A brief history of software — from Bell Labs to Microsoft Research

In the mid 1990s, I was (tangentially) part of an effort in Bell Labs called the "Code Decay" project. The hypothesis of this project was that over time code becomes fragile (more difficult to change without introducing problems), and that this process of decay could be empirically validated. This effort awakened me to the power of combining statistical expertise with software engineering expertise to address pressing problems of software production in a statistically valid manner. I will revisit some of the work we did in the Code Decay project at Bell Labs and then turn to what has been happening in this area in Microsoft in the last five years. In particular, I will trace how we have progressed from studying the data produced by product teams to validate hypotheses, to being actively involved with the product groups in creating and evaluating new tools and techniques for empirically-based software production.