Facebook Rank – A Review of Graph Search Ranking Factors

February 1, 2013

Facebook Graph Search has the potential to change how information about people, places, events, applications, and products are found and evaluated. This article will compare web search to social search to identify the differences between the two systems and then move on to a discussion of the ranking factors within a social search algorithm. Facebook Graph Search is still in Beta and subject to change. However Facebook has made statements about Graph Search as well as applied for patents that give an idea of what ranking factors may apply. It’s likely the algorithm will go through several noticeable changes during the course of the first year before stablizing into a schedule of incremental improvements followed by the addition of new services.

Internet Search
The current method for finding useful information is a citation based system (links) that analyzes the entire Internet and tries to algorithmically cull the most popular and useful information. An inherent flaw  in this method is that the citations are mostly created by the web publishers themselves, not the end users. To counter this bias the search engines have taken into account other signals to detect attempts to manipulate the algorithm as well as to correct for the inherent bias and weaknesses of a citational ranking system. The most important weakness of an Internet-wide citation based system of organizing the web can be said to be the inherent bias toward what web publishers feel is important, which may not reflect the bias of the searchers themselves. 

Searcher Bias
The bias of the searchers is what makes user generated content on sites like TripAdvisor, Amazon.com, Chow.com and Yelp so powerful, it’s the people themselves telling the others about products, destinations, events, food, and travel.  A comparison of the dynamic and timely nature of a system based on user generated content  versus a web based system of organizing information highlights the potentially stale search results of Internet Search, one that may not reflect the cultural tastes of those using the Internet Search system.  This lopsidedness is evident in various measurements of trends and buzz offered by the search engines themselves, which display trends by what people are searching for, not by what information is being cited the most. But if trends are related to what people are searching for, then how can a search system based on what websites are linking to (and not on what people are searching for) keep up?  

Google and Bing have attempted to keep up with search patterns by bolting on local search variants to their regular search experience, with Microsoft going so far as to add a side panel for interacting with “buddies” when searching. I won’t attempt to describe every nuance, that’s beyond the scope of this article. The purpose of this portion of the article is to highlight the key differences between Internet Search and the possibly disruptive Social Search that is coming. Now let’s take a look at a Social Search Citational System and the ranking factors associated with it.

Social Search Citation System
It may not be necessary to index the entire Internet in order to find answers if a system for retrieving useful information is pulling the data from a set that is entirely created, curated and validated by the consumers of the information. Rather than rely on the opinion of web publishers, a social search system relies on the judgement of the end user and their friends.  The accuracy and usefulness of the information may be improved because some of the citations used to generate the search results are created by peers of those who are searching. This may be more relevant than a citation by a web publisher who has no geographical, generational, or any other societal tie to the person who is doing the searching. In other words, if you want to know what is the best software for creating music on a laptop, the opinion of musicians and your peers may be a more accurate source of information than the opinion of web publishers who typically lag behind the experience of those who are using the software. Social search is a peer based system but Facebook has added more to the Graph Search in order to make it more useful.  

Facebook Graph Search Ranking Factors
Here is a link to a Patent Application filed on August 7, 2012 by Facebook entitled, Search and retrieval of objects in a social networking system. It describes a system that leverages Facebook data for the purpose of generating useful information via different algorithms.

The Social Graph itself – Connections
The social graph is a measurement of connections between a Facebook user and other objects within the system. Some connections are closer on the graph than others. One of the important ranking factors is the distance between the Facebook searcher and the object they are searching for. This has important implications or a business or individual who is seeking to be found within a social search environment.  Here is what the Patent itself says about the relationship of Facebook users and the objects being searched for within the Social Graph:

12. The computer-implemented method of claim 9, wherein the social graph has nodes corresponding to objects and edges corresponding to relationships of the objects, the user being represented as one of the objects.

Here is what the patent says about the Connections in distance and similarity: 

9. A computer-implemented method comprising: receiving a query from a user; submitting the query to a remote social networking system; and receiving, from the social networking system, a combined result set comprising objects matching the query, the combined result set comprising objects obtained from a plurality of search algorithms performed by the social networking system; wherein at least a plurality of the objects of the combined result set are ordered based at least in part on measures of affinities of the user for the objects, an affinity of the user for an object comprising at least one from a group consisting of: a distance on a social graph between the user and the object, and a similarity between the user and the object.

Action item: What connections means to you
Make it easy for Facebook users to interact with your website, including likes, sharing, signing up, etc. Create a profile or group on Facebook. Because connections are going to play a strong role in Facebook Graph Search, it’s important to start building those. Posting content, particularly content that is share-friendly is important. The more people share and like your content the more connections you build. Encouraging likes and friend building is going to be important. Reaching out to “friend” those strongly identified with a particular trade, hobby or niche should be helpful in expanding your circles of influence. 

What not to do – SpamLikes
Do not take shortcuts for building likes and friends. Inevitably the industry for building likes and friends will increase, with fake Facebook profiles, cash for likes, Facebook Friend rings and the like growing. Facebook is cleaning up the Likes ecosystem to improve accuracy by removing fake likes, to remove false citations. A citation based algorithm will function better when most of the citations are genuine. Facebook is already cleaning up SpamLikes, as noted in this discussion at WebmasterWorld.

Physical Distance
“6. The computer-implemented method of claim 1, wherein a measure of an affinity of the user for an object further comprises a physical distance between a geographic location associated with the user and a geographic location associated with the object.”

 Physical distance likely relates to events and local type searches. Facebook Mobile is a big driver for Facebook so it makes sense to roll geodata into their algorithm. If you are a restaurant, event space, a non-profit with fund raising events, a brick and mortar, even a gas station, having a geographically relevant Facebook page is going to be important.

Global Importance – Forget VIP, Facebook introduces the GIP
There are some Facebook users who will be designated as Globally important. This means that their influence will be more important than most other members. A famous chef, an author, a tastemaker, people who have large circles of friends and are strongly associated with a niche, topic, event, business will be designated as Globally Important. Here is what the patent says about Global Importance:

The global importance search sub-module 143 identifies, among objects considered to be of global importance, objects that match the query. The globally important objects need not have any specific relationship with the searching user, but rather are considered to be of general interest across users of the social networking system 100 as a whole. The objects to be placed within the globally important group 131C may be identified in different ways. For example, objects may be considered to be globally important if they have been accessed (e.g., viewed), tagged, posted, marked as having one or more fans, or otherwise designated as being of interest, some pre-specified number of times. Alternatively, actions such as accessing/tagging/posting may be used to calculate a score, and the objects with the top N scores may be selected as globally important objects, for some integer N. In one embodiment, the globally important objects, or references thereto, are stored in a distinct portion of the object store 110 so that they are readily available for searches by the global importance search sub-module 143. The global importance search sub-module 143 then selects as its result set, from among the objects of global importance, those objects that match the query according to some match algorithm.

…The last of the matching objects 210, an object 210C for a page 115 dedicated to Michael Jackson, is not of specific relevance to the searching user–e.g., is not a first-order or second-order connection of the user–but is considered of global importance given its sheer aggregate popularity

Global Importance – Action points
Not everyone can become a Globally Important Person (GIP?).  According to the patent, how many times your profile has been viewed, tagged, etc. all play a role in achieving the GIP designation. Essentially, the more people interact with your Facebook presence, the higher the possibility of being designated globally important. At the moment it appears to be an algorithmic process, which should allow it to scale.  Having your profile easily found and discovered and offering multiple ways to interact with it are key, including giving visitors something to like, friend, share, view, etc. But it’s not just about becoming Globally Important. Obtaining a reference, citation, link, by a Globally Important person or profile may be just as important. For example, having a globally important chef link to your New Orleans restaurant website or profile may help you rank better for a search for the best restaurant in New Orleans.

The Similarity Measure
The similarity measure is a way to match the person making a query to Facebook users, groups, or other “objects” in Facebook that share an affinity. This provides Facebook a way to provide results drawn from a data set that is larger than your immediate social group. Similarity measures includes hobbies, topics, events, music, food, etc. that you have indicated an interest in. Here is what the Facebook patent says:

The similarity measure quantifies how likely the searching user would be to find a particular matching object to be of interest, and may be calculated in different ways in different embodiments. For example, the similarity measure may be calculated by comparing a user profile of the searching user to the matching object, such as by noting that the user profile states that the searching user enjoys golf and that the matching object is a group 114 devoted to golf. Alternately and/or additionally, the similarity measure may be calculated by determining interests of the searching user based on past actions of the searching user within the social networking system, such as posting messages related to golf or using golf game applications 112, and comparing the determined interests to information about the matching object. The physical distance, graph distance, and the similarity measure can be considered, individually or collectively, to constitute an “affinity” of the searching user for the object in question.

Action items – Similarity Measure
Creating groups, belonging to groups, and creating strong and wide ranging content about a particular topic may be useful for increasing your ability to be matched to queries by affinity. The idea is to have your Facebook profile presence strongly identified with the product, event, topic, niche. If your niche is strongly identified to side topics, it might be useful to expand your affinity with those groups, so that your profile may be connected through the similarity measure with those groups making queries. But this remains to be seen and experimented with.

Facebook Graph Search is in flux
Graph search is currently in beta and is rolling out slowly. The intent is to find the right mix of algorithmic factors. What we see now may likely be different six months to a year from now, as they adjust to user satisfaction, insights into improving accuracy, and deal with unforseen methods of manipulating the system. Facebook Graph Search may become a way to build traffic to Internet websites. So exploring ways to make your site Facebook friendly today may be important tomorrow as Facebook Graph Search eventually comes out of Beta.

 

Article by

Previous post:

Next post: