Characterizing project scope attributes and their influence on the software estimation process

Abstract

Software estimation is a valuable practice, allowing organizations to predict and adapt for better project performance concerning cost and resources. Inaccurate estimates can lead to misalignment between stakeholders and loss of confidence by management in the software development process. This thesis focuses on software estimation processes and the relationship between scope attributes and their usefulness to the estimation process.

Semi-structured interviews were conducted with industry practitioners to collect data on their experiences of the estimation process. The experts characterized key steps involved in an estimation process and responded to a five-point scale survey to gauge the consideration of scope attributes, i.e., functionality, dependencies, and newness during estimation. Open-ended interview style questions were asked to understand the meaningfulness of system topology during estimation.

Based on the interviews it is identified that expert judgment based subjective assessment is the commonly used method for estimation. Additionally, release level estimation processes appeared to be informal and intuition-based, lacking analysis of systemic characteristics of project scope especially dependencies. Dependencies were often missed or insufficiently considered during the estimation phase as they required more thorough analysis and were considered only at later stages of project planning. Findings from the interviews also suggest that estimation mostly considers the prioritized scope items, making it inadequate to perform a topological assessment of dependencies. These and other findings are limited to the interviews conducted as part of this thesis. It opens possible avenues for researchers and practitioners to expand on the work by conducting more structured interviews or case studies to gather data on estimation processes, especially representation and assessment of scope dependencies within such processes. Additionally, experimental research can be conducted to determine how knowledge of dependencies impacts estimation processes.