Netflix CEO Ted Sarandos acknowledged that the service's app may bury shows users want to watch under content they do not, describing gaps in how streaming platforms help viewers find hits.
The admission points at a structural problem. Recommendation systems optimise for engagement with what is already popular, which works against new titles that need a critical mass of viewers before the algorithm will promote them.
Studios spend heavily on production and comparatively little on discoverability, and the gap has widened as libraries grew. A title that is not surfaced in its first week rarely recovers.
The fix is not simply better machine learning. Editorial curation, trailers placed inside the player, and social proof all influence what people start, and none of them scale as cheaply as a ranking model.
Netflix's incentive to solve this is direct: every unwatched title is amortised cost with no return.
