Analytics tapped to predict and prevent churn

IBM’s new analytics middleware promises to predict churn before it happens – perhaps with enough time to prevent it. It rapidly analyzes patterns in call activity, identifying social groups and yielding information to predict churn, target ad campaigns and detect fraud and abnormal activity.

Call pattern analysis can identify not only social groups among mobile subscribers but the "leaders" – or most influential members of each group. That's important for several reasons, one of which is the fact that, according to IBM's calculations, members of the group are more likely to churn if the leaders do and, if the leaders already use other carriers, followers are more likely to switch to the leaders' provider.

Mobile operators already mine call data to make such predictions, but historically they have stored that data in large volumes, culling analysis of it over a one- to three-month period. However, that may be too late to prevent some customers from churning, IBM said. With TABI, IBM is promising much faster analysis – on the order of days, or even culled in response to specific events.

Doing rapid analysis opens the door for other possibilities, IBM said, like for example, combining rapid analysis with location-based technology to analyze user behavior around a specific event while it's still happening – for example, examining mobile use around a football stadium surrounding a game.