A Bayesian approach to improving estimate to complete

The capability to develop a reliable ‘Estimate at Completion’ from the earliest stage of project execution is essential in order to develop a proactive project management. This paper provides a methodology to support the development of the Estimate at Completion in large engineering projects. In order to accomplish this aim, a model to formulate estimates at completion is presented which integrates through a Bayesian approach three knowledge sources: experts’ opinions, data from past projects and the current performance of the ongoing project. The model has been applied to three Oil and Gas projects in order to forecast their final duration and cost. These projects are characterized by a high level of size, uncertainty and complexity representing a challenging test for the model. The results obtained show a higher forecasting accuracy of the Bayesian model compared to the traditional Earned Value Management (EVM) methodology. Moreover, the estimates at completion calculated using the Bayesian model are not point estimates such as those calculated by EVM. In fact, the Bayesian approach leads to a probability density function for the forecasted final cost and duration. Hence, the project manager obtains an indication of the degree of confidence about the expected value forecasted which results in better quality information available for the decision making process.