August 19, 2010

Heading toward the Facebook recommendation engine

Filed under: facebook, lbs, local search, maps, yellow pages, yelp — Tags: — Rakesh Agrawal @ 9:13 am
Image representing Facebook as depicted in Cru...

Image via CrunchBase

There’s an interesting thread over at Mike Blumenthal’s blog on the effect of Facebook Places on the local reviews space.

My view is that reviews and updates will coexist, much as blogs and Twitter coexist. People who were less committed to reviews will migrate their activity to Facebook Places updates. But Places could lead toward the ultimate recommendation engine.

In the local space, there’s really only one review site that matters: Yelp. They’ve got a strong set of tools and an active and engaged community. New restaurants and bars, which are often of the most interest, will have a dozen reviews on Yelp a year before they even show up on many Yellow Pages sites.

There are three big challenges with Yelp:

  • It’s been too successful. Many restaurants have hundreds of reviews. Although Yelp provides great tools for analyzing the data, it can still feel overwhelming. It also discourages participation from more casual users. In the early days of Yelp, I was an active reviewer. That’s tapered off substantially — what’s the marginal benefit of me writing the 426th review of a place?
  • These aren’t my real friends. I don’t know how compatible their tastes are with mine. It also affects the propensity to write reviews. People are more likely to do something that helps their friends than something that helps a generic audience.
  • Skewed demographics. Yelp primarily caters to a young, urban demographic. If you’re a mom in the suburbs, its value is more limited.

Facebook Places lowers the bar to participation and ties it into real-life social networks. Instead of writing out a long review, a few clicks is all it takes. Combine that with Facebook’s large user base on mobile devices — its monthly uniques on mobile devices is 4x Yelp’s monthly uniques on the Web — and we’ll see a tsunami of local data. (For more on importance of massive amounts of data, watch Google’s Peter Norvig’s talk.)

While each blip may not be as rich as the data in Yelp, you could build a recommendation engine to infer a lot from that data.

If I see that a place I am considering visiting is regularly frequented by my friends with families, I can infer that it is good for kids. Positive reviews can be inferred by friends going back to a place regularly. There are some friends who I have negative taste relationships with. If I know that they’re regulars somewhere, I know not to go there. Facebook can also make recommendations based on places I’ve visited and the overlaps with places my friends have visited. Facebook also has real demographic information which could be used to tailor recommendations.

Status updates in the social network also prompt discussions. Even if the original poster doesn’t write a review, it may be followed up by “hey, I was thinking of going there. what did you think of it?” Facebook could also close the loop by prompting people to add star ratings, Like or add comments a few days after a check in.

When it comes to restaurant reviews and recommendations, most people are looking for “good enough”. While you could spend hours reading every Yelp review of several restaurants and possibly get a better answer, a recommendation based on your friends’ activity is probably nearly as good. Facebook has done really well with good enough; Facebook Photos dominates online photo sharing, despite many functional weaknesses when compared with flickr.

I built a prototype of this when I was at AOL Search and even with a few users in the system, it worked really well.

More on: Facebook, local search, Yelp

See also:

About these ads


  1. I strongly agree that there will be co-existance…not just because of time constraints/willingness but because of other factors as well…

    In rural America, the density is too low for any significant number of updates…in that situation reviews will hang on much longer.

    I also think that it will likely break down by industry…for entertainment, food venues, check-ins and updates make all kind of sense.

    It is not so clear to me that they make sense for the plumber that comes to you or for that service that you use once in a life time (ie a custom jeweler for the engagement ring or the personal injury attorney).

    Comment by Mike Blumenthal — August 19, 2010 @ 10:16 am

  2. I agree that check ins aren’t really relevant for service-related businesses.

    I’m not sure that that the rare cases you mentioned are well solved by existing review products. I still worry about how much those are gamed given low density.

    I’m much more likely to address those with a status update like “Anyone know a good jeweler in the area?”

    Comment by Rocky Agrawal — August 19, 2010 @ 10:44 am

  3. [...] Heading toward the Facebook recommendation engine [...]

    Pingback by Facebook Places is at the beginning of a long road « reDesign — August 20, 2010 @ 8:43 am

  4. How could Yelp get more people to log in?…

    Yelp is already doing many of the things that would increase logged in users: * International expansion. * Mobile. Absolutely killing it. * Facebook Connect. Making it easy to login. To some extent, I think Yelp is a victim of its own success. They’ve…

    Trackback by Quora — October 24, 2010 @ 11:10 am

  5. Is Yelp completely screwed now that Google have launched Hotpot?…

    No, putting up a tool is necessary but not sufficient in the local space. Many, many companies have provided tools to do ratings and reviews in local. (Including Google, for many years.) In order for a local tool to be successful, you have to have an a…

    Trackback by Quora — November 15, 2010 @ 11:53 pm

  6. [...] of my longstanding complaints about local review sites is that they require users to do too much work to figure out where to go. Google hopes to change that by providing recommendations based on your ratings and presumably the [...]

    Pingback by Google Hotpot takes a stand in Portland « reDesign mobile — February 19, 2011 @ 10:48 am

  7. [...] In some ways, this is an easier problem to solve than Web search. If you’re looking up answers for Jeopardy, there is usually only one right answer. And if Google can’t find it, you know right away. For a discovery-oriented local search, there is more than one right answer. And if the answer isn’t what you were expecting, you won’t know for hours and you might not blame Google. (The restaurant might have had an off night.) For more details, see my earlier post about making intelligent recommendations in local search. [...]

    Pingback by Local search is starting to get more social « reDesign mobile — February 23, 2011 @ 1:25 pm

  8. [...] Making intelligent recommendations requires having a lot of data. The easier you can make it, the more participation you will have. Few people will go through the trouble of writing detailed reviews, but 1-click ratings can provide important signals and will have a higher participation rate. See more on recommendation engines for local search. [...]

    Pingback by How the battle for local search will be won « reDesign mobile — February 25, 2011 @ 12:15 am

  9. [...] As you can see at the top of the screenshot, the ease of rating has the potential to generate a lot of quick hits of data — my ratio is about 4 ratings to each review. See my earlier post about using lots of small nuggets of data to make intelligent recommendations. [...]

    Pingback by Google Hotpot a strong competitor to Yelp « reDesign mobile — March 3, 2011 @ 4:36 pm

  10. [...] thousands of local businesses to identify those that you might be interested in. (See my post on recommendation engines.) Much like Google’s Hotpot, recommendations come from a number of different [...]

    Pingback by Foursquare 3.0 takes mobile ball to a whole new level « reDesign mobile — March 9, 2011 @ 3:19 pm

  11. Is Foursquare 3.0′s “every check-in counts as a recommendation” fundamentally flawed long-term?…

    No. With enough data, checkins are enough to provide good recommendations. Over time, crappy places will have fewer check ins per unique user versus good places. It’s more of a problem today when data is sparser. In my research, I found that the mean …

    Trackback by Quora — March 12, 2011 @ 5:47 pm

  12. Seems like this is what Facebook wanted a geolocation check-in company for. But there has to be a way of inferring whom your tastes are negatively correlated with (not just your friends!) and what dimensions (eg “good for kids”) constitute your taste.

    Comment by recommendation engineer — March 28, 2011 @ 10:06 pm

  13. How much value is a new-site rating worth to a site like Yelp or a Google HotPot?…

    It’s one of those “I know it when I see it” things. One thing that Yelp is really good at — and is very valuable to consumers — is getting reviews of newly opened businesses in a timely manner. As Joshua Greenough pointed out, information is more …

    Trackback by Quora — May 27, 2011 @ 3:31 pm

  14. [...] I’ve seen Facebook testing product concepts that point at its future direction in local. One test asks you to compare recent places you’ve visited. Another feature lets users contribute data on restaurants and other venues. Facebook realizes that most average people won’t go to the trouble of writing long reviews like those on Yelp. It’s better to collect small nuggets of data from massive audiences. See my post, Heading toward the Facebook recommendation engine. [...]

    Pingback by Why Google’s acquisition of Zagat matters « reDesign — September 8, 2011 @ 6:17 pm

  15. Today, Google announced its acquisition of Zagat, the company that publishes the venerable restau……

    Today, Google announced its acquisition of Zagat, the company that publishes the venerable restaurant and hotel review guides. It’s a terrific acquisition. If Google executes correctly, this deal could be as significant as the YouTube deal has been. (I…

    Trackback by Quora — September 8, 2011 @ 6:21 pm

RSS feed for comments on this post. TrackBack URI

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

The Silver is the New Black Theme. Create a free website or blog at


Get every new post delivered to your Inbox.

Join 227 other followers

%d bloggers like this: