Estimating project cost under uncertainty using universal generating function method

One of the important issues in project management is cost contingency estimation. Although this has been investigated in previous studies, the underlying assumptions on presuming specific probability distribution functions (like triangular and normal) for activity costs impose curve fitting preliminaries, which leads to approximation errors along with time inefficiency. This paper is sought to propose a fast and exact method to assess the risk of projects for any arbitrary probability mass function of activity costs obtained directly from historical data. The proposed method facilitates determination of cost contingency indicator curves that can help decision makers, planners, and contractors to select a reasonable contract value, and managing resources in predesign phase or during construction more efficiently. Our methodology consists of four main parts: (1) determining the universal generating function (UGF) of activity costs, (2) modeling the project network as a parallel multistate system, (3) determining the UGF of project cost, and (4) determining the probability mass function of project cost. The effectiveness of the proposed method is examined using sample projects, indicating that it is 5×105 times faster than Monte Carlo simulation in a 27-activity example, while the average costs obtained from the two methods is not different at a significance level of 0.0001. Moreover, the proposed method provides exact project cost probability distribution in discrete cases unlike Monte Carlo simulation, which does not have the ability to estimate extreme states, and, therefore, it can be a suitable substitution for Monte Carlo simulation.