Capstone: selecting and structuring a delivery method for a public facility

This capstone applies the whole course to one decision. You are on the owner's advisory team for a new hospital tower, and your job is to recommend a delivery method, explain why, and set out how the chosen method should be structured: how the partner will be selected, how it will be paid, which contracts will be used, how risks will be allocated, and how the owner will govern the project. The project and its numbers are invented, but the process is the one real public owners follow.

The project

Ridgeway Regional Health (an invented public hospital district in a US state) plans a new eight-story acute care tower on its existing campus, connected to the operating hospital at three levels. The tower will contain 240 inpatient beds, 12 operating rooms, an expanded emergency department and a central energy plant upgrade. The district is authorized by state law to use design-bid-build, CMAR and design-build, and the state's P3 statute covers health facilities. The estimated capital cost is $780 million in 2026 dollars, and the district wants the tower open in 2032.

Step 1: goals and screens

Workshops with the board, clinical leaders and facilities staff produce this ranking of goals: first, managing complexity and interface risk with the operating hospital; second, meeting the 2032 opening; third, cost certainty before the main construction commitment; fourth, clinical input and control over design; fifth, life-cycle performance. Lowest initial price through open competition ranks last, because the board has seen bid-and-claim cycles on earlier projects.

Screens: the district may use design-bid-build, CMAR or design-build under its enabling statutes, and a P3 under the state P3 law with a value-for-money analysis and board approval; progressive design-build is permitted under the design-build statute as amended. Federal funding is not expected, so federal procurement rules do not apply. Design-bid-build is legally available but scores poorly on the top two goals.

Market sounding produces three findings. Large general contractors are reluctant to take fixed-price risk for the connections to the operating hospital and the energy plant without extensive investigation. Two consortia express interest in a DBFM partnership but would price interface risk heavily. Mechanical and electrical trade partners are interested in early involvement because their capacity is limited and they want certainty of work.

Step 2: shortlist and risk analysis

The team shortlists four methods: CM at risk, lump-sum design-build, a DBFM P3, and an alliance or progressive model with early trade partner involvement. Design-bid-build and agency CM are dropped: design-bid-build because of interface risk and schedule, agency CM because the health corporation lacks staff to hold dozens of trade contracts next to an operating hospital.

# Ridgeway tower: expected risk cost by delivery method ($ millions, illustrative)
risks = {"Interfaces with operating hospital": 28.0, "Design development and clinical changes": 22.0,
         "Energy plant integration": 12.0, "Schedule delay": 18.0, "Trade market escalation": 15.0}
methods = {   # (share of each risk transferred to the builder side, premium on transferred risk)
    "CM at risk":       ([0.3, 0.2, 0.4, 0.5, 0.6], 0.25),
    "Lump-sum DB":      ([0.8, 0.6, 0.8, 0.8, 0.9], 0.45),
    "DBFM P3":          ([0.8, 0.5, 0.8, 0.9, 0.9], 0.50),
    "Alliance":         ([0.5, 0.5, 0.5, 0.5, 0.5], 0.10),
}
for m, (tr, prem) in methods.items():
    retained = sum(v * (1 - t) for v, t in zip(risks.values(), tr))
    priced = sum(v * t for v, t in zip(risks.values(), tr)) * (1 + prem)
    print(f"{m:12} retained {retained:5.1f}  priced {priced:5.1f}  total risk cost {retained + priced:6.1f}")

CM at risk   retained  59.4  priced  44.5  total risk cost  103.9
Lump-sum DB  retained  21.9  priced 106.0  total risk cost  127.9
DBFM P3      retained  22.3  priced 109.1  total risk cost  131.4
Alliance     retained  47.5  priced  52.3  total risk cost   99.8

The analysis reflects the market sounding: methods that push interface and integration risk onto the builder side attract high premiums, because bidders cannot control those risks well before the site and the existing hospital are investigated. The alliance shows the lowest total risk cost, but only on the assumption that the collaborative model works and shared risks are priced modestly; CM at risk is close behind. The team notes that the DBFM P3 would also bring finance and transaction costs, which a value-for-money test would need to justify.

Step 3: scoring and sensitivity

# Ridgeway tower: weighted scoring with sensitivity (illustrative)
import random
random.seed(1240)
w = {"Complexity and interfaces": 0.30, "2032 opening": 0.25, "Cost certainty": 0.15,
     "Clinical input and control": 0.20, "Life-cycle performance": 0.10}
s = {   # workshop scores, 1 to 5
    "CM at risk":  [4, 4, 3, 5, 2],
    "Lump-sum DB": [2, 4, 4, 2, 3],
    "DBFM P3":     [2, 3, 5, 2, 5],
    "Alliance":    [5, 4, 3, 5, 3],
}
crit = list(w)
def score(m, wt):
    return sum(wt[c] * s[m][i] for i, c in enumerate(crit))
for m in s:
    print(f"{m:12} {score(m, w):.2f}")
wins, lo, hi = {m: 0 for m in s}, {m: 9 for m in s}, {m: 0 for m in s}
for _ in range(3000):
    wt = {c: v * random.uniform(0.6, 1.4) for c, v in w.items()}
    t = sum(wt.values()); wt = {c: v / t for c, v in wt.items()}
    res = {m: score(m, wt) for m in s}
    wins[max(res, key=res.get)] += 1
    for m, v in res.items():
        lo[m], hi[m] = min(lo[m], v), max(hi[m], v)
for m in s:
    print(f"{m:12} first in {wins[m] / 30:.0f} % of trials; range {lo[m]:.2f} to {hi[m]:.2f}")

CM at risk   3.85
Lump-sum DB  2.90
DBFM P3      3.00
Alliance     4.25
CM at risk   first in 0 % of trials; range 3.61 to 4.07
Lump-sum DB  first in 0 % of trials; range 2.60 to 3.23
DBFM P3      first in 0 % of trials; range 2.66 to 3.39
Alliance     first in 100 % of trials; range 3.98 to 4.50

The alliance ranks first and stays first in every sensitivity trial, with CM at risk a clear second; CM at risk's best score in any trial is still below the alliance's typical score. The P3 and lump-sum design-build rank lower under almost any reasonable weights, because they score poorly on the two most important goals for this project. The team concludes that the choice is between a collaborative model and CM at risk, and looks at the owner's capacity and the market to decide.

Step 4: the recommendation

The team recommends a progressive, collaborative model: a multi-party IPD agreement adapted from the AIA C191 family, with the general contractor and the mechanical and electrical trade partners joining in a validation phase, followed by a target cost and shared profit pool. If the board or state law makes a multi-party agreement impractical, the fallback is CMAR under an AIA A133-2019 based contract with early trade partner involvement for mechanical and electrical work. The recommendation rests on four reasons: interface and integration risks are too uncertain to price well at a fixed price now; the market sounding showed contractors and trades would engage collaboratively but not take fixed-price interface risk; the district's goals favor managing complexity and clinical input; and value for money for a DBFM P3 is not demonstrated on current information.

Step 5: structuring the chosen method

# Ridgeway tower: target cost and profit at risk outcomes (illustrative, $ millions)
target = 745.0          # target outturn cost after development (cost of work incl. general conditions)
fee = 0.055 * target    # fee for corporate overhead and profit, at risk
gain_share, pain_share = 0.50, 0.50
def result(actual):
    diff = actual - target
    if diff <= 0:
        partner = fee - gain_share * diff
    else:
        partner = max(0.0, fee - pain_share * diff)   # partner risk limited to its fee
    owner = actual + partner
    return partner, owner
for actual in (715, 730, 745, 765, 790, 830):
    p, o = result(actual)
    print(f"actual {actual:4} -> partner fee {p:5.1f}, owner pays {o:6.1f}")
owner_budget = 780 + 0.06 * 780
print(f"Owner capital budget with 6 % owner reserve: {owner_budget:.1f}")

actual  715 -> partner fee  56.0, owner pays  771.0
actual  730 -> partner fee  48.5, owner pays  778.5
actual  745 -> partner fee  41.0, owner pays  786.0
actual  765 -> partner fee  31.0, owner pays  796.0
actual  790 -> partner fee  18.5, owner pays  808.5
actual  830 -> partner fee   0.0, owner pays  830.0
Owner capital budget with 6 % owner reserve: 826.8

The model shows that the owner's exposure grows quickly once the partner's fee is exhausted, at an actual cost of about $827 million, so the owner keeps a reserve above the target; here a 6 % reserve on the $780 million estimate, about $827 million in all, covers outcomes up to roughly the point where the partner's fee runs out. A realistic target from the development phase, strong governance and early work on the riskiest interfaces are what keep outcomes in the shared band.

Step 6: timeline

# Ridgeway tower: delivery timeline (months from business case approval; illustrative)
steps = [("Market sounding and model approval", 3), ("Partner selection (two stages)", 5),
         ("Development phase incl. early works", 14), ("Commercial close", 1),
         ("Construction", 40), ("Commissioning and transition", 6)]
t = 0
for name, d in steps:
    print(f"{name:38} months {t:3}-{t + d:3}")
    t += d
start_month = 1        # January 2026 approval assumed
year = 2026 + (start_month - 1 + t) // 12
print(f"Opening around month {t}, i.e. about {year}")

Market sounding and model approval     months   0-  3
Partner selection (two stages)         months   3-  8
Development phase incl. early works    months   8- 22
Commercial close                       months  22- 23
Construction                           months  23- 63
Commissioning and transition           months  63- 69
Opening around month 69, i.e. about 2031

The timeline reaches opening in about 69 months, late 2031, consistent with a 2032 target if the business case is approved early in 2026 and the development phase does not drift. The team flags the development phase as the critical risk to the timeline and recommends a firm schedule with decision gates.

Step 7: risks to the recommendation and their mitigation

Reflecting on the course

The Ridgeway exercise shows the main lessons of this course. A delivery method is a set of choices about organization, selection, pricing and payment, not a label. Each method trades off control, speed, certainty, collaboration and life-cycle accountability. Public owners work within legal, funding and fairness constraints that limit and shape their choices. Risk should sit with whoever can manage it best, and the market will tell you what it charges for the rest. A structured, documented selection process, tested for sensitivity and grounded in market feedback and an honest view of the owner's capacity, is the best protection against choosing a method by habit or fashion. CON-350 picks up where this course leaves off, with the detailed allocation, insurance and bonding of construction risk.