Facebook Graph Search: First Impressions
Facebook have always been in the possession of a vast amount of rich and meaningful data. The information accumulated is not limited to personal data, but also the relationships between “entities” such as people, objects, places, companies, photos and events. The interconnectedness of these entities on Facebook is the fundamental principle of the “Open Graph”, on which Graph Search is built. This is also the main principle of Google’s “knowledge graph”, and both systems aim to bring sematic meaning to information on the internet.
Until now the many relationships between objects on Facebook have been used primarily for targeting adverts. That is because Facebook user data makes it possible for companies to identify and target adverts towards very specific demographic groups. For instance being able to show wedding dress adverts to only recently engaged women.
Talk of using this data for the purposes of search has long been discussed, and on the 15th January Mark Zuckerburg announced “Graph Search”, Facebook’s very own search engine. In that announcement Zuckerburg stated “this is the coolest thing we’ve done in a while” and that Graph Search would soon be the “third pillar” of Facebook alongside the News Feed and timeline.
The Beta Test
This morning upon logging into Facebook I was notified that I could now use the new “Graph Search” feature, having joined a waiting list for the limited beta earlier in the week.
The New facebook search bar
Zuckerburg had made it clear that Graph Search was not a new web search of the kind we are used to with Google. That at least seems to be true; searches are indeed constrained to the data within Facebook ecosystem, and only the likes, shares, photos and other data people have shared with you. But the data available is far greater than simply your immediate social group, as people’s privacy settings are often set to “everyone” or “friends of friends”. Add to that the many millions of businesses, places and various other pages on the social network, and the power of Graph Search becomes apparent, and it’s possible incursion into the territory of Google.
The Art of Suggestion
Graph Search is heavily reliant on the formatting of search queries. As such Facebook gives you a big helping hand in the form of search suggestions. Typing in a term prompts a number of suggested queries which you must to use to perform your search. They’ve done this because the results of each search cannot ambiguous; they will always return results for specific entities and specific relationships based on actual data. This results in 2 stages to the search process, starting with the suggestions necessary to disambiguate whatever you type in.
As an example, searching for “friends who ski” suggests that I mean “My Friends”, as it’s possible to search based on other peoples friends. Many other possible matches are suggested, which as you can see in this case are not particularly useful. However, when Zuckerburg is using the word “serendipity” to describe a search engine, this kind of thing is expected.
Teething problems
As with any major product launch there are going to be problems, and to be fair this product hasn’t technically launched yet. Clearly the relationships between data on Facebook will grow and as it does relevant search results will increase in number and become more accurate.
The omission of any kind of spell check in Graph Search is difficult to forgive and will certainly lead to abandoned searches and confused users. I hope this issue is addressed early on as it great affects usability. The following example shows what happens currently with a misspelling, performing the same search with the correct spelling yeilded the desired results.
While we’re on the subject of confused users, with a feature so reliant on its search suggestions, this kind of incomprehensible suggestion really isn’t going to cut it:
Another big issue, at least in the beta, is that only certain types of entity are available to search. “Types” are another fundamental of graph search by logically grouping similar based on common attributes. All entities must be grouped into types in order for them to relate to other entities. At the moment there are a limited number of “place types” shown in this screen shot.
Clearly there are many more types of place than this, and as such Graph Search fails to recognise relationships of other types of place. For example, if you search for “web design companies my friends like”, because Web design companies does not exist as a type in the same way restaurants or shopping malls, Graph Search simply gives you a list of page names that match that term and an option to search Bing.
The results page
The social search results page provides lists of links to people and pages within Facebook that match your query. It’s not entirely clear how these are ranked. You’d expect that likes and shares would be a major factor, and I’m sure they are. However, Facebook knows who you interact with, what you like and lots of other information about you. As such I suspect that the results would be highly personalised for each person searching.
Each result includes the number of likes, check ins, a like button and buttons to search within the context of each result. For example if you search for museums in London, each result may have a link to search for Photos taken in that museum.
A right-hand side bar on the page includes the tools to further refine searches, a Bing map and options to extend the search. The Bing map is nicely integrated, clicking on markers moves you up or down to the appropriate result and hovering over a result on the page highlights the marker on the map. This sidebar is not always present, search results for images or videos take up the entire width of the screen, and instead you’ll find the refinement tools in an expandable box.
BING Web Search
I’ll be honest; I had assumed that Bing results would be more tightly integrated within the social search results. Instead Bing results are only triggered for certain searches as a kind of last resort option. Bing will clearly serve the purpose of catching more general queries and providing traditional search results from the web. It is worth noting however that Facebook likes and Shares seem to play a major role in the ranking order of the Bing results. Likes and shares are shown for each result and in general they do progressively decrease as you move down the rankings.
This partnering with Facebook could be a major win for Bing, and I have the feeling the Microsoft search engine will be a much more prominent player in search and SEO in 2013.
What does all this mean?
Graph search certainly has massive potential, but we do have to ask ourselves if people will use it and will they actually "get" it. There are also the inevitable issues of privacy and accuracy, which will likely affect it's usage. That said, the ability to search for and filter results in such an unambiguous way is very impressive. It is exactly this kind of targeted and semantic search that Google are trying to achieve with their knowledge graph. This is why HTML5 Rich Meta and the data highlighter tool are so important to Google’s future progression. In this area at least, with a ready made and ever expanding wealth of meaningful data, Facebook Graph Search has a major head start.
For business owners running a Facebook page, it’s time to be on your game when Graph search is finally released. You will really need to make sure your Facebook page is up to date with all the data you can add, making sure you’re in the correct category, have the right page name, URL and all the other data people may be searching for. Another key tactic will be getting those people with lots of friends to like you, that way your page is going show up in far more “pages my friends like” type queries, and the growth could be exponential. This is clearly going to open up a lot of opportunities and new social strategies, I for one look forward to seeing what happens next.
