To contain costs, it is essential to understand which factors drive costs over the longer term and can be controlled. In studies of software development, as a research community, we have not done an adequate job of differentiating causal influences from noncausal statistical correlations. In this podcast, Mike Konrad and Bob Stoddard discuss the use of an approach known as causal learning that can help the Department of Defense identify which factors cause software costs to escalate and, therefore, serve as a better basis for guidance on how to intervene to better control costs.