Conway’s Law states that “organizations that design systems are constrained to produce systems which are copies of the communication structures of these organizations” [Conway 1968]. Similarly, Fred Brooks argues in The Mythical Man-Month that the product quality is strongly affected by organizational structure [Brooks 1995]. Software engineering is a complex engineering activity. It involves interactions between people, processes, and tools to develop a complete product. In practice, commercial software development is performed by teams consisting of a number of individuals, ranging from the tens to the thousands. Often these people work via an organizational structure reporting to a manager or set of managers. Often failure predictions are obtained from measures such as code churn, code complexity, code coverage, code dependencies, etc. But these studies often ignore one of the most influential factors in software development: specifically, “people and organizational structure.”

With the advent of global software development where teams are distributed across the world, the impact of organization structure on Conway’s Law and its implications on quality are significant. In this section, we investigate the relationship between organizational structure and software quality via a set of eight measures that quantify organizational complexity [Nagappan et al. 2008]. For the organizational metrics, the metrics capture issues such as organizational distance of the developers; the number of developers working on a component; the amount of multitasking developers are doing across organizations; and the amount of change to a component within the context of that organization. Using these measures, we predict failure-prone binaries in Windows Vista. Some of the organizational measures are the following (for a more detailed list, refer to [Nagappan et al. 2008]):

The measures proposed here attempt to balance the various hypotheses about how organizational structure can influence the quality of the binary, some of which seem to represent opposing positions. A high-level summary of the hypotheses and the measures that purport to quantify these hypotheses is presented in Table 23-1.

Using random splitting on the entire Vista code base for 50 random splits, we obtained the precision and recall values as shown in Figure 23-2. The figure indicates the uniformity and consistency of the predictions of precision and recall. The values show little variance, explaining the overall efficacy of the models built and the ability of organizational metrics to effectively act as predictors of software quality. The precision and recall values were 86.2% and 84%, respectively. This is, as from our prior studies, the highest precision and recall in terms of predicting failures. This provided evidence that understanding the team structure in the organization is a crucial factor in predicting and understanding software quality. The interpretation lies in the fact that the organizational metrics were much better indicators of code quality, more than actual code metrics. This stresses the importance of getting the organization and team right for doing software development. These results also illustrate the impact of organizational structure on quality and the importance of planning for reorganization of teams to minimize the impact on software quality.

Figure 23-2. Precision and recall values for predictions using organization structure [Nagappan et al. 2008]