Software development control panels : concepts, a toolset and experiences

Abstract

This thesis discusses a concepts, guidelines, a toolset and experiences with defining and implementing on-line visual displays of software development progress, control panels. Control panels group sets of measurement visualizations providing a multi-dimensional, integrated view of development status. Control panels are developed at various organizational levels, serving the information needs of different stakeholders. Navigation between different control panels enables users to drill down to view more detailed information on, e.g., a particular project or on the progress of a particular development activity.

In the constructive part of the work, I describe concepts, and a toolset that can be used for defining and implementing control panels. In addition, I provide guidelines for grouping measures into control panels, and discuss how measurement programs can be partitioned into control panels. The toolset, built as a "proof-of-concept" consists of a central measurement database, a visualization client, and a set of data importers that transfer data from external corporate repositories into the measurement database.

The empirical part discusses our efforts to define and implement control panels in four high-technology companies in Finland. Framed in the context of "constructive action research", the cases are first described individually, according to the main steps of the action research cycle, then compared in a cross-case analysis. Our results—we succeeded in implementing control panels in only one of four organizations—indicate that the development of control panels requires a substantial investment and commitment by the organization, but that it can be feasible even in small organizations. The main obstacles hindering the implementation were organizational and human in nature, and more related to measurement program implementation in general than to tool support per se. Interestingly, the organizations did not consider measurement definition, i.e., the selection of "metrics" to be problematic. Success factors identified included management commitment, dedicated measurement personnel, and organizational change management.

While limited, our experiences with the use of control panels show that they were felt to be more valuable by management than by the collocated project personnel. We thus hypothesise that the usefulness of control panels increases with project size and geographical distance. In the future it would be interesting to develop control panels for distributed software development projects, in which informal status communication is more difficult.