It looks like Facebook has launched a new effort that addresses that classic social business challenge: finding expertise. Their graph search as they describe applies to a more generic look at searching than just expertise, but they are building a new way to turn natural language queries into answers based on all the structured and unstructured data that we have all populated into Facebook.
Image via CrunchBase
This is also referred to as Social Search placed in action the largest global social network, but has appeared before in many other systems. For example, CIO magazine recognized IBM back in 2008 for its tagging service to improve enterprise searching. Notably, Social bookmarking and tagging have reduced each internal search on average by 12 seconds. This may not seem much until you understand the scale: when multiplied by over 280,000 such searches a week, compounds to a significant amount of productivity time saved.
Expertise location remains one key area for why you would want people to interact within an organization, with partners or customers. It is the challenge of finding someone with relevant expertise for any given issue. The deeper and more specific you get into what you need, the more data and analytics you need to find the answers.
For Facebook, this data comes from the obvious, all the billions of posts, pictures, location, friends and links that you see as content, and the non-obvious relationships, indirection, frequency of interaction, interaction patterns and maps. The latter is social network analysis carried out with different methodologies, with a strong reliance on graph theory.
While graph search will likely create a new stream for Facebook (assuming a similar model to Adsense appears with the searches), and will apply well as yet another new approach to expertise location business directories – think yellow pages – there is a bigger challenge that Facebook does not solve.
Locating relevant information keyed on relationships
Facebook’s answer would be simple: put your entire company, all employee and customer interaction on Facebook. This is certainly practical for a wide variety, particular small businesses, especially if we all understood the privacy implications.
Most organizations, however, want to keep expertise information within their boundaries, because having that information is a competitive advantage. The reality then is that we need expertise location to stay competitive in today’s fast paced world. It matters in sales getting questions answered; it matters in marketing to reach influencers; it matters in product development to get deep input; it matters in the C-ranks obviously.
It is when you start looking into the complexity of the context often needed in these challenges when you begin to understand just how much more we need to know that beyond “Who do I know in Pasadena?” It becomes “Who do I know in Pasadena that has worked on this project with some success, know about this particular industry, is available to talk between Tuesday and Wednesday that I am there, and has the ability to deliver in following weeks?”
There are typical search related issues that describe depth of inquiry:
- Can we phrase what we are looking for? What if we aren’t entirely sure what it is?
- Can we bring in the context behind the inquiry? (without too much effort) What does it need to know about us and our work to describe that context?
- Where is that context that we know of stored and can you get access to that to bring into question? (going back to the structural aspects)
- Are we intentionally looking or need to browse and discover through serendipity?
- Do we understand how to evaluate the results we get?
- Can we refine the results (back to the first question)?
These pose challenges because we don’t always know what we want, or sometimes just don’t recall enough context. We need our secondary brains (aka computers) to help us recall some of that; do the footwork and cross-referencing; add the extra information; look up schedules; look up skills metadata; the list goes on. This is the reason why we kept making our computers smarter in the first place: to help us recall, remix and respond.
How are these any different than in standard Web search? For one much of this context can be keyed from relationship graphs. It becomes a social search because you are adding a particular dimension of context: the various types of relationship information that is relevant, applies and adds value to the context of what you need.
Then there are structural aspects that become part of the query:
- Across established relationships solely inside the company – each person’s known or established social network relationships across the company, simply in an online version.
- Across established and implied relationships inside the company – an employees established network but also implied network of people they may be associated with (e.g., being in the same workgroup, in the same office, in the same knowledge field), showing what information the system implied about that relationship
- Across established internal and external business partner relationships
- Across established internal and external customer relationships
- Across established internal and external industry & coopetitor relationships
Very often, that information is in a combination of your personal devices, your software applications, and the data you have access to on your organization’s network and beyond. The more we move down the list, the more we need to rely on defining what those relationships mean beyond one’s organization, and how third-party social networks like Facebook, Twitter, Google+, etc. become the pass through for the actual business relationship. We realize that there are many parties and places involved in the query.
Imagine, being able to discover through a search of relationship graph across multiple businesses and networks. The information about your social network map is out there, just spread across possibly many different software systems.
Yet, every social network company wants you to think that their domain of users will cover all you will ever need. There has never been a need for a single monolithic social network, regardless of the dreams of entrepreneurs in the tech industry. Yet, it is entrepreneurship that we really want. We want many models to be in place, and new startups and larger organizations alike to keep pushing innovative and competitive ideas.
The only true way to connect across all these is to have standards that interface across social networks: the one you run inside your company, the ones you use on the Web, the walled garden ones you use for specific activities, etc.
There will always be different fields, facilities, and factors in various social network systems, but the basics are becoming immutable. They have connections to people; they have metadata about these relationships; they have interactions across relationship; there is metadata about those interactions; there are even more compounding factors that emerge from those such as reputation and trust.
The information exists; we just need a way to connect them. The Facebook graph search system is a necessary improvement, but for many businesses, we need better enterprise search systems that incorporate this wealth of network information. This is typically the information we worry most about loosing, with employee turnover and job changes. This is the tacit volatile information is in our heads behind our collaborative relationships that can be very pertinent, for example in opportunity identification in Sales, influencer endorsement in Advertising, business partner collaboration, or project team recruiting in any kind of project management
The on-going debate right now if social search is better than web search focuses primarily on which company has a better business, Facebook or Google. Analysts from Opus Research, Altimeter Group, Pivotal Research Group, Gartner, Forrester Research, and others are debating different points that consider the results as well as third-parties who are impacted. For example, Susan Etlinger from Altimeter Group points that this shifts the business model of other search systems like Yelp where some of the context is missing. Yelp is social, relying on online user-generated comments and reviews of restaurants, an approach pioneered by Zagat years before. However, according to Etlinger, it doesn’t have enough context: “you really don’t know who the people are who are commenting.”
What Facebook has done successfully for years is encouraging people to share that context freely. You simply can’t do it overnight because it really does rely on many interactions over time. The challenge beyond collecting context is building the best algorithms to understand meaning and bringing the right context in response to specific queries. This works well for the public space of a massive social network where there may be a lot of discussion about the same topic over and again, but again in the enterprise, there are inherently fewer people who can contribute that information.
The way for businesses to succeed is to find ways to combine both this public context, as well as their own internal context. That means having an internal social search as well, and more importantly building interfaces between internal and various (not just Facebook) external social search engines.
