All experienced project managers know (and the inexperienced will learn soon enough) that software projects often go wrong. People also tend to agree that agile works, if done properly. But why exactly is this? Why do projects fail? What is it exactly that agile methods yield better results? In this talk we will attempt to present a rational model-based approach to answering these questions. The goal is to remove some of the mystery and hand-waving around project management and provide more accurate mental models for dealing with complexities around them.