The simple fact of the matter is that traditional efforts to predict attrition lack accuracy. Even in industries awash in customer data, correctly predicting at-risk customers has been enormously difficult. Successfully identifying these customers is all the more important because "vaccinating" a customer against attrition often requires expensive measures such as incentive and loyalty reward programs. This makes it even more critical that predictive systems become more robust.