Hacker News new | ask | show | jobs
by andrewcross 5114 days ago
Agreed. Not to say there isn't a problem with Yelp's practices (as have been documented elsewhere), but this article does not make a good case against them.

From an algorithmic standpoint, filtering out reviews from new users makes sense. Otherwise you'll get the owner and friends stuffing in good reviews.

Just look at what the data might look like for the 17 people. -They all sign up in a period of a few weeks. -They all review this one place and nothing else. -16 of the 17 never sign back in again.

It's pretty clear that it's a blatant rating stuff, so filter it out. Again, I'm not saying that the algorithm is perfect, but it makes sense to at least filter out these reviews.

2 comments

> (as have been documented elsewhere)

I don't think that really has been documented elsewhere. There have been lawsuits and allegations, but no proof, as far as I know, that Yelp has offered to compromise their reviews for cash. Yes, they give business owners willing to advertise some editorial control over their listing, but the extent of that control is fairly well-understood and not much more nefarious than Google's text ads. Correct me if there's documented evidence otherwise.

Instead, we get article after article of business owners complaining because their Yelp reviews suck and their obvious attempts to game them have been thwarted. But that's exactly what Yelp should be doing. They're useless if they can be gamed. They're also useless if it becomes well-known that they're pay-to-play, but again, I don't think there's actual evidence of that. They'd have to be pretty stupid to take that course.

>> Just look at what the data might look like for the 17 people. -They all sign up in a period of a few weeks. -They all review this one place and nothing else. -16 of the 17 never sign back in again.

Looking at the 17 filtered reviews, 16 of them are 5-star reviews from people with 0 friends and less than 5 reviews (mostly only have 1 review). There's also a 1-star review from a person with 0 friends and 1 review.

It seems like to me that their filtering algorithm is working pretty well.