Managing the Sources of Uncertainty: Matching Process and Context in Software Development

Abstract

There is increasing interest in the literature about the notion of a contingent approach to product development process design. This interest stems from the realization that different types of projects carried out in different environments are likely to require quite different development processes if they are to be successful. Stated more formally, a contingent view implies that the performance impact of different development practices is likely to be mediated by the context in which those practices operate. This article provides evidence to support such a view. Our work examines whether projects in which the development process matches the context achieve superior performance. We focus on two sources of uncertainty that generate challenges for project teams: platform uncertainty, reflecting the uncertainty generated by the amount of new design work that must be undertaken in a project; and market uncertainty, reflecting the uncertainty faced in determining customer requirements for the product under development. We develop hypotheses for how these sources of uncertainty are likely to influence the relationships between a number of specific development practices and performance. We then test these hypotheses using data from a sample of 29 Internet software development projects. Our results provide evidence to support a contingent view of development process design. We show that in projects facing greater uncertainty, investments in architectural design, early technical feedback, and early market feedback have a stronger association with performance. The latter relationships are influenced by the specific sources from which this uncertainty stems: platform uncertainty mediating the impact of early technical feedback and market uncertainty mediating the impact of early market feedback. Our results also indicate that while greater uncertainty is associated with making later changes to a product's design, this practice is not associated with performance. Our findings suggest that managers carefully must evaluate both the levels and sources of uncertainty facing a project before designing the most appropriate process for its execution. In particular, they should explore the use of specific development practices based upon their usefulness in resolving the specific types of uncertainty faced. Importantly, these decisions must be made at the start of a project, with purposeful investments to create a process that best matches the context. Reacting to uncertainty ex‐post, without such investments in place, is unlikely to prove a successful strategy.