Recommendation systems decide which items to show when a user has too many choices: videos, products, songs, jobs, restaurants, profiles, posts, or articles.

In system design interviews, what matters is the data flow rather than a list of algorithm names: collect signals, generate candidates quickly, rank them with richer features, apply business and safety rules, serve results with low latency, and learn from feedback.

The best answers make the tradeoffs explicit: relevance vs freshness, personalization vs diversity, offline quality vs online behavior, and latency vs model complexity.