This article presents findings of a post-mortem assessment of the accuracy of the percentage method used in estimating the cost contingency. The study was based on project data obtained from an organisation in Botswana, which for confidentiality reasons, was codenamed HouseBot. The study was motivated by premise that HouseBot, pre-sells houses before construction and in this mode there is a possibility of either overpricing or underpricing the housing units due to the estimated cost contingency’s contribution to the selling price. The study used 32 housing projects completed between 2011-2018, to assess the accuracy of the percentage method based on a contingency performance index (CPI). A CPI of zero percent (0%) indicates that the estimated cost contingency (ECC) exactly predicts the approved variation amount (AVA) found in the final account. An average CPI of 3.2% obtained leading to a conclusion that the percentage method used at HouseBot lacks the required accuracy. Despite the limitation of the study using a small sample of projects, the finding points to the need for the organisation to investigate possible use of alternative estimation methods, for example, probability based methods to increase accuracy.
Ssegawa et al. (2026) studied this question.