Assessing the Quality of Software Requirements Specifications

Abstract

Software requirements specifications (SRS) are hard to compare due to the uniqueness of the projects they were created in. In practice this means that it is not possible to objectively determine if a projects SRS fails to reach a certain quality threshold. Therefore, a commonly agreed-on quality model is needed. Additionally, a large set of empirical data is needed to establish a correlation between project success and quality levels. As there is no such quality model, we had to define our own based on the goal-question-metric (GQM) method. Based on this we analyzed more than 40 software projects (student projects in undergraduate software engineering classes), in order to contribute to the empirical part. This paper contributes in three areas: Firstly, we outline our GQM plan and our set of metrics. They were derived from widespread literature, and hence could lead to a discussion of how to measure requirements quality. Practitioners and researchers can profit from our experience, when measuring the quality of their requirements. Secondly, we present our findings. We hope that others find these valuable when comparing them to their own results. Finally,we show that the results of our quality assessment correlate to project success. Thus, we give an empirical indication for the correlation of requirements engineering and project success.