when a celebrity with 10 million followers posts.

Key Requirements

Functional

Home feed -- users see a personalized, paginated feed of posts from accounts they follow
Post publishing -- users create posts (text, images, video) that appear in their followers' feeds with low latency
Follow/unfollow -- users manage their follow list, which shapes their feed content
Ranking -- posts are ordered by relevance (recency, engagement, user preferences), not just chronologically

Non-Functional

Scalability -- support hundreds of millions of users, with some accounts having millions of followers
Latency -- the feed loads in under 300ms; new posts from followed accounts appear within a few seconds
Availability -- the feed must always load, even if some backend services are degraded
Consistency -- eventual consistency is acceptable; a post appearing a few seconds late in some feeds is fine

What Interviewers Focus On

Based on real interview experiences, these are the areas interviewers probe most deeply:

1. Fanout Strategy

The fundamental design decision: when a post is published, do you push it to every follower's feed (fanout-on-write), or do you pull from followed accounts when the user requests their feed (fanout-on-read)?

Hints to consider:

Fanout-on-write: pre-compute each user's feed timeline by writing the post to every follower's inbox → fast reads, expensive writes
Fanout-on-read: at read time, fetch recent posts from all followed accounts and merge → cheap writes, expensive reads
Hybrid approach: use fanout-on-write for normal users, fanout-on-read for celebrity accounts (millions of followers)
Discuss the storage cost: fanout-on-write duplicates every post N times (where N = follower count)

2. Handling Celebrity Accounts (Hot Keys)

A celebrity posting triggers fanout to millions of timelines. Naive fanout-on-write will overwhelm your workers and storage.

Hints to consider:

Mark accounts above a follower threshold as "celebrity" and exclude them from fanout-on-write
When a user requests their feed, merge their pre-computed timeline with a live fetch of celebrity posts
Use batching and rate limiting on the fanout pipeline to prevent worker overload
Discuss how to detect and adapt to sudden follower count changes (account going viral)

3. Feed Ranking and Personalization

A purely chronological feed is simple but not engaging. Users expect relevant content surfaced above purely recent posts.

Hints to consider:

Score each candidate post using signals: recency, author engagement history with this user, post engagement (likes, comments), content type preference
Run ranking as a lightweight model at read time over the candidate set (top ~500 posts from the last few days)
Cache ranked feeds with short TTLs and invalidate on new activity
Discuss the cold start problem: new users have no interaction history for personalization

Practice/Apple/Design a Social Media Feed Generation System

Design a Social Media Feed Generation System

System DesignMust

Problem Statement

Key Requirements

Functional

Home feed -- users see a personalized, paginated feed of posts from accounts they follow
Post publishing -- users create posts (text, images, video) that appear in their followers' feeds with low latency
Follow/unfollow -- users manage their follow list, which shapes their feed content
Ranking -- posts are ordered by relevance (recency, engagement, user preferences), not just chronologically

Non-Functional

Scalability -- support hundreds of millions of users, with some accounts having millions of followers
Latency -- the feed loads in under 300ms; new posts from followed accounts appear within a few seconds
Availability -- the feed must always load, even if some backend services are degraded
Consistency -- eventual consistency is acceptable; a post appearing a few seconds late in some feeds is fine

What Interviewers Focus On

Based on real interview experiences, these are the areas interviewers probe most deeply:

1. Fanout Strategy

Hints to consider:

Fanout-on-write: pre-compute each user's feed timeline by writing the post to every follower's inbox → fast reads, expensive writes
Fanout-on-read: at read time, fetch recent posts from all followed accounts and merge → cheap writes, expensive reads
Hybrid approach: use fanout-on-write for normal users, fanout-on-read for celebrity accounts (millions of followers)
Discuss the storage cost: fanout-on-write duplicates every post N times (where N = follower count)

2. Handling Celebrity Accounts (Hot Keys)

A celebrity posting triggers fanout to millions of timelines. Naive fanout-on-write will overwhelm your workers and storage.

Hints to consider:

Mark accounts above a follower threshold as "celebrity" and exclude them from fanout-on-write
When a user requests their feed, merge their pre-computed timeline with a live fetch of celebrity posts
Use batching and rate limiting on the fanout pipeline to prevent worker overload
Discuss how to detect and adapt to sudden follower count changes (account going viral)

3. Feed Ranking and Personalization

A purely chronological feed is simple but not engaging. Users expect relevant content surfaced above purely recent posts.

Hints to consider:

Score each candidate post using signals: recency, author engagement history with this user, post engagement (likes, comments), content type preference
Run ranking as a lightweight model at read time over the candidate set (top ~500 posts from the last few days)
Cache ranked feeds with short TTLs and invalidate on new activity
Discuss the cold start problem: new users have no interaction history for personalization