Most Internet users use Facebook to keep up with the “Mohans” and “Kumars.” With “Graph Search,” Facebook’s latest tool box offering, social connections can be further refined and configured into affinity lists. Graph Search team members discuss why the ability to sort, search and compile recommendations from publicly shared lists is an opportunity for enhanced social interaction.


As my husband races to his computer to search for an answer to settle an argument between us, I roll my eyes and wonder how we ever managed all these years without instant search and response.

The concept of a search engine is not new. In fact, it goes back at least to the 1940s, when visionary engineer Vannevar Bush published an article in the Atlantic Monthly called “As We May Think.” In the article, he predicted that “wholly new forms of encyclopedias will appear, ready made with a mesh of associative trails running through them, ready to be dropped into the memex and there amplified.” The memex that he referred to was an adjustable microfilm viewer that he imagined would have a structure similar to what we now know as the World Wide Web. What was unique about his proposal was that he was particularly interested in the way the brain links data by association instead of traditional hierarchical storage.

Search engines have been evolving over the last two decades, but I bring up Bush’s paradigm in view of what Facebook is now attempting to do. Until recently, search results were generated by standard keywords, and fell into broad categories such as People, Photos, and Groups. The “Typeahead” feature made Search a little more sophisticated by anticipating what a user might be searching for, and feeding likely matches in a dropdown menu.

Facebook’s new internal search engine, “Graph Search,” takes personalized searches one step further by customizing results and forming relationships, just as one might in real life. Unlike Google, Lycos and other search engines that bring together results from a web made up of online information from all over the world, Graph Search is a cosy microcosm, or a “social graph” in CEO Mark Zuckerberg’s words, that creates a network of one’s stated relationships with friends, favorite locations, photos and music.

The South Asian Element

Graph Search’s dynamic “Search Quality and Ranking” team is led by Sriram Sankar. Sankar is a Stanford alumnus who worked at Sun and Google, among other companies in the Valley, before joining Facebook. His dedication to the project and belief in the product is obvious when he says, “Search is my second profession (after compilers) and I’ve been lucky to have been mentored by the very best minds in this area.” For the last three years Sankar has been working on first building the search feature from the ground up to support all of Facebook’s search ranking needs, and then building the team around it, an enterprise that he calls “one of the most satisfying things I have done.”


Sandhya Kunnatur has been part of the “Search Infrastructure” team for the last year and a half. A graduate of National Institute of Technology Karnataka (NITK)—Surathkal, and Stanford University, Kunnatur loves to spend time exploring nature and has been using Graph Search to find fun places to visit.

One of the newer members on the team is Nidhi Gupta, who has been part of the Search team for the last six months. Enviably, this is her first job after graduating from Indian Institute of Technology (IIT)—Bhubaneswar.

Kedar Dhamdhere was part of the “Search Ranking” team at Facebook in 2012. Prior to that he worked at Google for six years where he was introduced to search engine technology. An engineer from IIT-Bombay, with a doctorate in computer science from Carnegie Mellon University, Dhamdhere gives a nod to his namesake Kedar raga as he expresses his love for Indian classical music.

Nilesh Dalvi is a Research Scientist on the Search team. An IIT-Bombay and University of Washington graduate, Dalvi worked as a Scientist at Yahoo! prior to joining Facebook. It’s easy to see why Dalvi would fall easily into this line of work that connects pieces of data—in his spare time, Nilesh is an avid puzzler!

Rajat Raina from the “Search Ranking” team has been at Facebook since 2009, and has worked on ads and search teams. A graduate of IIT Kanpur and Stanford University, Raina loves following the Indian cricket team.

Mitu Singh joined Facebook in 2010 and is part of the “Search Entities” team. He has worked on a few things since starting there, “but nothing as awesome as Graph Search.” He has an MBA from Massachusetts Institute of Technology—Sloan School of Management, which, he jokes, is something he desperately tries to hide from the engineers.

This enterprising team makes up some of the cogs in the big machine that has created the new feature at Facebook.

How Does Graph Search Work?

To understand how Graph Search works, consider that as a Facebook user, you have a profile that might contain your name, birthday, where you live, etc. You represent an entity. A group to which you belong is another entity; so is the place where you live. Each of these entities is related—these “relationships” may represent friendships (if the entities are people), tags (if the entities are photos), ownership (if the entity is a group) and so on.

Consider next that every friend of yours on Facebook is also an entity that is similarly connected to various people, places or applications. You already know that you can easily find a friend or you can find a place. What if you could do it all with one search in a way that connects your social circle?

Take this search for example. I’d like to find a couple with whom my husband and I can play tennis this weekend. If I use Graph Search to execute this query, it narrows the results to three people who meet the criteria of being my friend and liking tennis.

Team leader Sankar explains that Graph Search uses my connections, contextualizes my query, and customizes the output in a way that previous search engines could not. The results are biased to suit me and only me, since the same query by another user would bring a very different cross section of results. The search can be refined by other variables, such as gender and location, but what it did is save me the bother of sorting through my friends list (which, in my case, is not that long, but you know what it would feel like if you have 5,000 friends!).

What’s more, Graph Search uses natural language processing rather than traditional keyword searches. So I was able to type a search using my everyday language and the way I naturally speak instead of skewing it to a computer-preferred language or a checkbox. Graph Search has the ability to look at every single word in the query and index all of the user’s prior actions (such as likes and comments) and make a relationship between the words before producing the results.

The Sky is the Limit

In conversation with Research Scientist Raina, I learned that scalability was a challenge for the team. The vastness of the data that is already indexed is hard to comprehend, says Raina, and deciding how to handle the anticipated growth in data, and rank the results, were factors the team had to take into consideration. So far, the team appears to have succeeded and Raina chuckles that “the sky is the limit” regarding the variables one can use, as long as whatever is being searched is within the confines of what is publicly available. Even journalists can find a treasure trove of potential sources to use in their articles. Facebook stores more than 240 billion photos and one can get information on places and interests that are linked not only to trusted friends, but to public photos of cities or landmarks as well.

The Small Business Advantage

Moving out of the personal realm, it is likely that this application could allow small businesses to create a niche for themselves, and help them compete with the giant corporations. Ideally, a small business that has an accurately categorized page with detailed information on its services can acquire a fan base that will spread its reputation through Graph Search, the Internet’s version of “word of mouth.”

Graph Search does not currently support ad searches, but advertisers would want to use this to their advantage, and Facebook would presumably benefit as well, since advertisers make websites financially viable.

That said, Facebook is still viewed as a social hangout, a place to find your old buddies from school and college and stay in touch with colleagues and friends. As blogger and e-commerce CEO Philip Rooke writes, “The user expectation is still all about social connections and the mere ability to offer targeted commerce does not mean that the user will welcome or positively embrace the new Graph Search features and become more active consumers in Facebook.”

Shh, Privacy Please!

Also challenging in creating Graph Search was privacy, an ever-controversial topic among Facebook users. According to Raina, the team decided to handle user privacy by ensuring that Graph Search would not reveal anything that was not already visible on Facebook. In other words, your current privacy settings on Facebook will be honored by Graph Search.
Of course, as users of Facebook already know, you can control who can see your friends list, but your friends control who can see their friend lists.

What does that mean in real terms to the average user?

When a user does a search, Facebook responds to the query with a list of entities whose public or shared aspects of their profile match the search terms. With Graph Search, this means that what people intended to share with their Facebook audience (friends, friends of friends, or someone with a reason to search specifically for you) is now available to someone who was not looking for you, but happened upon you in their search. Even photos that you hide from Timeline will show up in searches unless your privacy settings prevent sharing.

True, it is all information you chose to share, but you feel less exposed when you think your information is there for an acquaintance to find than if you know that your information can be accidentally discovered by a stranger. (And perhaps it is a false sense of security, but we humans are good at hiding behind that!)

The problem is not unique to Graph Search; it is a risk with anything posted online. A few years ago, when I produced the school play at our public school, a journalist covered it for our online neighborhood newsletter. It was accessible to our neighbours, but it was unlikely that anyone else was looking for it. However, it made me a little uncomfortable when it showed up with the names of all the participating neighborhood children when I was searching for something else online.

Graph Search poses the same problem. As Adi Kamdar writes in, “There’s a difference between posting information for anyone to find and posting information to be searched and sorted. If you walk down a crowded public street, you are probably seen by dozens of people—but it would still feel creepy for anyone to be able to look up a list of every road you’ve walked down.”

It remains to be seen whether the Graph Search feature will encourage users to open up more and share more of themselves and their preferences to maximize the social networking experience, or tighten their privacy settings so they won’t unintentionally show up in someone’s Graph Search.


Another way to look at Graph Search’s functionality is to question whether the privacy in fact limits our search options! Social recommendations are wonderful for local restaurants or to find best soccer program for our kids. But if you are looking for something more obscure like the “most fun things to do in Mongolia,” the most content-rich answers are not necessarily going to come from your friends.

Guillaume Decugis, CEO of, believes that ultimately “Facebook will face a dilemma. Either Graph Search remains a private, social search that sticks to friends of friends, in other words potentially entertaining but limited, biased content that will not be the most relevant. Or, Facebook makes more private data accessible to Graph Search to make it more relevant, hurting privacy advocates in the process but perhaps more importantly losing what makes its core value proposition: connecting with friends in a trusted environment.”

Facebook launched the beta version of Graph Search in January 2013 to a limited number of users. If you wish to try it, you can ask to be on a waiting list, and you will be added as space opens up. At this time, it is only available in U.S.-English with future multi-language launches in the works.

As for privacy, it remains up to you, as always, to decide how much of your life you want to share online and with whom. In a generation that seems to favor and promote those who have a presence in the social media, Graph Search provides new opportunities. Just remember that opportunity and risk often go hand in hand and handle your online presence wisely.

Meanwhile, I’m off to play tennis with one of my blurred friends pictured in this article.

Gayatri Subramaniam is a San Jose-based instructional designer and writer. She is an ardent tennis fan who believes that if she had only been taller, stronger, faster, and blessed with more talent, she would’ve been a Grand Slam champion.

Vandana Kumar is a publishing executive with a 35-year track record in the industry. She leads the India Currents Foundation as President and CEO. As a new immigrant, she co-founded India Currents magazine...