House of Project Complexity—understanding complexity in large infrastructure projects

Abstract

This paper describes our conceptualization of complexity in large infrastructure projects. Since complexity itself is an emergent concept that is hard to pin down, we focus on the relationship between various project features and properties associated with complexity such as difficulty, outcome variability and non-linearity, and (non) governability. We propose a combined structural and process-based theoretical framework for understanding contributors to complexity—the ‘House of Project Complexity’ (HoPC). The formulation of the HoPC draws from a rich projects literature and is developed iteratively by first applying it to two trial samples and then to the main data set of 20 detailed case studies of infrastructure projects prepared for the IMEC study. A main contribution of this work is the conceptual distinction in the HoPC between ‘inherent project features’, ‘architectural features’, and their relationship with project outcomes and emergent properties—the ‘ilities.’ A second contribution is the separation of ‘inherent features’ into the technical and institutional domains, which are then developed in parallel fashion. A third contribution is to link complexity with the concept of project architecting. The HoPC can be generally extended to projects in the extractive industries, large manufacturing projects or other industrial megaprojects.

Keywords:

• Complexity
• infrastructure
• institutions and projects
• project architecture
• project shaping
• risk

Notes

1. Appendix 1 contains a detailed description of our initial exploratory analysis, including the scaling method for operational indicators for both T and O complexity, descriptive statistics, ANOVA analysis and regression results for the relationship between performance and T and O complexity. The regression results supported our intuition that the interaction of technical complexity and organizational complexity had a more important effect on project project's performance than their independent individual contributions. Specifically, project performance worsened in our sample (p < .01) along with an increase in the dispersion of performance, as the relative overall compound (T*O) complexity increased, with better fit than T or O individually or additive T+O.

2. We considered using qualitative comparative analysis (Rihoux and Ragin, 2009) given the small N and relatively large number of constructs in our sample, but found that the more widely used parametric analysis was adequate to demonstrate the significance of our results. In future studies with more detailed project coding, we most likely will use an alternative method.

3. It is well recognized that initially overambitious schedules are often imposed on major projects (see e.g. Priemus et al., 2008), and that this schedule pressure is itself a source of complexity, introducing the potential for correlated errors between the dependent and independent variable. However, deviations from publically announced schedules are the only information available in the public record.