This forum is about wrong numbers in science, politics and the media. It respects good science and good English.
Netflix seems to keep increasing the number of variables they are trying to tweak to optimize what they show you in their buckets.
I have found the "Because you watched " sections to be more useful, but I still roll my eyes and 90% of their suggestions.
There is a question for the epidemiologists to peek into. How well do the associations work for the Amazons and Netflix of the world. What is the Relative Risk that someone will watch a movie or by a product given than that they watched a movie with x, y, and z or looked for products a, b, c...
The marketing people can play games and see what sticks.
The epidemiologists play games and ignore all the people that don't die.
I don't think Amazon need a good working recommendation or suggestion system anywhere near as much as Netflix seem to do. As a customer of Amazon for about 15 years, I've never bought anything that was suggested to me by them. The suggestions that Amazon make are often strange, and even give the impression that they're trying to shift items that aren't selling particularly well. A list of screenshots of what are claimed to be the most ridiculous suggestions for other items to buy from Amazon is given on this link:
I have several books written by Ann Coulter. It doesn't keep me, my grandkids, my dogs, spiders, or mosquitoes away. It must not work.
Sadly, as a guy who works in this area, I can see how most of those happened.
Lost in the noise is all the noise of the person doing the search and the unnoise the programmer has to somehow insert.
Most of the suggested combinations I get are boring.