Using a Statistical Method to Compare Agile and Waterfall Processes Performance

Abstract

The agile principles and methods have become an important contribution in the Software Engineering area. Several works reported successful cases of agile adoption in the software industry and academic environment. Organizations can adopt the agile methods into their software development process using incremental or big bang approaches. The incremental approach allows the team members and stakeholders to assimilate the agile principles and techniques progressively. However, the use of this approach can omit supporting agile practices. In this case, process specialists usually recommend the big bang approach that allows teams to have a full agile practices experience. The current work contributes reporting an alternative approach based on the parallel development of projects using plan-drive and agile based processes. In this case, two development teams using different processes developed each project simultaneously. A set of ten real small projects were successfully delivered using this proposed approach and their respective results analysis, in terms of spent hours and defects, was performed using the Wilcoxon signed-rank test that consists in a nonparametric statistical method.