Netflix offered a million dollars to anyone who improved its recommender by 10%. The prize was awarded in 2009 to an ensemble of teams (BellKor’s Pragmatic Chaos).
The contest industrialized holdout practice, leakage awareness, and stacking. It also showed limits: the winning model was not deployed as-is due to complexity and product shifts (streaming).
Culturally, it legitimized ML as a consumer-business engine—a bridge to the next decade’s ubiquitous personalization.
