If you're involved in any projects that involve the implementation of statistical models or data analysis in general, you should read ComputerWorld's excellent article, 12 Predictive Analytics Screw-Ups. It's an excellent collection of common mistakes that can occur, and their outcomes (or lack of outcomes) in some real-world examples. While some of the mistakes seem obvious in retrospect (using predictors that are just transformations of the outcome, failing to clean data, fitting nonsensical models), they can be all easy to make in practice.
Learning from what goes wrong is an excellent way of avoiding future mistakes, but too often we only hear about the sanitized successes. Check out the article at the link below.
ComputerWorld: 12 Predictive Analytics Screw-Ups