Decision-making on the priority of requirements for software product lines with multiple business lines

Purpose: Prioritisation of requirements for software product lines can be an inefficient use of management time and effort and it is difficult to make the most effective decisions for the business while managing limited development resources. This research seeks to improve the efficiency and effectiveness of the decision-making process for prioritisation of requirements in matrix product development organisations. Through examination of the software release planning process in a global IoT semiconductor company and reviewing existing prioritisation methods, the research characterises the use case, identifies key challenges and creates and validates a novel method that can address those challenges.

Research Design: Participatory action research using mixed methods was conducted in two phases: one phase to investigate, define and tune a novel prioritisation method, and the second phase to validate that method. Both phases included experiments using real-world requirements data and interviews with domain experts. Investigations were also conducted with machine learning models on the requirements data, to assess how machine learning could contribute to the problem space.

Findings: Three challenges for prioritisation of requirements for software product lines with multiple business lines are identified and defined: scale, complexity and stakeholder discordance. The Thematic Hierarchy Process (THP) is proposed to address the identified challenges. The THP has as its basis the abstraction of requirements into themes which are understood at a business level and have some identified value. Machine learning models have the potential to enhance or replace this method in the future.

Conclusion: The THP is proposed as a novel method to be used to prioritise requirements for software product lines with multiple business lines. This research concludes that the THP can accurately predict the requirements that should be included in a software release and that it can improve the efficiency and effectiveness of the software release planning process.