Skip to content
Handbook navigation

Full-access lesson

Recommender Systems

About 75% of what Netflix members watched came from some kind of recommendation, the company's engineers wrote in April 2012, in a post titled "Netflix Recommendations: Beyond the 5 stars". In October 2006 it had framed recommendation as predicting those stars, and offered a million dollars for predictions of held-out ratings 10% more accurate than its own system's; Chapter 5 told how the prize was won in 2009 by an ensemble Netflix never shipped. In between, the product changed. A DVD arrived by post and was rated afterwards, if at all. A streamed title is played, abandoned or finished, and every play is evidence that nobody had to type in.

The complete curriculum

All 12 volumes, companion resources, interviews, and architecture reviews.

Executable engineering practice

Subscriber-only Python, Java, TypeScript, and SQL labs in the isolated runner.

Recommender Systems | KnowledgeOS