My eyes nearly popped out when I read this article in Cellular News titled "Is Predictive Modeling Key to Understanding Customer Loyalty?"  

The answer appears to be YES.  In fact, a South Africa based marketing insight services company, Knowledge Factory, convincably argues that predictive modelling techniques are the key to understanding customer loyalty.

 "………Uncovering churn patterns.   To help organisation's effectively manage their churn, Knowledge Factory uses predictive modelling techniques based on sound empirical evidence and strong analytics. "It's actually highly transparent," observes Moodley, "we take a firm's customer transaction data and then apply sophisticated mathematical models to uncover the churn patterns within it."

By designing and combining a wide variety of mathematical techniques, including artificial neural networks, statistical regression and decision trees, the company can determine the propensity of any individual customer to cease doing business with an organisation within a given time period.

"The results are often quite surprising," notes Moodley, "you might be under the impression that customers want the latest phones or trendier outlets, but a large number of those leaving your network might have experienced a considerable number of dropped calls in the last six months."

Is this the Holy Grail of Loyalty metrics that all the big brands has been looking for and trying to measure - somewhat unsuccessfully on their sites via stock web analytics?

"….Once we have discovered the reasons behind the churn you are experiencing, we can use them to devise analytical customer management strategies," explains Moodley. "This is because sophisticated predictive churn models on their own are not enough and organisations need to create informed strategies from them."

Reiterating the importance of creating and implementing customer management strategies, Moodley notes that "interesting results won't change anything and companies must be prepared to action the strategies we help them devise to really tackle churn."

Much of Knowledge Factory is based on the South African Market - like ClusterPlus*, MarketPlus*, ESP, etc.   I'm not sure Knowledge Factory will be immediately useful here in the US - or then again - maybe the technlogies can be used here - just with different data.

I would definitely say that Customer Loyalty metrics are not built into Web Analytics - it's difficult, if not impossible to measure loyalty using logfiles and javascript.   Maybe Knowledge Factory does have an answer worth taking a closer look at.