Simple generic models for cost-significant estimating of construction project costs

In this thesis, a new breed of hierarchical cost models is developed whose accuracy and complexity can be adjusted to suit the quality of data available at each stage of the design process. In their most detailed form, these models are as accurate as conventional ones but considerably simpler. The principle of cost-significance, which is used to derive the models, is explained. The models are based upon the finding that for defined project categories identical cost-significant work packages (CSWP) can be derived which contribute a constant fraction of the total project value. The techniques used to develop the models are described. A suite of cost models has been produced for three project categories; reinforced concrete bridges, major roads and trunk sewers. Each model comprises a few (no more than 27) CSWPs which need to be priced. The total value of the CSWPs is divided by an associated cost model factor to predict the total project value. The cost model factor represents the proportion of value contributed by the CSWPs. The ability of the models to predict total bill values was tested on virgin bills. The achieved accuracy was within ±7% (for two thirds of cases) at the detailed design stage and within ±8% (for two thirds of cases) at the feasibility design stage. The detailed model for bridges has been incorporated in a commercial estimating software package called BRIDGET. The software package is described. The variability of unit rates applied to cost-significant and non-cost-significant items (CSI) has been studied. Generally, unit rates applied to non-CSIs vary up to three times as much as unit rates applied to CSIs. Moreover, the difference in unit rates between similar CSIs is of no greater order than the typical fluctuations in rates expected from one tender to another in pricing a CSI. Since the CSWPs relate to the way costs are incurred on site the models can be used to compare estimated and actual costs and can provide an effective mechanism for feedback to estimators.