A Controlled Experiment on Comparison of Data Perspectives for Software Requirements Documentation

Requirements define what the stakeholders or end users want and what the system must have to satisfy their needs. The reason for poor requirements arises from a documentation format that paints an incomplete picture. Documentation methods lacking necessary details cause requirements engineers to waste time arguing over what to do and how to do it. This causes budget and schedule overruns. This research was conducted using the E-market application domain as a test context. Three documentation data perspectives were evaluated, namely entity relationship diagram (ERD), natural language (NL), and class diagram (CD), which are frequently applied to document stakeholders’ statements. Due to the lack of research in this area, a controlled experiment was conducted focusing on requirements documentation, in which ERD, NL, and CD were compared among 103 participants. The philosophy of this controlled experiment depends on the participant’s ability to transform each perspective into the use case model. This process will represent the efficiency of each perspective. It has been found that participants who used the ERD perspective had significantly higher scores in the experiment, reported a lower number of difficulties, and used less time than those who used the NL and CD perspectives. The study results indicate that the ERD is easier to understand, more helpful, and less time-consuming in documenting requirements than the other two perspectives. In conclusion, using the ERD perspective in developing the E-market application domain could be more effective for documenting preliminary requirements than the NL and CD perspectives.