How To Catch A Murderer, and Learn A Lot About Naturally-Occurring WOM Communities In the Process

Suresh Sood from the University of Technology, Sydney explains how to identify the natural WOM communities in moblie (3G) networks. Specifically, how do you identify the key people in these communities? (For example, if there's a crisis and a message needs to be sent out to key people right away, how would you identify who those well-connected people are?).

They looked at a visualization of A-B video calling data (A = dialing numbers; B = called numbers). Looking at nodes (in this case, callers and called) and links (the calls between A & B) and they call it "train of thought analysis". This method is often used in intelligence and criminal investigations (it was used to catch a murderer in Australia and is now being used for commercial purposes). It is also called "solving the backward problem". Solving the forward problem is creating a hypothesis and testing it. But to solve the backward problem you have the data already and then are working backwards to determine what the networks are.

Suresh explains how it's "resource intensive" meaning that it's expensive but it can also save a company a lot of money. There are also privacy concerns when using database information to identify the nodes.

Suresh seems to be talking about using the tools of social network analysis for commercial marketing purposes.