Hacker News Reader

While often ignored, item-to-item similarity recommenders work really well for new items and have been the basis for Amazon's and YouTube's early recommendation algorithms

https://glinden.blogspot.com/2011/02/youtube-uses-amazons-re...

View on HN

I think you may be misunderstanding the parent (I may also be misunderstanding though) - they're saying that with a new user (not a new product), you have no information for providing recommendations to them; you have no idea what they're interested in.

I was talking about new products/movies/whatever, not users. Sorry, "new entity" was too vague in retrospect.

My point was around the utility of marketing, at least in theory. If I release a new brand of baked beans, how would a recommender based on past sales (and combinations of sales) know to recommend it to anyone. Marketing allows it to go up the list (unfairly, according to the post I was replying to).

Ah, okay. Then yeah, the original reply is valid. I assumed you meant a new user to a site, since you can't recommend anything then no matter what you've built (without user tracking that can maybe have built up some kind of profile from elsewhere).

Reply on news.ycombinator.com