Comment by jorgeortiz85 | Hacker News Reader

Comment by jorgeortiz85 | Hacker News Reader

Parent Full thread

<shameless plug>

I work for Foursquare, so I'm obviously biased, but I think Foursquare Explore is a great way to find good places to eat.

https://foursquare.com/explore

Our recommendations are based on 30 million tips (short recommendations, <= 200 characters) left by our users all over the world. They also take into account the popularity of the venue, the likelihood of people leaving tips there, the time of day, the day of week, and many other factors.

If you have a check-in history with foursquare, or if you have friends that use foursquare, we'll make personalized recommendations. We'll recommend things we think you'll like based on your check-in history and your friend's check-in histories, and we'll highlight which of your friends have been to the places we're recommending, and how often.

The product isn't perfect yet, and there are definitely some big deficiencies if you try to use it as a Yelp replacement (notably, foursquare lacks: hours, webpage links, whether they deliver or not, etc). But it's constantly improving, and it's already pretty good.

</shameless plug>

I just tried it. Top recommendation was Chili's. I can't imagine anybody needs help discovering the clone chain suburban restaurant strip in their home city. It had 2 specials, and was nothing like anything else on my list or anywhere I have checked in, which I assume means it's actually an ad masquerading as a search result with no indication that top rank was bought.

In one sense, that seems kinda inevitable. Way more people check in to FourSquare (and possibly leave a comment/tip) than check in to, say, Fat Duck or Noma.

I suspect your local Starbucks has two orders of magnitude more checkins that the local specialty coffee roaster/espresso bar…

4sq employee here. It's not bought, just a bad search result. Can I ask what the home city is here?

Atlanta, GA, looking for 'dinner' near Buckhead.

There's still some value there. For some chains (Applebee's is one that comes to mind), some locations are actually pretty decent, while others are worse than fast food.

We don't charge merchants money for listing specials, and we don't sell search results. If and when we let merchants advertise on foursquare, all ads will be clearly marked as such.

There are a couple of reasons we might have recommended that Chilli's. Have any of your foursquare friends checked in there? That's a pretty strong signal for us, so that might be why it's showing up so highly. If not, it might be that a lot of people check in to that Chilli's.

That said, there's probably more we can do to detect chain restaurants and either boost them or bury them based on your personal likelihood of visiting chain restaurants. Unfortunately, all our ranking code currently works at the level of individual venues, so there's work for us yet on knowing that the venue is part of a chain and ranking accordingly.

No friends. You had 2 'specials' and a tip. I'm sure a lot of people check in there just due to volume, but volume strikes me as a questionable signal for an 'explore' feature - if everyone already knows it, I probably already know it too. I understand it's a tough problem, but it just looked like an unmarked paid result.

As you suspect, raw volume is a poor signal. We take a lot of other things into account: repeat visits, proportion of visitors that left a tip, the "sentiment" of the tips, etc.

I just checked this out. Near where I live in the suburbs I got mostly recommendations for chains which I expected since that's really all there is around here. When I get closer to the city where there are fewer chains the restaurants I consider legit that I would recommend show up.

Just for kicks I tried San Francisco since Yelp did such a terrible job with recommendations when I was there last time. Sure enough I got a host recommendations that I would consider trying next time I'm out there.

I think for this to be better you could tweak the relevancy based on different psychographic profiles. For example maybe show different results for a foodie, family, and maybe a "safe mode" for out-of-towners which would display a mix of results from the prior two.

For what it's worth, I just gave it a try and it seemed to make some pretty good recommendations (including places my friends have been to a lot but I've never heard of).

I don't really use recommendation sites/apps, because I've never found any of them to be useful (I find Yelp to be worthless), but I'd actually use this.

I totally didn't even know that was a thing in Foursquare. I use Foursquare almost exclusively to catalog where I go out to places, so I can remember later when the last time I was there was.

As someone who doesn't work at Foursquare, I agree! If you're friends use Foursquare, Explore is a great way to find places to eat nearby.

Even if none of your friends have ever written a tip, I've found that seeing that one of my friends has been to restaurant 3 or 4 times is great way to find good places to eat or drink.

Reply on news.ycombinator.com