A Novel Approach to Capture Similarity in Capital Project Benchmarking: An Application to Healthcare Projects

Abstract

For credible project benchmarking, like-for-like project comparison is a prerequisite to set realistic targets for improvement, especially when a heterogeneous sample of healthcare projects is compared to another. However, the current method for determining the groups of similar projects relies on an ad-hoc technique that can lead to suboptimal target settings for improvement. To address this issue, this research proposes a novel approach to capture similarity for capital project benchmarking by leveraging Classification and Regression Trees (CART). This research focuses on healthcare projects. The data collected from a total of 89 healthcare projects were used to construct the trees by selecting a set of critical and flexible features that are closely associated with two metrics representing cost and schedule performance of the selected projects. The effectiveness of results derived from the proposed method was validated through statistical methods and comparative analysis. The proposed method allows for more targeted performance comparisons by capturing similarity using flexible sets of meaningful features, which reduces the search space of determining a group of similar projects. The new approach is, thus, expected to help organizations gain better insights into their relative performance position when benchmarking their capital projects.