Profiling prerelease software product and organizational performance

Abstract

Background: Large software development organizations require effective means of quantifying excellence of products and improvement areas. A good quantification of excellence supports organizations in retaining market leadership. In addition, a good quantification of improvement areas is needed to continuously increase performance of products and processes. Objective: In this chapter we present a method for developing product and organizational performance profiles. The profiles are a means of quantifying prerelease properties of products and quantifying performance of software development processes. Method: We conducted two case studies at three companies-Ericsson, Volvo Group Truck Technology, and Volvo Car Corporation. The goal of first case study is to identify risky areas of source code. We used a focus group to elicit and evaluate measures and indicators at Ericsson. Volvo Group Truck Technology was used to validate our profiling method. Results: The results of the first case study showed that profiling of product performance can be done by identifying risky areas of source code using combination of two measures-McCabe complexity and number of revisions of files. The results of second case study show that profiling change frequencies of models can help developers identify implicit architectural dependencies. Conclusions: We conclude that profiling is an effective tool for supporting improvements of product and organizational performance. The key for creating useful profiles is the close collaboration between research and development organizations.