I don't know, I never found it very good myself?

Netflix had a million-dollar engineering challenge to improve their recommendation algorithm, awarded the prize, then did not implement it. Literally because it was hard.

They were going to run another challenge, then it got scuppered due to privacy concerns. Without violating privacy, it was hard. There was no effort to resurrect the engineering challenge in a way that would accommodate this issue.

Netflix doesn't do a good job, and that's mostly because they don't value the outcome of improved recommendations, given that it is a monumental and non-trivial engineering challenge. And they don't want to.

All of their actions show this.