Netflix is an interesting example since it also has devolved so much over the years. Back in the olden days before streaming was a thing there was a table view of all their content. The table allowed you to sort by the different columns and I think you even were able to filter. I loved that view. It became more hidden and then went away.
They also had an API to pull content info. I used the API to write a little Ruby script that just listed all their movies ordered by how much Netflix thought I'd like them. This worked really well since their predictions were surprisingly accurate. More surprising was that many of the movies that they predicted I'd like a lot were never recommend by their own UI. Of course the API went away soon after as well.
Ever since I noticed that Netflix's UI wasn't suggesting what they accurately thought I'd like I've been wondering what they are actually optimizing for. Around that time there was also a lot of buzz around their contest to predict this stuff. Why would they put all this effort in and then not use it? Is their content that they benefit from more of I watch it? Are there licensing deals were Netfkix pays when I watch? Are they saving infrastructure and data transfer costs if everyone streams the same stuff?
All this has gotten way worth when they moved from making DVDs to streaming.
Edit: I think what might originally have been a driver for this when they went to streaming is to hide how small the library actually was. We all had still been used to a world where we could watch almost everything we wanted via DVDs from a single service. That's almost unfathomable in our new world where most perks are at least subscribed to a handful of services and still don't have the same access, especially to movies, as we had with Netflix DVDs.
Edit2: another problem is of course that what I might like isn't necessarily what if watch in any given moment. I might give Andre Rublev five stars, but an I likely to watch a 3+ hour movie with subtitles on a whim on a week night?