A cost estimation model for improving the budget estimates of industrial plant construction projects

It is well established that construction projects do not meet their cost budgets. Part of the problem is because cost budgets are not based on definitive cost estimates that are founded on detailed design and engineering, but rather they are based on preliminary cost estimates prepared when the level of project definition is less than 40% at its best. While the expected accuracy range of preliminary cost estimates is between -20% to +30%, past research found that on average over 85% of estimates come closer to the higher end of accuracy range at 28%. This praxis introduces a multiple linear regression model that utilizes some of the project cost elements to estimate its construction cost. This new hybrid (stochastic/deterministic) cost estimation method improves the preliminary cost estimates of industrial plant construction projects. The new model can estimate construction costs within 15% accuracy, which is comparable to results from definitive cost estimates, yet at less than 40% project definition. While definitive cost estimates require detailed project definition for a full set of 15 cost elements, the new model requires only 3, site works, electrical works and auxiliary equipment works. It can be applied earlier in the life of the project to compute improved preliminary cost estimates, providing the project stakeholders with more accurate cost budgets. The model was developed using 32 definitive cost estimates, for small industrial plant projects under US$50 million, and their actual construction costs. It was tested using 4 different datasets reserved for model validation. It has an R2(adj) and R2 (pred) of 82.34% and 79.41% respectively, and an overall regression p-value of 0.000.