Facebook Graph Search Whitepaper
6 likes6,431 views
Facebook Graph Search is Facebook's internal search engine that utilizes user profile and connection data to return relevant search results. The document discusses how Graph Search works currently, focusing on searches for restaurants, movies, and places recommended by friends. It also speculates how Graph Search may evolve to include more customized content from brands and how content, profiles, and the value of user actions like "Likes" could impact search rankings. Each user's authority and the quality of their connections and engaged fan bases may determine how much their actions influence search results.
Facebook Graph Search Whitepaper
- 1. FACEBOOK SEARCH HOW CONTENT AND THE VALUE OF THE ‘LIKE’ WILL AFFECT RESULTS IN GRAPH SEARCH 2 Designing with Grids
- 2. EXECUTIVE SUMMARY Facebook Graph Search is the search engine inside Facebook. You can use it to find photos, restaurants, places, and new and old friends. It works by joining up all the different data points you and all your friends have entered into Facebook and returning relevant results based on how they are all connected. For example, you can search for ‘restaurants in Atlanta my friends like’. The algorithm uses all your friends’ check-ins and ‘likes’ related to restaurants in Atlanta to give you a list of restaurants that, if you can trust your friends, you’ll like too. This work explores the notion that SEO for Graph Search will be a battleground for companies and brands vying for the top spot in search results just like in Google. However, there will be a difference in how Facebook determines which company, brand, or piece of content ranks better than others. I propose that each individual and brand on Facebook will have an authority score, similar to Klout®. This ultimately will mean that the value of each person’s like, or opinion, will vary. Brands, companies, and individuals will then be judged based on the quality of their fan base, and by extension, the quality of the friends of each of their fans. How qualified a fan is will reflect how relevant the fan’s interests, personal details, such as education and place of work, and Facebook connections are to the brand. 2 3 Designing with Grids
- 3. facebook Graph Search: What is it and where’s it headed? By Adam Westin, Associate Director, Head of Search, Edelman UK Follow on Twitter: @adwestin Add on Google+ Find on Linkedin Find restaurants, music, Facebook Graph Search (GS) rolled out to all Engphotos, activities, and meet lish speakers in the US in August 2013. It uses the new friends all through details of your profile people you already know. and the details of all your The search engine in friends’ profiles to make finding photos, places, Facebook, called Graph and things easier within Search, makes this possible Facebook. Facebook has an ever-growing wealth of personal content. Its motivation to make it sortable and findable is intuitive. Improving the user experience on Facebook means people will spend more time. If at the same time, Facebook can start taking market share from traditional search engines and review sites, it’ll reap the benefits of the ad revenue that comes with it. The social giant is smartly taking a conservative approach to global rollout as it understandably has not only a number of kinks to work out, but also needs to spend time defining what Graph Search will ultimately become. For now, Graph Search {Engine} Results Pages (GSERPs) show unsurprising results: the content for which you have searched. The interesting part is thinking about what GS will become, particularly within the domain of content in GSERPs. For brands and marketers, it will be about what the actions are that are needed to rank well in GS in order to increase visibility of their content and deepen brand engagement. For Facebook, it’s about providing a useful utility with which paid advertising can seamlessly integrate. Facebook has shareholders and stock price to worry about, so it will be no surprise when monetization emerges within GS. 3 4 Designing with Grids
- 4. Content Currently Present in Facebook Graph Search Graph Search results for movie searches are fairly basic; a literal list of movie titles ranked based your friends’ ‘likes’. At the present time, the content for brands/companies shown in GSERPs is pretty narrow in focus and feels a bit like Yelp®, which is an urban reviews site with recommendations by locals for restaurants, shopping, and activities. The current search results are fairly ridged in that customised content from brands, or anyone for that matter, isn’t included. For example, you can search for ‘movies my friends like’ and the results are just a list of FB pages for each of those movies. There’s not yet any video trailer content, no IMDb write-up, or anything else beyond just movie pages. But again, Facebook is focusing on utility and relevance of the results first and foremost, so it’s naturally taking its time. Recommendations from people you trust, a.k.a. your friends Speaking of utility, one of the most common use cases is searching for restaurants your friends have been to (checked in), or recommended (liked the company FB page). The same concept applies to dentists, vets, and most any other business category. You can search for positive things like “restaurants in London my friends like”, which will return restaurants in Atlanta that your friends like. No surprise there. However, GS won’t give a result for negative searches, such as “restaurants in Atlanta my friends hate.” Instead, you’re given a Bing search result complete with related search suggestions and paid search ads. Facebook may choose to entertain including results in GS for negative queries, especially where relevant to reviews and recommendations for restaurants, stores, bars, etc., but for the moment, just as it doesn’t have a “Don’t Like” button, it’s opting to stay positive and happy. 4 5 Designing with Grids
- 5. In any case, GS is still in its infancy. The seen, but rest assured it will happen. With output is limited to a certain set of query that will come the inevitable clamor among parameters and by the details of your con- brands to rank at the top of the results for nections. This won’t always be the case, given searches. The route to the top is easy though. With an eye on revenue generation, in concept: unique, relevant content. The Facebook will find a way to start integrating winners will be those brands that figure out other content within the GSERPs. The trick what their audience wants, and gives it to will be to be able to do so while still putting them. the user experience first. Relevance for the user is going to be the single most imporLet’s establish a key parameter first, tant factor in making GS a success, which though. Facebook isn’t trying to replicate will be judged by user adoption. To this end, Google, at least not Facebook is being smart with how it’s feel- yet. Google indexes In the short term, Graph Search ing out what GS will ultimately become. The the entire web, is a threat to review sites. In current utility focuses on usability and func- while Facebook tionality. It provides something that users utilizes data people the future, it will be a threat can get from nowhere else. From a reviews have entered and to search engines for market and recommendations perspective, this the relationships means advice from people users know and therein. Facebook share of long-tail searches trust. From a data management perspec- doesn’t need to tive, it’s a way to conveniently sort all the provide the answers to all things, just content, primarily photos, based on your answers to some of the things people connections, locations, etc. For Facebook, care most about. The result is an immethe key will be figuring out how to expand diate threat to sites like Yelp® and Open the scope of content that is served up in Table® for reviews and recommendations. the search results, while still maintaining Looking forward, it this will also be a threat relevance. to search engines for market share of long-tail searches, which are searches Exactly how Facebook will start blend- that contain multiple keywords. Consider ing in content from brands remains to be the following example in the figure below: Current WAY OF SEARCHING: STEP STEP STEP STEP STEP ® ® 5 6 Designing with Grids
- 6. FUTURE WAY OF SEARCHING: STEP Think about how impactful it would be if Facebook could give relevant, real-time answers to this type of question. Google would lose those types of searches overnight. Facebook GS can’t give the full result just yet for the above query, but it will in the future. The implications this has for all brick and mortar locations are huge. Companies will need to ensure they provide Facebook with all the information that customers would factor into a purchase decision. From our Italian example above, this would include seating capacity, real-time reservation availability, operating hours, restaurant genre, and location. From a tactical perspective, this would mean that just like fully filling out Google+ and Bing Local profiles, owners will need to ensure technical site markup like Open Graph tags and schema.org are included in their site coding. These act as behind-the-scenes signposts to search engines and Facebook to help them recognise things like an address in the contact us section of a website. This helps take out some of the guesswork on the part of the engines by telling them explicitly how to treat the information contained in address tags, or phone number tags, etc., and display it properly to consumers. For Facebook to get to the point where it can confidently answer Bob’s Italian restaurant query, the social giant is going to need a lot of help from brands in the site markup arena, but what this does is present an opportunity for those brands and companies that are quick to act, as they will be the ones to reap the rankings rewards. Content creation for Graph search in the future Local profile info is a no-brainer, but what about content creation going forward? Content strategies are obviously unique to each business or brand. If done correctly, they are based on audience and search behavior insight and should resonate with the target market in such a way that the actions produced align with the brand’s business objectives. That’s a longwinded way of saying connect with the right customers, in the right way, so that they benefit and so does the brand. This concept is not novel. People want content they find useful, be it informative or entertaining. The key question is how will Facebook decide what content is most relevant? Beyond keyword and technical context it can infer from a piece of content, Facebook will look at several main factors for ranking search results, as explained in the sections to follow. 6
- 7. Personal Profiles Will Expand to Include More Granular Data It’s no secret that people share an incredible amount of personal information on Facebook and other social networks already. We all know far too much about people’s coffee drinks, their babies eating schedules, and status updates that start out like this: “To the guy in front of me in line at the checkout…” The point is, we should expect the level of granularity in personTo build authority on al profiles to expand. Not only will you be able to fill in that you live in a certain city, work at a Facebook, a dog food brand certain company, and you have 3 siblings, (and should seek to acquire not here are links to their profiles), but you’ll also be able to fill in the fact that you drive a black BMW just any fan, but fans who or have a golden retriever for a pet. Think about are actual dog owners what this means in terms of targeting for a Mercedes dealership or a company that makes dog food. They literally will know who their audience is. This is relevant as a ranking factor because it will allow Facebook to dissect and analyze who is engaging with a certain piece of content. Using the dog food brand example, what this means is that 100 likes from dog owners on a dog video will mean more to the brand in terms of building authority than 100 likes on the same video by cat owners. Facebook profile fields will get more granular. Expect to be able to fill in things like make/ model of your car and the breed of dog you own. The result is a marketer’s dream. 7
- 8. Each Person’s Like Will Be Valued Differently Right now, all our ‘likes’ are counted equally. A teenager’s like on a brand’s fan page carries the exact same weight as the CEO of a billion dollar company in terms of determining the popularity and value of the brand on Facebook. However, for the brand, raw fan count paints a skewed picture of its audience because depending on the brand and its target audience, it may actually value the teenager’s ‘like’ over the CEO’s, or vice versa. In the future within Facebook, this will likely change. Facebook will use your connections, data, and information from other social profiles (which we will willingly provide because that’s just what we do) to create a Klout®-type evaluation of each person’s profile. Each person’s ‘like’, or comment, will carry a different weight. However, it will be more robust than Klout® because it will be able to understand not only that you have authority, but that you have it in certain areas and not others. Effectively, your opinion (your ‘like’), will vary in importance depending on your relationship with the content with which you are engaging and the people in your social profile. In turn, each of us, and brands, will be judged by the quality of the company we keep. To that end, content will be judged based Brands will be judged based on on the authority of the producer of the conquality of their fan profiles and tent, be it brand or an individual author. It will be valued using a number of factors, but the extent to which the fans primarily on follower profile. Brands with a engagement with their content. higher number of qualified fans will have a higher authority than competitors, all else What defines a qualified fan will being equal. What defines a qualified fan be different for every single brand! will be different for every single brand! So, if two brands have the exact same followers, the brand for which the follower profile is more relevant would be given a higher authority score in Facebook’s eyes. What this means is that the value a person’s ‘like’ contributes to the fan profile for company A is different than the value that same person’s ‘like’ gives to company B’s fan profile. Consider the following example of how the power of each person’s ‘like’ will vary: John is an avid coin collector. He lists ‘coin collection’ and ‘rare coins’ under his interests in his profile. He works at Rare Coins, Inc., also in his profile. Many of his friends on FB also are interested in coin collection. To Facebook, John is seen as having a certain level of authority about coins. Therefore, when John ‘likes’ the fan page of Coin Collectors Monthly, his ‘like’ contributes more ranking authority to the page than an average like. By the same token though, if John were to like a fan page for a helicopter manufacturer, being that he’s not an expert in helicopters, his ranking authority contribution to that page would not be as significant. 8
- 9. What Does This Mean For Brands? What this means for brands is that similar to how Google judges part of a website’s authority by looking at the value of links pointing to that site from others (quality over quantity), Facebook will start looking at the quality of fans a brand has and how engaged they are with the Page’s content. Note that amplifying content with paid media will continue to play a pivotal role in consistently reaching the right fans with the right message to keep them engaged. That said, factors by which fan profiles will be judged will include, among other things: How influential/authoritative they are (e.g. celebrity status) with respect to the target audience Interests, hobbies, etc. which are relevant to the brand brand Who their social connections are and what relevant interests their friends have Where they live and work and where they went to school For brands seeking to build their authority on Facebook and rank well in Graph Search, targeted influencer outreach activities will increase. This will be so much more than just traditional influencer identification because we’re no longer just dealing with trying to acquire a single person. Instead, brands must consider the implications of the social connections of that influencer. The social profile of each person will serve as validation for that person’s influence and authority and will be the measure of whether that person is truly relevant to the brand. Conveniently, this is also how Facebook will keep people from trying to game the system by buying ‘likes’ because you can’t easily fake a legitimate social profile. In summary, doing well in Graph Search will require brands to focus not only on creating quality, relevant content, but also to commit to acquiring the right type of fan. The two concepts go hand in hand, as qualified fans will be attracted to strong content. The best course of action is to forget about trying to rank well and instead just focus on providing customers with what they want. High rankings will be a natural byproduct. By Adam Westin, Associate Director, Head of Search, Edelman UK By Adam Westin, Associate Director, Head of Search, Edelman UK Follow on Twitter: @adwestin Add on Google+ Find on Linkedin Follow on Twitter: @adwestin Add on Google+ Find on Linkedin Adam leads Edelman’s search practice in Europe, focusing on driving real results with ROI-driven strategic thinking and insights-led content creation. 9 10 Designing with Grids