In this paper, data envelopment analysis variable returns to scale (DEA VRS) model is applied to data collected on 79 software development projects from a leading CMMI level 5 organization. We divide overall software effort into software development effort, software quality conformance effort (EoC), and software maintenance non-conformance (EoNC) effort due to poor software quality at delivery time. Partitioning effort into software development and software quality metrics provides us a comprehensive model to measure productivity of software projects and to identify best practice projects. Some of positive productivity drivers from the DEA best practice efficient projects point to good customer rapport and application familiarity. Inefficient projects had problems such as customer requirements volatility, and the use of unfamiliar technology. The DEA results identify 12 “best practice” projects that can be emulated for software process improvement. Additionally, our results point to approximately 50 % potential for productivity improvement in software projects to get to the level of “best practice” projects. This study shows that including EoC and EoNC as inputs has a positive impact on the best practice frontier.
All Science Journal Classification (ASJC) codes
- Information Systems
- Business, Management and Accounting (miscellaneous)