Are We More Productive Now? Analyzing Change Tasks to Assess Productivity Trends during Software Evolution

Organizations that maintain and evolve software would benefit from measuring productivity in an easy and reliable way. This could allow them to determine if new or improved practices are needed, and to evaluate improvement efforts. We propose and evaluate indicators of productivity trends that are based on the premise that productivity during software evolution is closely related to the effort required to complete change tasks. Three indicators use data about change tasks from change management systems, while a fourth compares effort estimates of benchmarking tasks. We evaluated the indicators using data from 18 months of evolution in two commercial software projects. The productivity trend in the two projects had opposite directions according to the indicators. The evaluation showed that productivity trends can be quantified with little measurement overhead. The methodology is a step towards making quantitative self-assessment practices feasible even in low ceremony projects.