Toward Predicting Software Development Success from the Perspective of Practitioners: An Exploratory Bayesian Model

Abstract

Software development project managers may need to focus on particular issues during the development process for a variety of reasons, including limited resources. This study utilized a survey to ask software practitioners, who are at the core of development, to provide insight into some of the important early non-technical issues of software development, including those related to sponsor/senior management, customer/users and requirements management. Proposed relationships among these early items, and their relationship to software practitioners’ perception of project success, were quantitatively represented through a proposed Bayesian Belief Network. The concept of ‘success’ was derived from a pilot study of practitioners and was ‘defined’ as (a) there is a project plan, (b) the project is well planned, (c) practitioners have a sense of achievement while working on a project, (d) practitioners have a sense of doing a good job (i.e. delivered quality) while working on a project, and (e) requirements are accepted by the development team as realistic/achievable. The proposed causal model provided quantitative evidence that reaching agreement with customers/users on requirements, a high level of customer/user participation, and users who make adequate time for requirements gathering have the largest direct impacts on project success among the investigated items. The proposed model identified the following as the critical chain of events for success: (1) Having a sponsor throughout the project, (2) users who make adequate time for requirements gathering, (3) a high level of customer/user participation in the development process, and (4) agreement on requirements between customer/users and the development team.