Today we focus on social networks to reach our customers or for more effective internal collaboration. But we are just scratching the surface of what they can do for organizations. Here's a look at how understanding the social graph can have broader implications for the enterprise.
For several years now, I've heard repeated noises from the enterprise information management sector about the importance of social media to their long-term strategies. It seems that, most often, visions of social media integration for the enterprise are either focused on reaching customers or potential employees by utilizing the new platforms like Facebook; and on building social, collaborative features into the corporate intranet.
Both are worthwhile goals, but as a researcher (partly) in this field, it's obvious to me that there's a whole lot more to social network utilization than these initial goals. I want to talk about a different perspective on the matter; a whole new type of analysis regarding internal and external corporate social networks which I believe will ultimately become fundamental to effective resource management in large enterprises.
This new area of analysis is not really new at all... it’s been kicking around research labs, university departments and the dusty Sociology shelf at the library for years. What will be new, whenever it happens, is the application of it to the enterprise domain. What I'm talking about is recording and analyzing the social graphs created by personal interactions in the enterprise.
Online social networks are defined by people 'following' or 'friending' each other. Draw a bunch of dots on a piece of paper (representing people), and a bunch of arrows pointing from dots to other dots (representing relationships), and you're drawing a social graph.
Social graphs are remarkably informative in their own right, and tend to have similar properties in different environments, and for these reasons they have been studied for years, since at least the 1960s and Stanley Milgram’s so-called “6 degrees of separation” experiment.
We can draw these graphs based on acquaintanceship (Facebook), loose professional ties (email), collaboration history and so on. We can apply different weighting schemes to these networks based on, for instance, frequency of interaction between people; and analyze the properties of these graphs to find interesting patterns and flows of information, identify important actors and so on.
Applications of Social Graph Analysis for the Enterprise
I mentioned that the simple act of recording these graphs can be informative. It turns out that if two people are linked in such a graph, a third person who is linked to either of them is more likely to be linked to the other. (In other words, it's more likely that a friend of my buddy is also a friend of mine, than it is that some other random dingbat, unknown to my friend, is a friend of mine.) We tend to cluster together, and our social networks tend to be reasonably tight.
This clustering phenomenon is well known, and indeed pretty intuitive. But consider this phenomenon in light of collaborations within the enterprise. We might analyze a collaboration network in the enterprise, and find that it was rather incestuous -- that the same people tended to collaborate on projects. We might discern, algorithmically and analytically, that expertise in different locations was not being utilized correctly, because the people who should be collaborating simply don't know of each others' existence, or of each other's expertise.
We might try to work out why, and analyze the communications social graph (say the network generated by emails sent between locations and departments). We might find that collaboration between locales was a function of communication between those locales, and that the right people who should be collaborating didn't know about each other because there simply was no communication between their locales.
This type of analysis is the domain of the enterprise social graph. Indeed, it's hard to see how certain optimization problems like these, which could just as easily remain hidden, can be solved without graph analysis and modeling.
Give the Postman a Raise!
Another example is finding key personnel in an enterprise whose importance is not illustrated by traditional evaluation metrics. As mentioned, social graphs tend to be cluster-y. Another characteristic tends to be a sparse number of links between clusters. Taking an extreme example, consider a bunch of villages in the Andes, where people from one village rarely go to another village. All those people exist in a social graph in clusters... they know the people in their own village, but not anyone in other villages.
But there's a postman for the mountain, and this poor guy has to climb all over the mountains to go to all the villages to deliver mail. The postman only knows one or two people in each village, and doesn't work any harder than any of the farmers who toil in the fields. He's just a regular guy. But look how important he is to the whole operation! He's the one who spreads news from village to village.
If you were to look at his work rate and number of connections, you wouldn't think he was anything special. But if you were to take him out of the picture, the whole communication system of the mountain would shudder to a halt.
Enterprise divisions and locations aren't villages in the Andes. But, as is well known from social network analysis, there are often critical points of contact and critical individuals, which form links between different groups, and their importance can easily be overlooked in many forms of empirical analysis.
Every interaction between people or organizational units in an organization can be examined in isolation, but also in the context of a complex system. Methods of deriving useful information from social and interaction graphs are in continual development; these will eventually be translated into new analytical tools for the enterprise. This article just hints at a couple of applications where these forms of analysis might prove useful -- without question there are better, more interesting examples of how such analysis will be used. And we’re completely ignoring all aspects of temporal analysis, finding hidden influencers, and all external aspects of such analysis here (such as the networks created by interactions by our enterprise and the whole environment of suppliers, competitors etc. etc. in which organizations exist).
I’m sure you can think of some killer application for analysis of interactions. Hit me up on twitter (@johnconroy) or in the comments below.