I can’t remember all the details, but from what I remember when I looked this up a few years ago, yes, Netflix did throw the whole thing out the window. They said that the winning algorithm would require more processing power to run, or expensive code changes or something, and so they decided their current engine was good enough.
Then Netflix switched strategies away from highly-personalized recommendations to a dumber algorithm that is more likely to recommend more popular movies, which ends up retaining subscribers better. The reasoning was that users were more likely to cancel their subscription if they had one obviously bad recommendation than if they had multiple so-so recommendations. So rather than recommending, for example, 80% great movies and 20% bad ones, Netflix wants to recommend 5% great movies and 95% okay ones.
Then Netflix switched strategies away from highly-personalized recommendations to a dumber algorithm that is more likely to recommend more popular movies, which ends up retaining subscribers better. The reasoning was that users were more likely to cancel their subscription if they had one obviously bad recommendation than if they had multiple so-so recommendations. So rather than recommending, for example, 80% great movies and 20% bad ones, Netflix wants to recommend 5% great movies and 95% okay ones.