The Social Network Analysis gives you a heuristic on the coordination needs between developers on different teams. The idea is based on Conway’s law - a project works best when its organizational structure is mirrored in software. Using the Social Network Analysis, you now have a way to ensure that your organization matches the way the system is designed with respect to the work the developers do.
The Social Network Is Build from How the Code Evolves¶
The social network paths are mined from how your codebase is developed. You see an example of a social network in code in Fig. 191.
The network is built by identifying developers that repeatedly work in the same parts of the code. The more often they work in the same parts of the code, the stronger their link in the network. Note that CodeScene filters developers with weak links since they would clutter the visualization (you can change the threshold as described in Project Configuration).
Define Your Development Teams¶
The social network lets you identify developers that should be close from an organizational perspective. The visualization in Fig. 191 shows an example of an organization with 8 development teams. If you hover over a developer, you highlight their peers that tend to work in the same parts of the codebase. You use this information to evaluate how well your organization supports the way the codebase evolves.
That also means you want to compare your organizational chart with the information in the generated social code network. Any discrepancies has to be understood.
Align Your Architecture and Organization¶
In a perfect world most of your communication paths would be between developers on the same team. That is, the teams have a meaning from an architectural perspective; People on the same team work on the same parts of the codebase. They share the same context, know each other and have a much easier time coordinating their work.
However, sometimes the world looks radically different. Have a look at Fig. 192.
The visualization in Fig. 192 shows an organization with severe coordination problems. Since the data has been made anonymous to protect the guilty, you cannot read the names of the teams or developers. But you still see that the organization has four teams with a high degree of inter-team coordination between virtually every developer. In practice, this isn’t an organization with four different teams. Rather, it’s an organization with one giant team of 29 developers with artificial organizational boundaries between them. The resulting process loss due to coordination needs is likely to be severe and lead to inefficient development, quality issues and code that’s hard to evolve.