Wednesday, February 18, 2009

The Awesome Power of Network Analysis

While CRM technology is a wonderful thing which, if its set up right, lets salespeople manage their days more efficiently and effectively, lets managers get a better picture of their operations and lets customers be treated at least consistently, there is a point where the users of a CRM system ask themselves a very simple question:

"So what are we going to do with all this information we're now capturing?"

In fact this question can be thrown at most modern business application tools. The obvious answer is Business Intelligence, a fancy way of saying business report generation. This post is about a powerful but still underutilised technique that falls into this camp called 'Network Analysis' or more specifically 'Link Analysis'. Here is the wikipedia article for those interested.

The only industries I know of that have used link analysis to its full effect are government intelligence agencies and pharmaceutical marketing companies. Essentially the method consists of looking at interactions between people and concluding information in regards to those interactions. For example, if I'm a drug company that wants to promote my latest drug to the medical community, I can either send information to every doctor in the phone book or I can work out which ones are the ones the other doctors go to for information and just send them the details. Similarly, if I am a government intelligence agency trying to disrupt communication in an organised crime ring, if I can find out the people who everyone else go to for information, knock those people out and suddenly no one is talking.

So what relevance does this have to CRM given the lack of drugs or criminal activity involved in the average implementation? Lots!

In a CRM system we capture email interactions, company hierarchies, company-company relationships, salesperson-contact relationships and so on. All of these are fair game.

For example, a customer could be playing two salespeople off each other. Such activity can easily be picked up with this technique. You may want to know who customers go to or other members of staff go to when they need to know something. Again, this can easily be discovered and while some of these things could be determined with an intelligently crafted query to the database, you can go further to ask things such as which support staff consistently work together to solve customer issues?

So lets see an example. First of all, I recommend downloading NodeXL from codeplex.

This is an Excel 2007 template which allows the magic to happen (apologies to the 2003 folk).

You can import your own data from another workbook (spreadsheet), such as one exported out of CRM, use Outlook as the source or even point the spreadsheet to twitter. Essentially you need to populate the Vertex 1 and Vertex 2 columns in the Edges workbook to make it work. When you have your data in there, press the 'Read Workbook' button on the right-hand pane and a graph will appear. You can change the layout type to something that makes sense and if it is too cluttered, you can use the 'Tie Strength' column to weed out the weaker links (the higher the tie strength, the stronger the link).

My plan was to use the CRM VPC demo for the data source for this blog but the data is so sparse in it that there is really no point, so I pointed the spreadsheet towards my Outlook pst and after filtering the tie strength to a manageable level I got the following.

So what are we looking at? Well basically the little dots (vertices) are email addresses and the links and emails sent to or from those email addresses. So the big spot in the middle is my current email address. Other big vertices represent things like work email addresses, close friends, work colleagues and so on. Through literally thousands of emails, the graph quickly shows us the important people in my life. 

Similarly, collaborations are quickly shown as triangles/quadrilaterals etc. In this graph, they are typically of the type (employee 1, employee 2, client) or (employee 1, client stakeholder 1, client stakeholder 2). All excellent information to know when re-engaging with a customer to determine who will likely want to be 'in' on the project.

Finally, the arrow directions tell us the flow of information i.e. who was sending the emails to whom. For the main spot in the middle, the links above the spot are typically newsletters and the links below are typically one-way communication ranging from my wife's work email address (she is not allowed to send personal emails from work) through to unsubscribe requests.

CRM is all about communication and the management of that communication. Network/link analysis allows us to divine information about those communications not readily discernible through other methods. Download NodeXL and have a play one afternoon and I guarantee you'll think of half a dozen different uses for it in your business before the siren sounds.

1 comment:

Marc Smith said...

Thanks for recommending NodeXL!

May I recommend running the "Calculate Graph Metrics" feature and then trying the "Auto-fill Columns" feature in NodeXL to map display attributes like color, shape, size, and opacity to data attributes like degree, betweenness, or density (clustering coefficient). This will allow the chart to display more data about the nodes and edges.


Marc (for Team NodeXL)