Netflix recommendations add 4% more viewing time, but only if they're good

Researchers built a model to measure how much extra time people spend watching Netflix because of personalized recommendations. It turns out that the current system keeps people watching 4% longer than a simpler system would. The biggest gains come from recommending shows that are popular but not universally known.

This paper quantifies the value of personalized recommendations in a massive streaming service. It shows that the algorithms Netflix uses are not just showing people what they want, but actively increasing overall consumption. The findings suggest that the effectiveness of these systems depends heavily on the quality of the recommendations, with mid-tier content benefiting the most.