Recommending for Long-Term Member Satisfaction at Netflix
Netflix enhances its recommendation system by using contextual bandits and proxy rewards to optimize long-term member satisfaction.
MAIN POINTS
- Netflix aims to enhance long-term member satisfaction through improved recommendation algorithms beyond short-term metrics.
- Proxy rewards are used to align recommendations with long-term satisfaction, overcoming retention's limitations.
- Delayed feedback prediction helps refine proxy rewards and improve the recommendation system's effectiveness.
TAKEAWAYS
- Contextual bandit models help Netflix personalize recommendations by considering immediate and delayed user feedback.
- Proxy rewards are crucial for capturing user satisfaction beyond click-through rates.
- Predicting delayed feedback allows for timely updates to recommendation policies, improving user engagement.