Improving organization design by clustering project actors to mitigate the risks induced by complex project structure

Abstract

Project complexity can lead to uncertainty in decision making and ambiguity in the interdependencies and intricacies of project structure. A change or negative impact on one element of the project can propagate throughout the entire network of elements. These phenomena, coupled with concurrent engineering practices, can lead to time delays and cost overruns. The existing hierarchical structure of organizations is simply unequipped to handle this type of complexity. This is especially true for projects that encompass members from different departments of an organization or from different organizations altogether. This research thesis proposes to cluster project actors, or to form complementary teams, based on their interactions and interdependencies with each other. Creating coordinated, effective communication channels between members of a project can alleviate the negative effects induced by complex project structure. The Design Structure Matrix (DSM) is used as a modelling tool to represent the interactions between project actors, which are based on the number of deliverables exchanged between each member. An agglomerative clustering technique is then used to place the project members into groups. The proposed clustering algorithm was developed and implemented in MATLAB. The algorithm permits the project manager or the user of the algorithm to impose certain restrictions, such as the size of the teams, which members must be placed together, and which members must be kept apart. The algorithm also allows critical loops to be inserted into clusters. It produces as output the clustered communities of actors as well as the percentage of the total number of interactions that occur within the clustered groups. The latter is an evaluation of the performance of the algorithm.