Revisiting Project Delivery Performance

Final Report

December 31, 2018

Keith Molenaar, Ph.D. Bryan Franz, Ph.D.

K. Stanton Lewis Professor Assistant Professor
University of Colorado Boulder University of Florida
kieth.molenaar@colorado.edu bfranz@ufl.edu

Summary
This report presents the results of a study to examine and compare the performance of design-bid-build
(DBB), construction manager at risk (CMR) and design-build (DB) project delivery systems in the U.S.
building construction industry. By leveraging data from 212 projects completed over the past ten years,
we use a best subset analysis to explain the variance in five measures of performance: unit cost, cost
growth, schedule growth, construction speed and delivery speed. From these regression models, we
calculate the average expected performance of each project delivery system. The results clearly show that
the DB projects were delivered faster and with lower cost and schedule growth than the CMR and DBB
projects in the data set. In addition, the completed unit cost of DB projects was comparable to DBB and
slightly less than CMR projects. Our results are generally consistent with findings from the Construction
Industry Institute’s seminal report from 1998 that compares performance across delivery systems.
However, our modeling does indicate that, with the exception of delivery speed, the gap in performance
between DBB, CMR and DB has narrowed over time. We then interviewed owners and project
participants from the best and worst performing projects within the data set. Across the best performing
projects, interviewees frequently cited the owner’s emphasis on a relational project culture and having
previously contracted with the same architect or contractor as being vital to project success. Across the
worst performing projects, interviewees identified consistent challenges, including a lack of experience
with the project delivery system, poor communication between the owner and contractor, and
understaffing or turnover within the project team. These interviews validated our regression models, as
well as provided insight into how to improve the likelihood of a successful project, regardless of the
project delivery system.

1. Purpose
Twenty years ago, the Construction Industry Institute (CII) published a report titled, “A Comparison of
U.S. Project Delivery Systems,”1 which benchmarked the performance of design-bid-build (DBB),

1 Sanvido, V. and Konchar, M. (1998). “Project delivery systems: CM at risk, design-build, design-bid-build.” The Construction
Industry Institute, RT-133.
construction manager at risk (CMR) and design-build (DB) projects. The report examined data from 351
projects of varying size, sector, complexity, and location that were completed between 1984 and 1998.
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 public and private owners decided to deliver projects in the
construction industry. The value of non-residential construction projects in the U.S. undertaken using a
DB delivery system increased from an estimated 24% in 19962 to 38% in 20143. By 2021, DB delivery is
expected to encompass 44% of construction spending, with the greatest increases being seen in the
manufacturing, highway and education construction sectors4. This growth has been primarily customer-
driven, in that more owners are becoming aware of and choosing alternative forms of delivery for their
projects because of their reported performance advantage.
Now, in the nearly two decades since this seminal report, there has been considerable interest in updating
these performance benchmarks. The purpose of our 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.

2. Background
There is a wide array of project delivery systems in the U.S. building construction industry. However, the
most common project delivery systems continue to be DBB, CMR, and DB. The distinction between
these systems has blurred somewhat, yet there is general agreement in the characteristics that define them:
(1) the number of contractual relationships with project stakeholders; and (2) the timing of involvement of
those stakeholders. Recent research has also identified patterns in (3) the contract payment terms and (4)
stakeholder selection criteria that are commonly used in each delivery system5. Figure 1 provides a
graphical comparison of the contractual relationships within each delivery system.

Design-Bid-Build Construction Manager at Risk Design-Build

Owner Contract
Owner Owner
Coordination

General Design
Designer Designer CM/GC
Contractor Builder

Design Specialty Design Specialty Design Specialty

Consultants Trades Consultants Trades Consultants Trades

Figure 1. Project delivery systems

2 Tarricone, P. (1996). Design-build it, and they will come. Facility Design and Management, September, 60-63.
3 Duggan, T. and Patel, D. (2014). Design-build project delivery market share and market size report. Design-Build Institute of
America and RS Means Reed Construction Data Market Intelligence, Norwell, MA.
4
FMI. (2018). Design-build utilization: Combined market study. Design-Build Institute of America.
5 Franz, B. and Leicht, R. (2016). An alternative classification of project delivery methods in the United States building

construction industry. Construction Management and Economics, 34(3).

2
Under DBB, an owner first contracts with a designer, and with the design to nearing 100% completion,
contracts with a general contractor (GC) to build the project. The GC is typically selected on the basis of
their competitive bid price, which becomes their lump sum contract amount. The designer represents the
owner’s interests throughout the construction phase, exercising oversight of the GC.
Under CMR, an owner first contracts with a designer and then with a GC or CM once the design is
between 20 to 60% complete. Since the scope is still being defined at that time, the owner selects the GC
or CM based on a combination of their qualifications, plan for the work and fee. The GC or CM’s
contract may begin as a cost reimbursable, but is often converted to a guaranteed maximum price (GMP)
or lump sum later in the project. Because of their early involvement, the GC or CM is expected to
coordinate closely with the designer to provide constructability guidance, estimating and scheduling
services.
Under DB, an owner has a single contract with a design-builder, who may be a firm offering in-house
design, engineering and construction services, or teamed design and contracting firms. The owner
commonly selects the design-builder based solely on their qualifications or through a combination of cost
and technical proposal (i.e., best-value procurement). The design-builder may be contracted when the
design is between 0 to approximately 20% complete. Depending on the level of design completion at
selection, the design-builder’s contract may be either lump sum or cost plus a fee with or without a GMP.
In the 20 years since the seminal 1998 CII study, additional studies have continued to explore the
performance of these three project delivery systems. Most confirm that DB projects outperform CMR
and DBB in the public infrastructure sector6, as well as on sustainable projects7 and healthcare projects8.
However, there are a few studies that report only minor or no difference between delivery systems in
terms of productivity9 and cost growth10. Most recently, the intermediate factors of team integration and
group cohesion were studied as an explanation for why certain delivery systems perform better than
others11. Team integration was defined as the degree to which the owner, designer and builder engage in
collaborative activities, while group cohesion was the degree to which those same stakeholders developed
a sense of shared task commitment and interpersonal alignment with one another. The authors concluded
that both team integration and group cohesion are important to project success, regardless of the project
delivery system. Specifically, the project delivery system provides the framework or “potential” for
success, while the project team is critical determinant in reaching that potential. To complement these
research efforts, new performance benchmarks for DBB, CMR and DB are needed.

6
Molenaar, K., Songer, A. and Barash, M. (1999). Public-sector design/build evolution and performance. Journal of
Construction Engineering and Management, 15(2).
7 Korkmaz, S., Riley, D. and Horman, M. (2010). Piloting evaluation metrics for sustainable high-performance building project

delivery. Journal of Construction Engineering and Management, 136(8)

8 El Asmar, M., Hanna, A. and Loh, W. (2013). Quantifying performance for the integrated project delivery system as compared

to established delivery systems. Journal of Construction Engineering and Management, 139(11).

9 Ibbs, C., Kwak, Y., Ng, T. and Odabasi, A. (2003). Project delivery systems and project change: Quantitative analysis. Journal

of Construction Engineering and Management, 129(4).

10 Thomas, S., Macken, C., Chung, T. and Kim, I. (2002). Measuring the impacts of the delivery system on project performance:

Design-build and design-bid-build, NIST, Austin, TX.

3
3. Methods
This research leverages an existing data set from a previous study11, wherein a large sampling of projects
(204 projects) were collected via survey questionnaire and extensively verified via conversations with the
owners of each project. The projects within this data set were distributed across the U.S. (see Figure 2)
and all completed after 2008. The distribution of project delivery systems used on these projects was 42
for DBB, 78 for CMR, 81 for DB and 3 classified as integrated project delivery (IPD). The 3 IPD cases
were removed from the data set prior to our analysis. To increase the sample of DBB projects to be more
consistent with the number of CMR and DB projects, we supplemented the existing data set with 11
additional DBB projects that met the same inclusion criteria as the previous study.
The final data set for our analysis contained 212 projects, 53 of which were DBB, 78 were CMR and 81
were DB. The projects ranged in size from about 10,000-square feet at the smallest to over 550,000-
square feet at the largest, with 57% falling below 150,000-square feet. Of these projects, 62% were
publically-funded, either through federal, state or local municipalities.
To remain consistent with the 1998 CII study methodology, we classified each facility by their use into
families that closely align with project complexity. In order of increasing complexity, 9% of projects in
the data set were considered light industrial, 13% as multi-story dwelling, 45% as simple office, 12% as
complex office, 5% as heavy industrial and 16% as high-technology. A complete descriptive summary of
the data set by project delivery system is provided in Appendix A.

Number of Projects
1 33

Figure 2. Geographic distribution of projects in data set

To compare the project delivery system performance, we repeated the same methods used in the 1998 CII
study. From the 212 projects, a best subset regression analysis was performed for each measure of
performance: unit cost, cost growth, schedule growth, construction speed and delivery speed. This
analysis identified sets of variables that explained the greatest amount of variation in each measure.
Table 1 summarizes the variables included in the analysis. Indicator coding was used for categorical
variables and one level was removed to avoid multicollinearity in the regression models. Replicating the
1998 CII study procedure, several interaction terms were also created to account for moderating effects of
facility type and contract terms. When specifying the best subset procedure, the project delivery systems
11Franz, B., Leicht, R., Molenaar, K., and Messner, J. (2017). Impact of team integration and group cohesion on project delivery
performance. Journal of Construction Engineering and Management, 143(1).

4
(DBB, CMR and DB) and facility types (e.g., light industrial) were included in every model. The set of
variables resulting in the highest adjusted R-squared and lowest Mallow’s Cp was selected for each
performance measure. An ordinary least square (OLS) regression was performed using each set of
variables to predict the corresponding performance measure and derive coefficients of the regression
equation. Projects with a high Cook’s distance (D > 1), indicating a disproportionate effect on model fit,
were removed and the OLS regression was re-run if necessary. These equations were then used to
calculate the average performance for projects delivered under each project delivery system, when all
other variables are held constant.

Table 1. Summary of predictor variables used in the best subsets regression procedure

Predictor Variable Coding

Project Delivery System
CM at Risk 1=Yes, 0=No
Design-Build 1=Yes, 0=No
Timing of Builder Involvement Percent of design completed when the builder was hired, 0 to 100
Negotiated Builder Selection 1=Yes, 0=No

Facility Type
Light Industrial 1=Yes, 0=No
Multi-Story Dwelling 1=Yes, 0=No
Simple Office 1=Yes, 0=No
Complex Office 1=Yes, 0=No
High Tech 1=Yes, 0=No

Project Characteristics
1 / Area (Log10) Inverse Log10 transform of the gross building area
Initial Unit Cost (Log10) Log10 transform of the project’s contracted unit cost
Deep Foundation System 1=Yes, 0=No
LEED Certification Goal 1=Yes, 0=No
High Complexity 1=Yes, 0=No
Average Complexity 1=Yes, 0=No
Percentage New Construction Percent of project that was new construction, 0 to 100

Team Characteristics
Public Owner 1=Yes, 0=No
High Administrative Burden 1=Yes, 0=No
Participation in Goal Setting Proportion of project team that participated in goal setting, 0 to 1
Participation in BIM Planning Proportion of project team that participated in BIM planning, 0 to 1
Excellent Team Chemistry 1=Yes, 0=No
Excellent Communication 1=Yes, 0=No

Interactions
CMR*Light Industrial Project was both CMR and Light Industrial: 1=Yes, 0=No
CMR*Multi-Story Dwelling Project was both CMR and Multi-Story Dwelling: 1=Yes, 0=No
CMR*Simple Office Project was both CMR and Simple Office: 1=Yes, 0=No
CMR*Complex Office Project was both CMR and Complex Office: 1=Yes, 0=No
CMR*High Tech Project was both CMR and High Tech: 1=Yes, 0=No
DB*Cost Plus Project was both DB and Cost Plus: 1=Yes, 0=No
DB*GMP Project was both DB and GMP: 1=Yes, 0=No

5
To provide validation for the regression analysis and offer additional insights into performance, we also
conducted interviews with a sample of respondents from the best and worst performing projects. These
projects were identified by considering the upper and lower quartiles of the performance measures in the
data set (see Figure 3). The best performing projects were those found in at least three of the following
ranges: the lower quartile for cost growth, the lower for schedule growth, the upper quartile for
construction speed and the upper quartile for delivery speed. The worst performing projects were those
appearing in at least three ranges at the opposite end of the distributions: the upper quartile for cost
growth, the upper for schedule growth, the lower quartile for construction speed and the lower quartile for
delivery speed. Unit price was not considered in this classification because of its strong dependence on
facility type.
Using these criteria, we classified 24 projects as “best” performing and 16 projects as “worst” performing.
Each of these projects was contacted and, of the best performers, 9 respondents (38%) agreed to a follow-
on interview. Seven of the respondents (44%) on the worst performers agreed to participate. We
performed semi-structured interviews over phone or video conference with each of the respondents, using
a guiding list of questions (see Appendix C). Notes taken during the interview were entered into a
spreadsheet. Afterwards, the projects were cross-compared with one another to identify commonalities
both within and between the best and worst performing project groups.

Figure 3. Best performing (green shaded areas) and worst performing (red) quartiles by performance
measure

6
4. Results
Descriptive statistics. Prior to the regression analysis, a simple comparison of the median and mean
project performance measures was prepared. Table 2 summarizes this comparison, both overall and by
delivery system. The standard deviation (Std. Dev.) is a measure of the variation or dispersion of the
measure, where a larger standard deviation suggests that the data is spread over a wider range of values.
Because of the relatively large standard deviations of these measures, we caution against drawing
conclusions based on this summary alone. The best subsets regression results offer a more accurate
comparison because they consider and control for differences in all project characteristics, while holding
the impact of the project delivery system constant.

Table 2. Median and mean performance by project delivery system

Std.
Performance Measure n Median Mean
Dev.
Unit Cost ($/ft2) 204 387 422 239
DBB 52 431 448 245
CMR 73 427 442 227
DB 79 327 388 243
Cost Growth (%) 203 1.15 3.34 7.9
DBB 53 1.90 3.23 8.6
CMR 72 0.91 3.99 8.5
DB 78 0.85 2.81 7.2
Schedule Growth (%) 212 0.00 4.38 15.2
DBB 53 2.49 6.29 13.2
CMR 78 0.22 5.41 16.8
DB 81 0.00 2.16 14.7
Construction Speed (ft2/mo.) 211 5,844 8,845 8,443
DBB 52 3,893 5,562 4,698
CMR 78 6,825 10,626 9,614
DB 81 6,292 9,237 8,592
Delivery Speed (ft2/mo.) 211 3,634 5,777 6,175
DBB 52 1,721 3,214 3,419
CMR 78 3,655 6,270 6,377
DB 81 4,704 6,915 6,175

Best subset regression analysis. The best subsets procedure allowed us to identify the combination of
variables that best explained the variance in each performance measure. Regressing those variables on
each performance measure resulted in equations (see Appendix B) that may be used to predict and
compare project success.
Cost results. Table 3 represents a comparison, in the form of a percent difference, of the average cost
performance for projects delivered under each project delivery system, as predicted by their regression
models. With respect to unit cost and cost growth, DB has the best performance (see Table 3) on
average. When all other variables were held constant, projects using DB are expected to cost 1.9% less
per square-foot when compared to CMR, and 0.3% less when compared to DBB. Similarly, projects

7
using DB are expected to average 2.4% less cost growth than a similarly scoped project using CMR and
3.8% less cost growth than a project using DBB. When compared to DBB, CMR is expected to cost 1.6%
more per square-foot and have 1.4% less cost growth on average. The coefficient of determination (R2)
represents the amount of variation in each performance measure being explained by its regression
equation. In the case of unit cost, we are able to explain 99% of the variation in unit costs across projects
in the data set. This indicates a high level of certainty in the model and we can be more confident in the
results shown. The lower R2 value for cost growth (22%) indicates that we were not able to fully explain
a large portion of the variation in that measure with the variables available in our data set.

Table 3: Predicted cost performance comparison

DB vs. CMR vs. DB vs.

Performance Measure CMR (%) DBB (%) DBB (%) R2
Unit Cost 1.9 less 1.6 more 0.3 less 99
Cost Growth 2.4 less 1.4 less 3.8 less 22

The standardized regression coefficients (β) can be used to compare the relative influence of predictors in
the model on the performance measure. For unit cost, the most influential predictors of performance were
the initial contracted unit cost (β = 0.267), team chemistry (β = -0.006), contract payment terms (β = -
0.006) and facility type (β = 0.006). Given the direction of the relationships and variable coding, this
means that a lower cost per square foot was found on projects with:

 Lower initial contracted unit cost at the start of the project

 Excellent team chemistry among the owner, designer and builder (GC, CM or design-builder)
 Open book contracting terms, such as a cost plus a fee or GMP
 Lower project complexity
For cost growth, the most influential predictors were the project delivery system (β = -1.85), facility type
(β = -1.72), contract payment terms (β = -1.67), timing of builder involvement (β = -1.48) and team
chemistry (β = -1.31). In other words, reduced cost growth was more likely to be found on projects with:

 A DB project delivery system

 Lower project complexity
 Open book contacting terms, such as a cost plus a fee or GMP
 Earlier timing of builder involvement
 Excellent team chemistry among the owner, designer and builder (GC, CM or design-builder)
Schedule results. Design-Build was the best performing project delivery system in terms of schedule
growth, delivery speed and construction speed (see Table 4) on average. This table summarizes the
percent difference in average schedule performance for projects delivered under each project delivery
system, as predicted by their regression models. When all other variables were held constant, projects
using DB are expected to have 3.9% less schedule growth than a comparable project using CMR and
1.7% less schedule growth than a project using DBB. On average, DB projects are delivered 13% faster
during construction and 61% faster from design through final completion when compared to CMR
projects. DB projects are also delivered 36% faster during construction than DBB and 102% faster over
the entire project duration. Similar to the cost results, the R2 values suggest greater certainty in the

8
regression models that produced the construction speed (88%) and delivery speed (89%) results. There is
still a large proportion of the variation in schedule growth, approximately 79%, that could not be
explained by the variables in our analysis.

Table 4: Predicted schedule performance comparison

DB vs. CMR vs. DB vs.

Performance Measure CMR (%) DBB (%) DBB (%) R2
Schedule Growth 3.9 less 2.2 more 1.7 less 21
Construction Speed 13 faster 20 faster 36 faster 88
Delivery Speed 61 faster 25 faster 102 faster 89

In order of decreasing influence, schedule growth was predicted by the participation in goal setting (β = -
4.31), timing of builder involvement (β = -2.74), facility type (β = -2.30), owner type (β = -2.16) and
foundation system (β = 1.86). Allowing for the direction of the relationships and variable coding, this
means that reduced schedule growth was found on projects with:

• Participation of the designer and builder (GC, CM or design-builder) in project goal-setting

• Earlier timing of builder involvement
• Lower project complexity
• Public funding source
• Simpler foundation systems, such as slab-on-grade
For construction speed, facility size was most influential (β = -0.383), followed by the project delivery
system (β = 0.149) and initial contracted unit cost (β = -0.092). In other words, the rate of completion of
the construction scope, from notice to proceed to final completion was faster on projects with:

 Larger gross square footage

 Higher initial contracted unit cost at the start of the project
 A design-build or CM at risk project delivery system
For delivery speed, the same predictors were also influential, but in a different order. Facility size (β = -
0.355), initial contracted unit cost (β = -0.081) and project delivery system (β = 0.064) had the most
impact on delivery speed. Therefore, the rate of completion of the entire project, from design initiation to
final completion was faster on projects with:

 Larger gross square footage

 A design-build or CM at risk project delivery system
 Higher initial contracted unit cost at the start of the project
Case study comparisons. A total of 16 semi-structured interviews were performed with respondents, 9
of which were owners or builders on a project that we classified as a best performer within the data set
and 7 that we classified as a worst performer. A brief summary of each interview is provided in
Appendix D. Beyond simply confirming the influential variables from the regression analysis, the
interviews also explored additional factors leading to the success or failure of the project.

9
 Best performers. Across the most successful projects, there were four factors that appeared with
regularity (see Figure 4): early team formation, relational culture, clear expectations and repeat clients. A
description of these factors and several examples from projects are summarized below.
Early team formation meant that the designer and builder (GC, CM or design-builder) were both procured
and collaborating by the design development phase of the project. Of the best performing projects, 78%
described this factor, compared to only 29% of the worst performing projects. Most of the respondents
we interviewed described the primary benefit of early team formation as the ability to engage in rigorous
pre-planning for the construction phase.
A relational culture was present when the attitudes and values of project participants aligned with
concepts of cooperation and collaboration. The owner placing a strong emphasis on creating a relational
culture was referenced in 78% of the best performing projects, compared to only 14% on the worst
performing projects. Each owner had a different approach to shaping their project’s culture. One owner,
in particular, made their expectations clear during the first meeting. They challenged each member of the
design and contractor’s teams to not tolerate arguments or unprofessionalism, be willing to express their
opinion and to treat each other fairly. Another owner described a constant drive for greater accountability
and one project successfully implemented periods of co-location that began during the schematic design
phase. Regardless of how it was achieved, the culture on these projects emphasized the project’s success
over the success of any single firm or organization.

Factors characterizing 7 interviews from the Factors characterizing 9 interviews from the
worst performing projects Chart Title best performing projects

Frequency -100%
100% -80%
80 -60%
60 -40%
40 -20%
20 0%
0 20%
20 40%
40 60%
60 80%
80 100%
100%

Bring the team together early Early team formation

Develop a relational project culture Relational culture

Communicate expectations Clear expectations

Repeat use of designer/builder Repeat clients

Lack of experience Lack of experience

Poor communication Poor communication

Understaffing or turnover Turnover

Figure 4. Comparison of factors characterizing the best and worst performing projects

Clear expectations meant that the owner, designer and builder had an unambiguous understanding of
what will or should be done to have a successful project. This factor appeared only in the best performing
projects and was discussed in over half (56%) of our interviews. On one project, having clear expectation
was cited as a reason that the team was able to recover quickly from a delay due to poor soil conditions.
The team was aware of what the owner expected of the project and were able to respond more quickly.
This factor was closely linked having the owner as a repeat client, since an understanding of owner’s

10
expectations could be obtained on a previous project and carried forward to the current project to shortcut
a learning period within the team.
Repeat clients meant that the owner had worked alongside either the designer or the builder, or both, on
prior projects. The majority (56%) of the best performing projects and a few (29%) of the worst
performing described having this factor. Because of these prior working relationships, the owner and
builders that we interviewed were more comfortable communicating and, specifically, more willing to
share challenges or problems encountered on the jobsite with other team members.
Worst performers. There was less agreement across the least successful projects; however, there were
three factors nearly unique to this set of projects (see Figure 2): lack of experience, poor communication
and turnover within the team. A description of these factors and several examples from projects are
summarized below.
A lack of experience with either the project delivery system or project management in general was found
to be factor common to most of the worst performing projects. Inexperience was cited on 57% of the
worst performing projects as the underlying cause of poor planning and inconsistent quality of installed
work. In one case, the inexperience originated with owner’s project manager who lacked experience with
the DB project delivery system and was unsure of their role in the process, which led to delayed decisions
and poorly conveyed expectations. In another, the project manager for the contractor was new to the
company and assigned to a project with tight deadlines. He had limited experience and, as a result,
managed reactively to current problems, losing sight of the overall schedule.
Poor communication meant that communication among the owner, design and builder was lacking in
either frequency or quality, or both. This deficiency was cited on 29% of the worst performing projects
and only 11% of the best performing projects. The causes of poor communication cited in our interviews
varied, although there was often a “trigger” event, such as an unexpected condition or conflict that led one
team member to retreat from the team. In one case, poor communication was more simply attributed to
having an international owner. This case presented language, cultural and time zone barriers that the team
was not entirely able to overcome.
Turnover meant that key project team members were transitioned away from the project, either because of
a career move or staffing change within the owner, designer or builder’s firm. Turnover and the resulting
periods of understaffing was a challenge on 14% of the worst performing projects. On one case, frequent
turnover meant that few team members retained a complete understanding of the project. This placed a
heavy workload on members of the project team and forced long hours. The result was increased stress
and animosity among the owner, designer and builder.

5. Discussion
This study compared the performance of DB, CMR and DBB delivery systems for a data set of
contemporary projects, using the same methodology as the seminal 1998 CII study. The best subset
procedure allowed us to identify the sets of variables that explained the most variation in five measures of
project performance: unit cost, cost growth, schedule growth, construction speed and delivery speed. Five
separate regression models were built from these subsets of variables. We use these models to calculate
the average expected performance for each project delivery system in Table 5. A direct comparison to the
1998 CII study results is also shown in Table 5. We were able to obtain similar levels of certainty (R2) in
our regression models as the 1998 CII study. While DB is the best performing delivery system in both

11
studies, we found that the performance gap in unit cost, cost growth and schedule growth has narrowed
somewhat over the past 20 years. On the other hand, the gap in construction speed and delivery speed
between DB and CMR and DBB has widened significantly.

Table 5: Comparison of results between 1998 CII study and this study

1998 CII Study This Study

DB vs. CMR vs. DB vs. DB vs. CMR vs. DB vs.
2 2
Performance Measure CMR (%) DBB (%) DBB (%) R CMR (%) DBB (%) DBB (%) R
Unit Cost 4.5 less 1.5 less 6 less 99 1.9 less 1.6 more 0.3 less 99
Cost Growth 12.6 less 7.8 more 5.2 less 24 2.4 less 1.4 less 3.8 less 22
Schedule Growth 2.2 less 9.2 less 11.4 less 24 3.9 less 2.2 more 1.7 less 21
Construction Speed 7 faster 6 faster 12 faster 89 13 faster 20 faster 36 faster 88
Delivery Speed 23 faster 13 faster 33 faster 87 61 faster 25 faster 102 faster 89

Through a more detailed examination of the best and worst performing projects, we also identified several
common factors. The following are recommendations derived from those factors that can be applied,
regardless of the project delivery system, to improve the likelihood of a successful project.
 Bring the team together early: Owners who seek early involvement, not only of the primary
builder, but also of key DB or design-assist specialty contractors, realize more successful
projects. Engaging the core project team members in the design process, before advancing
beyond the schematic design phase, is critical to garner the full value of construction input and to
begin building a cohesive project team.

 Develop a relational project culture: Owners who create a culture of trust within a project team
have a higher probability of success. They can begin building relationships of trust through
qualifications-based procurement and open book contracting. The use of the same designer and
builder on multiple projects, as opposed to low-bid selection on a project-by-project basis, can
jump start a project culture by carrying forward existing relationships.

 Communicate expectations: Early team involvement and a relational project culture provide an
opportunity for exceptional communication. The most successful projects use this opportunity to
set clear expectations at the onset. Treating project goal-setting as a team activity with the owner,
designer, builder and key specialty trades ensures team alignment. Similarly, co-location is an
essential tool on complex projects to facilitate communication and manage expectations
throughout the process.

 Engage in succession planning: The least successful projects in this study experienced
disruptive turnover in key team members from the owner, designer and/or builder’s organizations.
While some level of turnover is unavoidable, the teams that employ qualified project managers
and actively plan for departures have a higher likelihood of success. This means developing a
deep understanding of the roles and responsibilities of each other team member.

12
6. Conclusions
This report provides an empirical investigation into the performance of the three most common project
delivery systems. The results of best subsets regression analysis clearly showed that, on average, DB
projects were delivered faster and with lower cost 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. Our regression models ranged in their ability to explain the variance of these performance
measures and we have greater certainty in the unit cost, construction speed and delivery speed
comparisons. We then identified the best and worst performing projects in the data set and performed
interviews with their project participants to identify common factors. Across the best performing
projects, interviewees frequently cited the early team formation, the owner’s emphasis on a relational
project culture, setting clear expectations and having contracted with the same architect or contractor on a
previous project as being vital to their success. Across the worst performing projects, interviewees
mentioned factors that hindered performance, including a lack of experience with the project delivery
system, poor communication between the owner and contractor, and understaffing or turnover within the
project team. These interviews validated our regression models, as well as provided insight into how to
improve the likelihood of a successful project, regardless of the project delivery system.

Acknowledgements
We would like to thank the Charles Pankow Foundation and CII for their support of this research. We
also thank our collaborators Dr. John Messner and Dr. Robert Leicht at Penn State, Dr. Behzad Esmaeili
at George Mason University and Bradley Roberts at the University of Florida. Finally, we wish to thank
the members of our Advisory Panel: Bill Quatman (Chair), Greg Gidez, Stephen Mulva and Michael
Pappas, who provided valuable advice during the research. Any opinions, findings and conclusions or
recommendations expressed in this paper are those of the authors and do not necessarily reflect the view
of the sponsors, collaborators, or Advisory Panel members.

Definitions
Construction Speed: The square-feet per month rate of completion for construction activities. It is
calculated by dividing the gross square footage of the building by the actual construction duration in
months.
Cost Growth: The percent change in project cost from design initiation to final completion. It is
calculated by taking the final project cost, including both design and construction contract, less the initial
contracted project cost and then dividing by the initial contracted project cost and multiplying by 100.
Delivery Speed: The square-feet per month rate of completion for both design and construction activities.
It is calculated by dividing the gross square footage of the building by the actual project duration in
months.
Schedule Growth: The percent change in project duration from design initiation to final completion. It is
calculated by taking the final project duration, including both design and construction activities, less the
initial contracted project duration and then dividing by the initial contracted project duration and
multiplying by 100.

13
Unit Cost: The cost per square foot of the completed project. It is calculated by taking the final project
cost, including both design and construction contracts, divided by the gross area of the building. All cost
data was indexed by project location and to June 2014 prices prior to analysis to control for regional
differences in prices.

14
Appendix A: Existing Data Set Variable Summary

n Median Mean Std. Dev.

Gross Area 211 116,400 188,710 230,710
DBB 52 68,488 129,813 168,265
CMR 78 155,000 254,880 278,654
DB 81 120,000 162,802 199,329

Gross Area
Light Industrial 19 165,000 232,095 325,896
Multi-Story Dwelling 28 125,500 191,757 197,584
Simple Office 100 90,572 147,722 180,446
Complex Office 24 82,500 175,439 187,036
Heavy Industrial 11 195,000 145,240 126,684
High-Tech 29 116,888 326,157 340,504

Unit Cost
Light Industrial 18 433 417 195
Multi-Story Dwelling 28 286 293 141
Simple Office 97 380 430 238
Complex Office 23 321 339 182
Heavy Industrial 11 355 495 419
High-Tech 27 542 573 209

Cost Growth
Light Industrial 18 2.96 6.07 8.39
Multi-Story Dwelling 28 1.38 3.73 7.68
Simple Office 96 1.90 3.81 7.88
Complex Office 22 0.00 0.17 4.32
Heavy Industrial 11 2.86 3.49 5.37
High-Tech 28 0.00 1.97 10.58

Schedule Growth
Light Industrial 19 0.00 -0.26 6.77
Multi-Story Dwelling 28 1.09 3.77 17.75
Simple Office 100 1.60 6.33 16.65
Complex Office 24 -0.07 0.60 10.24
Heavy Industrial 11 0.00 1.19 3.99
High-Tech 30 0.00 5.54 16.48

15
 n Median Mean Std. Dev.
Construction Speed
Light Industrial 19 9698 12507 12413
Multi-Story Dwelling 28 6707 10393 9198
Simple Office 100 4610 6562 6322
Complex Office 24 8094 9847 6979
Heavy Industrial 11 9534 9953 7536
High-Tech 29 8610 11571 10638

Delivery Speed
Light Industrial 19 6676 8573 9253
Multi-Story Dwelling 28 5186 6184 4794
Simple Office 100 2447 4347 5205
Complex Office 24 4348 6031 4594
Heavy Industrial 11 6391 7029 5953
High-Tech 29 4880 7617 8104

16
Appendix B: Regression Model Results

Performance Measure
Unit Cost Schedule Construction Delivery
Cost Growth
Predictor (Log10) Growth Speed (Log10) Speed (Log10)
Constant **0.076 **32.50 ***47.40 ***7.762 ***7.962
CMR 0.008 -1.32 2.85 0.055 -0.151
DB -0.001 -3.77 -1.62 ***0.132 ***0.305
Light Industrial 0.004 0.73 -9.23 -0.033 -0.088
Multi-Story Dwelling -0.014 -3.67 0.51 -0.030 -0.094
Simple Office -0.007 -2.45 -0.93 -0.047 -0.095
Complex Office *-0.019 *-5.74 -7.96 -0.045 -0.059
High Tech 0.007 1.38 -4.13 0.029` 0.067
Timing of Builder Involvement -0.038 -0.077
Public Owner *-5.47 0.039 0.038
Percentage New Construction -11.61 *-20.37 0.094
Deep Foundation System *4.87 0.042
CMR*Light Industrial 0.131 ***0.428
CMR*Multi-Story Dwelling 0.024 7.13 *0.279
CMR*Simple Office **0.257
CMR*Complex Office **0.190 **0.332
CMR*High Tech *-0.029 -6.57 -0.066 0.164
Negotiated Builder Selection -0.027
DB*Cost Plus *-0.021 -3.49
DB*GMP *-0.013 *-3.34 1.86
Initial Unit Cost (Log10) ***0.981 *-4.81 ***-0.307 ***-0.342
1 / Area (Log10) ***-16.87 ***-18.29
Administrative Burden 1.38 *-0.055
Participation in Goal Setting 0.019 4.43 ***-23.10
Participation in BIM Planning -4.79
Excellent Team Chemistry **-0.015 *-3.04 -3.94
Excellent Communication 3.00
R2 98.9 21.3 21.5 88.2 88.9
R2-adjusted 98.9 12.5 14.2 87.2 87.8
*p < 0.05, **p < 0.01, ***p < 0.001

17
Appendix C: Sample Interview Questions

PROJECT PERFORMANCE INTERVIEW

Purpose of the research: To understand the participants, your experiences will improve the
performance differences of various project delivery confidence in statistical models that we develop to
methods and to identify the specific factors that owners compare project delivery performance.
see as contributing to the success of their projects.
Confidentiality: Your responses to interview questions
What you will do in this research: If you decide to will be kept confidential. Your name and company
volunteer, you will be asked to participate in one affiliation will not be known outside of the research
interview. You will be asked several questions. Some team and will not appear in any publications or
of them will be about project performance. Others will presentations related to the findings of the research.
be about the organizational factors and use of Notes from the interview, without your name, will be
technology that contribute to performance. The kept until the research is complete.
research team will be taking notes during these
interview, but you will not be recorded. Participation and withdrawal: Your participation in
this study is completely voluntary, and you may refuse
Time required: The interview will take approximately to participate or withdraw from the study at any time.
30-60 minutes. You may withdraw by informing the research team that
you no longer wish to participate. You may skip any
Benefits: This is a chance for you to tell your story question during the interview, but continue to
about your experiences concerning a project that you participate in the rest of the study.
recently completed. Taken together with other

To Contact the Researcher: If you have questions or concerns about this research, please contact:
Dr. Bryan Franz, School of Construction Management, University of Florida, 573 Newell Drive, Gainesville, FL
32611; Email: bfranz@ufl.edu; Phone: (352) 273-1161.

Agreement:
The nature and purpose of this research have been sufficiently explained and I agree to participate in this study. I
understand that I am free to withdraw at any time.

Signature: _____________________________________ Date: __________________

Name (print): ________________________________________________

Q1. Why did you choose the project delivery system Q5. How and when were project goals established? Do
used on this project? you feel those goals were shared by the designer
and builder?
Q2. How did you decide when to hire the designer and
builder? Q6. How was the chemistry between yourself, the
designer and the builder? Did you do anything
Q3. How overlapped were design and construction
special create or foster strong relationships within
activities?
the team? When, if ever, do you feel the team
Q4. Did you run into any unexpected challenges on the “clicked?”
project?
Q7. What do you believe most strongly influenced the
performance of this project?

18
Appendix D: Interview Summaries
Best Performing Projects
[1] This project was a 7-story apartment building of approximately 500,000 square feet. The
project was built using a DBB project delivery method and the interview was conducted with
the owner’s representative. This delivery method was chosen because the owner was
comfortable with its execution and gained familiarity by delivering multiple prior projects in
the same way. The owner was very involved in their role and maintained control of the project,
but was very open to the team they worked with and open to creative design solutions. This
project brought project team members together very early, during pre-construction and felt that
early pre-construction services were critical to project success. The project did have a major
problem with unexpected site conditions that had the opportunity to increase costs well outside
of the budget, but with cooperation and teamwork the project was able to recover from these
unexpected costs and was completed under budget. Having the designer and builder together
early and knowing their expectations was a significant factor in becoming a team, upholding
their commitments, and the ultimate success of the project.

[2] This project was a 23-story, 1,250,000 square foot, medical facility built to meet LEED
standards. The project was completed using a CMR delivery method and the interview was
conducted with a member of the construction management team. This method of delivery was
chosen because the project team members felt that it best allowed for pre-planning. The
designer and CM were brought in early and co-located, as well as empowered to make critical
decisions. This project had an aggressive overlap in design, planning, and construction, but
with a large management team located at the site, the project moved at a rapid pace. The
project ran into unexpected site conditions due to inadequate fill, but the overwhelming amount
of pre-planning performed by the project team minimized the disruption. The project used a
real-time punch list method that assisted in keeping the project on schedule. Because of the
size of the project, the amount of planning, and the co-location of the project team, there was an
early development of high team chemistry that made reaching the project goals seem “like a
forgone conclusion.”

[3] This project was a 12-story office building of roughly 310,000 square foot that was built to
LEED standards. The project was completed using the DB delivery method and the interview
was conducted with the owner. They chose this delivery method because of familiarity. The
designer, builder and select specialty trades were all hired early in the delivery process. This
project was designed and then place on-hold before construction due to the recession. This
project had a very strong relational culture and the expectation of the owner was a high level of
success. The interviewee described making a “charge” to the project team during their first
meeting, where they challenged each member of the design-builder’s team to not tolerate
arguments or unprofessionalism, be willing to express their opinion and to treat each other
fairly. The interviewee stated that the penalty for non-compliance was being black listed from
future projects.

[4] This project was a 275,000 square foot, 2 story athletic building. The project was constructed
using the DBB delivery method and the interview was conducted with the owner. This method
was chosen due to funding issues, which required the design to be completed and then more
funds to be raised prior to construction. Project team members worked very well together and
 the general contractor was motived by the hope of future work with the owners, which led to
higher performance at the most advantageous price to the owners. The project did have issues
with needing to import structural fill and dealing with some groundwater infiltration, as well as
having a subcontractor go bankrupt, but with strong leadership and sense of commitment the
project overcame these obstacles. The interviewee stated the there was a previous relationship
with the designer and that, with the motivation of future business for the contractor, the team
came together and worked well toward common goals. The interviewee felt that the high
visibility of the project in the community helped to create a positive “can do” attitude that
assisted in making this project successful.

[5] This project was a 6 story, 335,000 square foot, laboratory-research building that was built to
LEED standards. The project was completed using a CMR delivery method and the interview
was conducted with the owner’s representative. This delivery method was the owner’s
preferred method for larger scale, higher cost projects. The project had little to no overlap of
design and construction activities. The project ran into issues regarding under funding, which
was overcome by creative planning, some team members agreeing to lower their fee, and cost
savings using lean construction methods. The project team worked very well with each other
and had strong relationships even outside of the project. The interviewee felt that this
contributed to the flexibility of the team and assisted them in overcoming the projects funding
challenges.

[6] This project was a 165,000 square foot, 2-story office building. The project was completed
using a CMR delivery method and the interview was conducted with a member of the
construction management firm. This delivery method was chosen because the owner had a
prior relationship with the CM. The interviewee felt that the project was better described as a
hybrid of CM and DB. Critical path trades were brought on early in design-assist roles to help
plan the project. However, the project was delayed on construction activities until a completed
design was submitted due to permitting. The interviewee felt that the largest issue this project
faced was delays and difficulties with the glass and glazing work. The project also suffered
from weather delays, but these were overcome by modifications to the schedule, which allowed
the project to come in with a faster than estimated construction time. The interviewee felt that
the project greatly benefited from the familiarity of the owner, designer and CM team, as this
was the 15th project they had completed together.

[7] This project was a 1-story manufacturing facility of approximately 126,000 square feet. The
project was completed using a DB delivery method and the interview was conducted with a
member of the design-build team. This delivery method was chosen based on the owner’s
preference for this delivery method and a previous relationship with the design-build firm. This
project brought the whole construction team together prior to the submission of the proposal.
The design and construction process was aggressively overlapped and described by the
interviewee as a “rolling” process. The project suffered only minor unexpected drainage issues,
but otherwise was described as a good project. The project goals were set early and the team
understood the importance of not just meeting these goals, but of exceeding them. The
interviewee mentioned that the project team worked extremely well because they were an
existing team that the design-builder had used on many projects before.

20
 [8] This project was a 1- story, 343,000 square foot, manufacturing facility that was built to LEED
standards. The project was completed using a DB delivery method and the interview was
conducted with a member of the design-build team. This method was chosen due to a long
standing prior relationship between the owner and design-build firm. Much of the construction
participants were “valued partners” of the design-build firm and thus had an established
relationship. These participants were involved early in the process. The design for this project
was completed during months of construction stoppage due to weather conditions. The biggest
challenge for this project was meeting the owners design requirements for portions of the
building to that meet “food grade” standards. The design-build firm benefits from being able to
partner project managers and architects from project to project, over many years, this allows for
strong teamwork and efficiency. The project team was large and thus allowed for many minds
and hands to address problems in a quick and timely manner, never really allowing the project
to fall behind.

[9] This project was a 195,000 square foot, 5- story medical building. The project was completed
using a DB delivery method and the interview was conducted with a member of the design-
build firm. This method was chosen due to the owner’s need to get the building to market fast.
The interviewee felt that DB allowed the building process to move quickly and not become
“bogged down.” The project moved at a record pace in the interviewee’s experiences, which
was assisted by having participants brought on to the project without the need for bidding. The
project’s only issue involved the owners wanting to install medical equipment prior to full
completion of the building and thus causing “traffic jams.” The interviewee described using a
white board schedule method on site, with each trade contractor having their own board, to
keep them on task and on time. This goal-setting method, combined with $500 fines for
missing weekly progress meetings, allowed for the project to come in 3/5th quicker than similar
projects. The project benefited from the owner, design-builder and trades all having prior
experiences working together and the owner empowering the design-build team to pick the
best, not the cheapest, trades.
Worst Performing Projects
[10] This project was a 35,000 square foot, 1-story educational building built to LEED standards.
The project was completed using the DBB delivery method and the interview was conducted
with the owner. This delivery method was chosen because of the downward market, which the
owner felt would provide greater competition from contractors and eventual cost savings. The
builder was hired at 75% CD to assist with value engineering and guide the project’s scope
based on costs. The project had issues with site conditions, requiring large amounts of
structural fill to be imported before construction could begin. The team had meetings prior to
construction to establish goals and the interviewee felt that their approach of openness and
understanding assisting in the team having good chemistry.

[11] This project was a 1-story recreational building of approximately 8,000 square foot that was
built to LEED standards. The project was completed using CMR as a delivery method and the
interview was conducted with the owner. This delivery method was chosen for familiarity and
because they felt that the level of quality and service were much higher than traditional
delivery. This project was a smaller, new construction portion of a larger residential remodel
project. Because of this the designer and CM had been working to together previously on the
other buildings prior to the start of this new construction. Some members of the project team

21
 had little experience, this being their first project in a leadership position. The project had some
setbacks due to design flaws, which required large amounts of repairs and rework after
discovery of the issues. The interviewee felt that the CM’s project management team was not
properly staffed and spread thin with contributed to the projects issues.

[12] This project was a 10,000-square foot, 1-story recreational building that was built to LEED
standards. The project was completed using a DB delivery method and the interview was
conducted with the owner’s representative. This project suffered from being outside of the
normal procedure for this owner due to funding constraints. The interviewee was not able to
provide much information about the project due to being placed in an observation position in
preparation for the transfer of ownership to his firm once the project was completed. The
interviewee did opine that it was “the worst project” of their career.

[13] This project was a 4-story, 50,000 square foot remodel of a medical building into a multi-unit
residential building. The project was completed as DBB delivery method and the interview
was conducted with the owner. They chose this delivery method because they felt it gave them
the best price and allowed for the most competition during procurement. Project team members
were brought on early to assist in pre-construction and assist with current market information.
This project had difficulties with overlapping design and construction activities due to funding
restrictions. The projects biggest challenge was not carrying enough contingency funds and
thus the project was over on budget and time. This was the owner’s first renovation project and
they had little experience with this process and with the other team members. This interviewee
felt that many lessons of how not to conduct a project were learned and that it was important to
financially tie the designer and builder into the project using the contract to develop a level of
ownership in the project success.

[14] This project was a 35,000 square foot, 2-story recreational building that was built to LEED
standards. The project was completed using a CMR delivery method and the interview was
conducted with the owner. This delivery method was chosen because allowed the project to
progress faster and was allowed by the project funding requirements. The CM firm was able to
hire trades that they had prior relationships with, as long as the owner could review their
proposal. Construction activities started while the design was progressing, which assisted in
addressing one of the project’s main challenge, preexisting utility structures that had to be
worked around. This was the interviewee’s first time using this delivery method. The
interviewee felt that many lessons were learned from the mistakes made in this project’s
delivery and that, as they have continued to use the CMR delivery method over the following
projects, it has become a much smoother and more successful delivery method for them.

[15] This project was a 1-story, 4,300 square foot, office building. The project was completed using
a CMR delivery method and the interview was conducted with owner’s representative. This
method was chosen because the owner prefers to have separation between the designer and
builder and prefers some conflict between members, which they feel keeps members more
accountable. There was no construction and design overlap in the completion of this project.
The project used different materials from other projects which slowed the projects delivery.

[16] This project was a 25,000 square foot, 1-story commercial building built to LEED
standards. The project was completed using a DBB delivery method and the interview was

22
conducted with a member of the general contracting firm. This delivery method was chosen
due to a previous relationship between the builder and the owner in the past, where they had
used this method before. The interviewee stated that there were no unexpected or significant
challenges faced by the project. The interviewee thought this was a well performing and
delivered project, contrary to what the data showed.

23