Accuracy in design cost estimating
The level of achieved accuracy in design cost estimating is generally accepted by researchers as being less than desirable. Low accuracy has been attributed to the nature of historical cost data, estimating method and the expertise of the estimator. Previous researchers have suggested that the adoption of resource based estimating by designers could eliminate data and method-related problems. The work in this thesis has shown that this will not solve the problem of inaccuracy in estimating. A major problem in assessing accuracy in design cost estimating has been the absence of a generally agreed definition of the'true cost' of a construction project. Hitherto, studies of accuracy in design cost estimating have relied solely on the assessment of errors using the low bid as a datum. Design cost estimators do not always focus on predicting the low bid. Rather, they may focus on the lowest, second lowest, third lowest or any other bid, mean/median of bids, or sometimes, on just being'within the collection'. This has resulted in designers and researchers having different views on the level of achieved accuracy in estimating. To resolve this problem, an analysis package, ACCEST (ACCuracy in ESTimating), was developed to facilitate 'fair' assessment of accuracy in design cost estimates. Tests - using cost data from 7 offices, the ACCEST package and the OPEN ACCESS II package on an IBM PS/2 - have shown that error in design cost estimating (averaging 3.6% higher than the predicted parameter) is much lower than portrayed in construction literature (averagel3% higher than the low bid). Also, false associations between project environment factors (such as geographical location, market conditions, number of bidders, etc.) and the level of achieved accuracy has been developed by researchers through using the low bid as a datum. Previous researches have also demonstrated that design estimators do not learn sufficiently from experience on past projects. A controlled experiment on design cost estimating information selection was designed to explain this occurrence. Failure to learn, and the persistent use of information on one project for estimating, has been shown to result from the method of information storage in design offices, the illusion of validity of inaccurate rules and over-confidence resulting from inaccurate assessment of individual expertise. A procedure for aiding learning from experience in design cost estimating has been suggested. Finally, the work has shown that by distinguishing between different trades, and selectively applying different estimating strategies, based on the objective evaluation of the uncertainty associated with cost prediction for ear h trade, error in design cost estimating could be further reduced. Two formulae for predicting tender prices using data generated from historical cost estimating experience are represented.