Automating Feedback for Requirement Changes in Agile Systems Development

Abstract

Context: Managing requirements in large-scale agile systems becomes harder and harder as companies and products grow in an ever-changing environment where quick and effective responses to change are highly important. While previous on automated analysis of requirements documents exists, there are no obvious ways to automatically validate changes in requirements in agile environments where multiple changes are made in parallel to each other. Meanwhile, automated unit and integration tests are essential for agile practices to keep the main branch of the workflow clean. Objective: This thesis investigates what can be done for requirements when it comes to automation, specifically in terms of requirements management and feedback. This is done through the investigation of the main problems of requirements verification and validation in agile systems development, when changes are often made in parallel to each other, that can be addressed using automated feedback. Then, an exploration of which potential solutions can be incorporated to improve the quality of the requirements, when changes are made in parallel to each other, are suggested and evaluated by requirements engineering experts. Method: When conducting the study, three cycles of design science research were applied. Each cycle consisted of a regulative cycle containing the following steps: problem investigation, solution design, solution validation, solution implementation, and evaluation. The data collection of the thesis consisted of interviews, workshops, a survey, and a literature review. Conclusion: Multiple solutions to the discovered problems were identified and rated valuable by requirements engineering experts. The solutions were categorized as interdependency feedback, feedback on change history, language feedback, or technical feedback based on the evaluation of requirements that they provide. These solutions are provided in the form of guidelines for practitioners and researchers to apply the findings to their own tools. The guidelines are constructed from the solutions rated as valuable by experts during the research. Suggestions of the points in time, in git-based workflows, where they should be applied, to which receiver to send the feedback, and a discussion of abstraction levels are further included in the guidelines.