Empirical comparison of weighted and non-weighted reference class forecasting

This work investigates the potential for improved accuracy of a cost-estimating method called weighted reference class forecasting (WRCF) with respect to the traditional reference class forecasting (RCF) method. WRCF is a modification of the RCF method, which uses historical project cost data to produce project cost estimate uplifts. WRCF uses work package-level cost deviations to overcome a limitation of RCF: the unavailability of reference data. However, WRCF has not been validated, raising concerns about the method’s reliability. This case study uses real project data from a utility company in simulated project portfolios and cross-validation to assess the validity of WRCF. The study contributes to the robustness of RCF in practice by providing empirical evidence on the performance of WRCF. Results indicate that WRCF produces more conservative cost estimates than the traditional RCF. In addition, WRCF cost estimates have a higher level of certainty than expected, potentially leading to inflated cost estimates beyond what is intended. This study emphasises the importance of practitioners selecting appropriate WRCF percentiles to avoid excessive conservatism in cost estimates and overallocation of resources. Further research is needed to assess WRCF’s validity across diverse industries and regions, providing insights for practitioners in estimating project cost uplifts.