It aligns with my experience as well. I keep getting recommendations for shows I have already disliked. If it's due to a low number of options, and not due to their recommendation system, then surely they don't need such a complex recommendation system after all.
They used to have a large catalog. But then almost everyone pulled out to start competing services. So now it's 90% Netflix films/series, 7% old movies, 3% recent-ish movies.
Yep! Hence why I imagine a review & like/dislike system wouldn't be an order of magnitude worse than their ML system but without having to pay millions in ML salaries, complicate the pipelines, and have a crazy number of microservices. I doubt the total business value for Netflix can truly be greater than the tech investment.
