Meet Facebook Graph Search #22B
Here’s what’s in store according to the SMX West conference agenda:
“In January, Facebook unveiled a major move into search space: Facebook Graph Search. Powered by what people’s friends are liking and sharing, Facebook Graph Search can make recommendations ranging from places to eat to products to buy. In this session, we’ll look at how Facebook Graph Search works and how search marketers can tap into this developing new service.”
Moderator: Danny Sullivan, Founding Editor, Search Engine Land (@dannysullivan)
Speaker: Loren Cheng, Product Manager, Facebook, search team
Danny asks the audience who has Graph Search, and a few raise their hand. The rest raise their hand when Danny asks if you’ve signed up and are waiting for an invite (including myself). This kind of session is a challenge to program since so few people have actually seen it. As Loren takes the podium he explains he’s going to show us the Graph Search experience.
The mission of Facebook is to make the world more open and connected. Up until now we’ve connected to what we already know. But the map we keep at Facebook is one part of an enormous graph, and there are parts of this graph that you might be interested in – movies you want to see, restaurants you want to try. They want to help you discover these things.
The graph is big and changing:
- 1 billion people
- 240 billion photos
- 1 trillion connections
Three pillars (tools) to answer questions people have on the social graph:
1. News Feed: What are my friends saying today and how can I connect with what’s happening in the world?
2. Timeline: Who is this person, what’s their story and what’s the identity they’ve curated for the world?
3. Graph Search: What’s the next book I should read? Or ask any question you want, take any slice or view of the graph I want.
Web search vs. Graph Search: In Web search, a top movies query will give you a list of resources. On Graph Search you get more specific with your questions. Movies liked by people who like my movies, or movies liked by people who like film, movies liked by people in Internet marketing!
What can you search for? People, photos, places, interests.
Let’s look at Graph Search. There’s a bold blue bar along the top that you use to search for one of the listed items. If you search “my friends” it will come up with a list of people and relevant info, they are specific about the info they include, and in this search it tells you how long you’ve been friends and mutual interests, mutual friends, etc. A search for “restaurants” has different types of relevant info, like hours, address, checkins.
If you want to throw a Star Wars movie night, search for “friends who like Star Wars”. The results have friends who live a little too far away. So he changes the query to “friends who like Star Wars and live in sf” and the friends list changes. Also note that they accept different inputs, so “SF” will work for San Francisco. He adds “movies liked by” to the front of the last query and he sees that these friends also like The Matrix, so now he has his double header.
It’s quick to pivot, like from a population of people, to movies. They built it in a natural language interface to allow for that flexibility of use and pivot for the info they want.
“Product managers who have been founders and are not my friends” is a search that’s going to find him people off his graph.
You used to have to go to a user’s timeline to find photos of a user. Now you can type “photos of my friends” and the result is a collage of great photos from good friends. “Photos of my friends in national parks” which is signaled by check-ins as well as the description of the photo; they key in to different available signals. “Photos of my friends before 1990” is a very cute result set. And you can of course search for photos of places, “photos taken in berlin, germany in 1989” shows the breaking down of the Berlin Wall.
“TV shows my friends watch” or “music liked by people who like the beatles”. “Books read by software engineers at facebook” is a search you could do to get into the culture and mindset of a new company you’re starting at.
“Restaurants in San Francisco” and you’ll see 3 bodies of info they pull from.
1. A business owner who provides the address, open hours, a description of the business.
2. Facebook info at large, like total number of check-ins, star ratings
3. Social info like “likes”
These results seem kind of touristy. What about finding more local trendy hot spots. Find out the restaurants liked by people who work at eleven madison park (a trendy up-and-coming restaurant).
“Restaurants my friends like” shows up Baretta and it’s got high star reviews and he sees his friends have been there. When he clicks to see what friends of his have been there he sees his friend Mike and remembers he and Mike have similar food taste. In Mike’s info here it shows other checkins so he clicks that and sees all the places he’s checked in. From here, in the right side he specifies to only display restaurants, and now Mike is his virtual tour guide for restaurants to go to in SF.
This is just one example we’ve gone deep into. There’s nothing about this info that is available to you here that wasn’t before, but it makes it easier to get to.
Privacy on Graph Search
Every piece of content has its own audience.
Most content is not public.
You can only search for data that is open to you, and only people, places and interests.
Marketing with Graph Search: New ways to be discovered
More ways to be found, by location, categories, photos, videos, connections and interests. Improved results, like bigger photos and social context.
Share your information. Create a Page and make sure it’s categorized correctly and the info about the page is complete. Add name, vanity URL, category and info in About section. Add 620×620 pixel profile photo. Share your address. Add locations for each outlet and opening hours.
Strengthen your connections for ranking improvement. Search results are based on connections of the person searching. It’s not the number of connections but the strength of those relationships that are important.
Currently can’t buy ads on Graph Search. Sponsored Results still available in the search bar.
Sign up for the beta at facebook.com/graphsearch
Q: What about the people who don’t share a lot of info on Facebook?
LC: We’ve found that this makes it more likely that people will check-in and use Facebook like this since they know their friends will benefit from it.
DS: I’ve liked 3 books on Facebook and like 500 on Good Reads.
LC: Good Reads has an Open Graph integration, and we’re working on feeding that in to Graph Search now. We don’t feel we need to build a book reading application. We want to create the social fabric for these things. Integrating Open Graph is definitely coming in future versions.
DS: What about online only business, advice for them?
LC: Info about a page, even if online only, is important. You can have just as rich conversations online only as a local business. More info online means more pushing Graph Search results.
Q: Do you use geotagged info in a photo to show it up in places search?
LC: They don’t use any embedded data in a photo, just what the user chose to geotag because again it’s about what the user is comfortable sharing.
DS: What about searching posts?
LC: Posts, Open Graph and other languages are what they’re working on hard now.
Q: Can you provide a way for brands to monitor what other people are saying about their brand or around their network that might look bad for them.
LC: They look for areas of the graph that aren’t clean, and have ways to find these outliers. The graph isn’t as useful if it isn’t clean.
A while ago, people were report a lot of inappropriate images. They had a team of humans that were reviewing them, and it turns out that a lot of these images weren’t objectionable at all (violence, copyright violation, etc). It was people that didn’t like the way the photo made them look that a friend had posted. That’s the flow that led to allowing people to untag themselves.
DS: Dislike button? Or following without liking?
LC: It’s important to remember that when people share, those you share it with understands it in context. Sharing publicly or with a more constrained audience, they respect that as well. When you “like” Mitt Romney and all your friends know you supported Pres. Obama, they understand that like was meant more like an “ironic button.”