I have replaced uses of the Standard Deviation with the Mean Absolute Deviation at work on several occasions, for just the reasons described here. It often leads to substantial improvement in predictive validity, in some cases fixing a broken process.