Choosing a delivery method

By now it should be clear that no delivery method is best in general. Each is a bundle of trade-offs, and the right choice depends on what the owner is trying to achieve, what the project is like, what the market will offer, which risks the owner can manage, and what the owner's own team can do. This lesson brings those factors together into a structured selection process of the kind public owners now use, and looks at how risk allocation differs across methods, a topic CON-350 treats in more depth.

Start with the owner's goals

The first step is to state what success means for this project and rank the goals. Owners often say they want everything: lowest cost, fastest schedule, highest quality, least risk and full control. Forcing a ranking is uncomfortable but essential, because the methods trade these off against each other.

Then look at the project, the market and the owner

With goals ranked, the owner considers factors that can make a method unsuitable regardless of goals. Some are pass/fail: a method the owner lacks legal authority to use, a funding condition that rules a method out, or a market where no contractor will accept the risk. Others shift the balance.

Market sounding is valuable here. Before choosing, many public owners meet potential bidders, describe the project and the methods under consideration, and ask how the market would respond. If no general contractor will take fixed-price ground risk on a tunnel, a lump-sum design-build will attract few bids or very high ones. If local trades are fully booked, a construction manager may struggle to get competitive trade prices.

Comparing risk allocation

One way to compare methods is to list the project's main risks, estimate their expected cost, and ask who carries each under each method, and how efficiently. A risk transferred to a party that cannot control it is usually priced at a premium; a risk retained by the owner costs its expected value plus whatever the owner loses through poor management. The aim is not to transfer as much risk as possible, but to give each risk to the party best able to manage it at the lowest cost.

# Expected cost of retained and transferred risk for a $150 million hospital addition (illustrative, $ millions)
risks = {   # risk: expected cost if retained by the owner
    "Design errors and coordination": 6.0, "Construction cost on known scope": 5.0,
    "Ground conditions": 3.0, "Schedule delay": 4.5, "Systems integration": 3.5,
}
# share of each risk transferred to the builder side under each method, and premium charged on transferred risk
methods = {
    "Design-bid-build": ({"Design errors and coordination": 0.0, "Construction cost on known scope": 1.0,
                          "Ground conditions": 0.2, "Schedule delay": 0.5, "Systems integration": 0.3}, 0.30),
    "CM at risk":       ({"Design errors and coordination": 0.3, "Construction cost on known scope": 0.9,
                          "Ground conditions": 0.2, "Schedule delay": 0.5, "Systems integration": 0.5}, 0.25),
    "Design-build":     ({"Design errors and coordination": 1.0, "Construction cost on known scope": 1.0,
                          "Ground conditions": 0.5, "Schedule delay": 0.8, "Systems integration": 0.8}, 0.35),
    "IPD":              ({"Design errors and coordination": 0.5, "Construction cost on known scope": 0.5,
                          "Ground conditions": 0.5, "Schedule delay": 0.5, "Systems integration": 0.5}, 0.10),
}
for m, (transfer, premium) in methods.items():
    retained = sum(v * (1 - transfer[k]) for k, v in risks.items())
    priced = sum(v * transfer[k] for k, v in risks.items()) * (1 + premium)
    print(f"{m:17} retained {retained:5.1f}  transferred and priced {priced:5.1f}  total {retained + priced:5.1f}")

Design-bid-build  retained  13.1  transferred and priced  11.6  total  24.7
CM at risk        retained  11.1  transferred and priced  13.6  total  24.7
Design-build      retained   3.1  transferred and priced  25.5  total  28.6
IPD               retained  11.0  transferred and priced  12.1  total  23.1

In this invented example the methods' total risk cost differs by a few million dollars, and the method that transfers the most risk is not the cheapest, because the market charges a premium for taking it. IPD looks cheapest here because the shared risk is assumed to be priced with a small premium, which is true only if the collaboration works. The numbers are only as good as the assumptions, but the exercise forces a useful discussion of which risks matter and who should hold them.

A weighted scoring matrix

Most structured selection tools combine pass/fail screens with a weighted scoring matrix. The owner chooses criteria from its goals and factors, weights them, and scores each candidate method against each criterion, usually in a workshop with its advisers. The Project Delivery Selection Matrix developed for transportation projects by the Colorado Department of Transportation and the University of Colorado, which the Federal Highway Administration has promoted, uses this approach, as do guides from the Design-Build Institute of America, the Construction Management Association of America and owners in both countries.

# Weighted scoring of delivery methods with a sensitivity test (illustrative)
import random
random.seed(340)
criteria = {"Schedule": 0.25, "Cost certainty": 0.20, "Owner control": 0.15,
            "Complexity and risk management": 0.25, "Owner capacity fit": 0.15}
scores = {   # 1 (poor) to 5 (excellent) from a selection workshop
    "Design-bid-build": {"Schedule": 2, "Cost certainty": 4, "Owner control": 5, "Complexity and risk management": 2, "Owner capacity fit": 5},
    "CM at risk":       {"Schedule": 4, "Cost certainty": 3, "Owner control": 4, "Complexity and risk management": 4, "Owner capacity fit": 4},
    "Design-build":     {"Schedule": 5, "Cost certainty": 4, "Owner control": 2, "Complexity and risk management": 3, "Owner capacity fit": 3},
    "Progressive DB":   {"Schedule": 4, "Cost certainty": 3, "Owner control": 3, "Complexity and risk management": 5, "Owner capacity fit": 3},
}
def total(m, w):
    return sum(w[c] * scores[m][c] for c in w)
base = {m: total(m, criteria) for m in scores}
for m, t in sorted(base.items(), key=lambda kv: -kv[1]):
    print(f"{m:17} {t:.2f}")
# sensitivity: perturb each weight by up to +/-40 % and renormalize, 2,000 trials
wins = {m: 0 for m in scores}
for _ in range(2000):
    w = {c: v * random.uniform(0.6, 1.4) for c, v in criteria.items()}
    s = sum(w.values()); w = {c: v / s for c, v in w.items()}
    wins[max(scores, key=lambda m: total(m, w))] += 1
print("Share of trials each method ranks first:", {m: f"{n / 20:.0f} %" for m, n in wins.items()})

CM at risk        3.80
Progressive DB    3.75
Design-build      3.55
Design-bid-build  3.30
Share of trials each method ranks first: {'Design-bid-build': '0 %', 'CM at risk': '73 %', 'Design-build': '2 %', 'Progressive DB': '25 %'}

The base scores put CM at risk first, narrowly ahead of progressive design-build. The sensitivity test shows how often each method stays first when the weights vary: if the leading method wins most trials, the choice is robust; if two methods trade places often, the owner should look harder at the criteria that separate them, or accept that either would be reasonable and decide on other grounds, such as market appetite or staff experience.

Common patterns

Although every project is different, experience suggests some common matches between project types and methods.

Documenting and revisiting the decision

Public owners should record the selection process: goals and weights, screens applied, methods considered, scores, risk analysis, market feedback and the final rationale. The record supports approvals, audits and lessons learned. It also helps if circumstances change. A method chosen in a competitive market may need revisiting if the market tightens; a design-build chosen for speed may need reconsidering if a funding delay removes the schedule pressure. Revisiting a decision before procurement starts is cheap; changing method after a procurement fails is expensive.

After completion, owners that compare outcomes, including cost growth, schedule growth, quality, claims and stakeholder satisfaction, against the reasons a method was chosen build an evidence base for future decisions. Many large public owners now publish such reviews.

Hybrid and packaged approaches

Large programs do not have to use one method for everything. Owners often package a project into several contracts with different methods: early works by design-bid-build, a main building by CM at risk, a central plant by design-build with a performance guarantee, and equipment by owner pre-purchase. Transit agencies commonly split civil works, stations, systems and vehicles into separate packages with different methods. Packaging lets each part use the method that suits its risks, but it creates interfaces between contracts, and someone, usually the owner or its program manager, must manage them. A packaging decision should therefore be scored like a delivery method, with interface risk counted explicitly.