An Experimental, Tool-Based Evaluation of Requirements Prioritization Techniques in Distributed Settings

In this paper, we compare and analyze three common techniques for prioritizing software requirements: Analytic Hierarchy Process (AHP), Cumulative Voting (CV) and Likert Scale Technique (LST). These techniques are based on a ratio scale and are applied within this research as part of a hierarchical approach to requirements analysis on different levels of abstraction. For the systematic evaluation of these techniques in distributed settings, a controlled experimental setting was developed and carried out via the Internet. Therefore, a particular Web application was developed and data from 199 subjects was collected. The overall results indicate that LST is a simple, fast, and well-scaling prioritization technique, whereas slightly less precise than the other two techniques. However if accuracy is an important criterion, and a more complicated and slower technique is accepted, CV has proven to be most adequate. For the AHP, particularly when used with many requirements, a recommendation cannot be given because of poor scalability.