Twenty years ago, the Construction Industry Institute (CII) published a report titled, “A Comparison of U.S. Project Delivery Systems,” which benchmarked the performance of design-bid-build (DBB), construction manager at risk (CMR) and design-build (DB) projects. The report examined data from over 350 projects of varying size, sector, complexity, and location that were completed in the mid-1990s. The analysis revealed that DB projects outperformed both DBB and CMR in terms of unit cost, cost and schedule growth, and all metrics relating to the speed of delivery. These results had a profound impact on how projects were delivered in the construction industry and, in the decades since this seminal report, there has been considerable interest in updating the benchmarks. The purpose of this research effort was to provide new benchmarks for DBB, CMR and DB performance by repeating the same methodology employed by the authors of the 1998 CII report with a data set of contemporary projects.
The Charles Pankow Foundation (Grant #02-18)
This research followed the same data collection and analysis procedures outlined in CII’s 1998 report, which relied solely on quantitative methods. Data from a sample of 212 vertical construction projects completed between 2008 and 2013 were collected via survey questionnaire and verified for accuracy with the project owners. Using this data, a best subset analysis was performed for each measure of project performance, including unit cost, cost growth, schedule growth, construction speed, and delivery speed. This analysis identified sets of predictor variables that explained the greatest amount of variation in each measure. An ordinary least square (OLS) regression was then performed using the best subset of variables to predict the corresponding performance measure and derive coefficients of the regression equations. By varying the project delivery system, these equations were then used to calculate the average expected performance for all the projects in the data set, when all other variables in the model are held constant. In this way, the effect of the project delivery system was isolated, and the performance of each system was compared to one another.
The results showed that, on average, DB projects were delivered faster and with nominally lower cost growth and schedule growth than their CMR and DBB counterparts. The completed unit cost of DB projects was also comparable to DBB and slightly less than CMR projects. However, the modeling does indicate that, except for delivery speed, the gap in performance between DBB, CMR, and DB has narrowed over time.
By examining the regression models themselves, several other variables, in addition to the project delivery system, contributed to these performance differences. Projects with excellent team chemistry and open book contracting terms (e.g., cost plus a fee and GMP) were, on average, more likely to have lower unit costs and cost growth. On average, projects that involved both the designer and builder in early goal-setting had reduced schedule growth. These findings provide further evidence that the project delivery system is not the sole contributor to project performance and these human-related factors are, in some cases, equally as important.