Netflix is a popular streaming service that recommends TV shows to
millions of users.
Their recommendation system is a big reason why people stay on Netflix
and keep watching.
How Netflix’s Recommendation System Works
Netflix uses a mix of different methods to recommend shows:
- Collaborative Filtering: Suggests shows based on
what similar users have liked.
- Content-Based Filtering: Looks at the show’s genre,
& actors
- Behavioral Data: Tracks what you watch, search for
and how long you watch.
Netflix shows recommendations in sections like: - “Because you
watched…” - “Top Picks for You”
Scenario Design Analysis
Here’s the Scenario Design, answering three questions:
Who are the users? | Millions
of people who pay to watch Netflix.
What are the users trying to do? | Keep people watching
so they don’t cancel their subscription.
How does the site help them? Shows them stuff that
matches their tastes and helps them find new favorites.
Should We Do Scenario Design for Both? Yes!
Netflix’s goal (keep users engaged) and the user’s goal (find
something good) are similar but not exactly the same.
Ideas to Improve Netflix’s Recommendations
Here are a few ways Netflix could get even better:
- Mix in More Surprises: Show some titles outside a
user’s usual picks to help them discover new things.
- Explain Why It’s Recommended: Show messages like
“Because you liked” more clearly.
- Mood Filters: Let people pick a mood like “funny,”
or “serious,”
- Better Feedback Options: Let users say what they want more or less
of, not just thumbs up or thumbs down.
- Social Recommendations: Let people see (if they want) what’s popular
with friends.
References
Gómez-Uribe, C. A., & Hunt, N. (2015). The Netflix
Recommender System: Algorithms, Business Value, and Innovation. ACM
Transactions on Management Information Systems. Link
Netflix Research. Personalization and Recommender
Systems. Link
Netflix Tech Blog. System Architectures for Personalization
and Recommendation. Link
Bhattacharya, M., & Lamkhede, S. (2022). Augmenting
Netflix Search with In-Session Adapted Recommendations. Link