Alliances and collaborative contracts: one agreement, shared results

Chapters 3 to 5 moved the commercial result of the works step by step from the contractor alone towards the owner and contractor together. Under a lump sum the builder keeps every dollar saved and pays every dollar lost; under the target-cost contract of Chapter 5 the two share each dollar inside the bands. An alliance takes the last step. The owner and its key designers, contractors and suppliers sign one agreement, are paid their costs on open book, share the result against a jointly developed target, and make the project’s decisions together in a leadership group in which every member has a voice.

This chapter places the alliance on the spectrum of collaborative arrangements, sets out the three-limb payment as it would apply to the distribution centre, and works the share table at seven outturns to show where the cap on painshare bites and who pays after it does. It then designs the key result areas that reward performance beyond cost, puts a number on what the alliance costs to run, reviews what Australian evidence says about alliance outcomes, and ends with a schedule measure that paid for the plan instead of an improvement on it.

From partnering to alliance

Collaborative arrangements form a spectrum rather than a single contract type. At one end, partnering adds a charter, joint workshops and an agreed issue-resolution process to an otherwise conventional contract; the charter usually has no legal force and the commercial terms are unchanged. In the middle sit contracts with collaborative terms and a shared commercial result, such as the target-cost contract of Chapter 5 with early contractor involvement. At the other end is the alliance: a single agreement signed by the owner and all the key participants, open-book reimbursement of cost, a shared result against a target outturn cost, decisions taken unanimously by a leadership group drawn from every party, and, typically, commitments by the parties not to sue each other except in narrow cases such as wilful default.

Lahdenperä (2012) compared project partnering, project alliancing and integrated project delivery and found that they share a set of features in different combinations: early involvement of the parties who will build, open books, shared risk and reward, joint decision making and, in the stronger forms, a multi-party agreement. Walker and Lloyd-Walker (2015), drawing on interviews with 36 experienced practitioners, set out a taxonomy of relationship-based procurement in which alliancing sits at the most collaborative end. Both make the same practical point: the label matters less than which features are actually present and whether the money follows them.

The alliance is also the clearest written form of what Macneil (1980) called relational exchange, which Chapter 2 contrasted with discrete exchange. A long, uncertain project cannot be fully specified at signature, so the parties rely on the relationship to fill the gaps. A lump sum tries to price the gaps in advance; an alliance accepts that they cannot be priced and writes the relationship itself into the agreement.

Australian guidance and published forms

Australian public owners have used alliances widely for infrastructure, and national guidance exists for them. The arrangement described in the National Alliance Contracting Guidelines and widely used in Australia separates the owner from the non-owner participants (NOPs), develops the target outturn cost (TOC) jointly with the selected participants before it is committed, and pays the NOPs in three limbs. The national guidelines were first issued in 2011 and revised in 2015, after the Productivity Commission’s 2014 inquiry into public infrastructure. They cover when an alliance suits a project, how participants are chosen, and the commercial framework, and they are written for public owners. This chapter describes the arrangement only in general terms; the share ratio, cap and key result areas used below are the case’s own, not values taken from the guidelines.

A published multi-party form also exists: the NEC4 Alliance Contract, launched in June 2018, is a single contract signed by the client and all its alliance members together. It is named here only as an example; its clauses are not discussed.

How the owner selects alliance participants, and how the TOC is developed and reconciled, is covered in Alliance contracting: choosing partners before price and keeping open-book costs disciplined. This chapter does not repeat it. It works the commercial arithmetic that follows once the TOC is set.

The TOC is also the number the owner commits to. In Australian practice it is usually approved by the owner at the end of a development phase, before delivery is committed, so the owner can still decline to proceed if the target is unaffordable. Committing before you know: spending, committing and the gate plan in Stage Gates and Stopping Rules by the Numbers sets out how a gate plan separates spending on development from committing to delivery.

The three limbs in the case

This is an illustration. All amounts are A$ million excluding GST. Suppose the manufacturer delivers the works as an alliance. The members are the owner, the designer, the builder and the conveyor and controls integrator. The racking supplier-installer and the warehouse management system (WMS) vendor stay outside the alliance on fixed-price supply contracts to it: their scopes are standard products, well defined and small, and bringing two more firms into the leadership group would add more coordination than it removes. Their interfaces stay in the register of Chapter 1, managed by the alliance as a client would manage them.

The three NOPs are paid in three limbs:

• Limb 1: their actual direct costs of the project, reimbursed in full on open book and audited. The TOC for limb 1 is 13.50, the P50 direct cost used throughout the book.
• Limb 2: corporate overhead and profit, a fixed 1.080 (8% of the TOC), the same fee as the cost-based models of Chapters 4 and 5.
• Limb 3: a 50/50 share of any underrun or overrun of actual limb 1 against the TOC. The NOPs’ painshare is capped at their limb 2, so at worst they lose their whole fee but never their direct costs; their gainshare is uncapped.

On top of the three limbs sits a key result area (KRA) pool of ±0.30, paid or deducted according to performance on four measures beyond cost. The NOPs’ half of limb 3 and the KRA payments are divided among the three of them in proportions set in the agreement; this chapter treats them together.

Setting the TOC at the P50 is a deliberate choice: it is the cost the alliance is as likely to beat as to exceed, so neither side starts the job with a gain or a loss already built in. Reading the distribution: medians, percentiles and P-values in Reference Class Forecasting by the Numbers explains how to read a P50 from a forecast distribution and why it differs from the P80 that an owner funds.

Figure 1How the DC alliance pays its non-owner participants

1. 1Limb 1: actual direct costreimbursed in full on open book; target outturn cost A$13.50m
2. 2Limb 2: corporate overhead and profitfixed A$1.080m, 8% of the TOC
3. 3Limb 3: share of underrun or overrun50/50 against the TOC; NOP pain capped at A$1.080m, gain uncapped
4. 4KRA pool±A$0.30m on safety, quality, schedule and operational readiness

Painshare and where the cap bites

Below the target every dollar saved is split evenly: the owner pays 50 cents less and the NOPs earn 50 cents more. Above the target every dollar of overrun is also split evenly, until the NOPs’ share reaches their whole limb 2. That happens when the overrun reaches twice limb 2: 2 × 1.080 = 2.160, an actual limb 1 of 15.66. Beyond that point the NOPs earn nothing more and lose nothing more, and the owner pays every extra dollar.

Figure 2Owner payment and NOP profit at seven outturns, KRA score zero

Actual limb 1	Under (+) or over (−) TOC	NOP gain or pain	NOP profit	Owner pays
12.50	+1.000	+0.500	1.580	14.080
13.00	+0.500	+0.250	1.330	14.330
13.50	0.000	0.000	1.080	14.580
14.00	−0.500	−0.250	0.830	14.830
15.00	−1.500	−0.750	0.330	15.330
15.66	−2.160	−1.080	0.000	15.660
16.50	−3.000	−1.080	0.000	16.500

At an actual limb 1 of 15.00, the owner pays 15.330, which is 0.750 above the 14.580 it would pay at target: exactly half of the 1.500 overrun. At 16.50 the owner pays 16.500, which is 1.920 above 14.580, so it bears 1.920 of the 3.000 overrun, or 64%, and its share of each further dollar is 100%.

The comparison with Chapter 5 shows what the cap gives away. Under the target-cost contract’s bands, the contractor’s share of overrun rises to 80% beyond 110% of target, and at an actual cost of 16.20 the owner pays 15.525. Under the alliance at the same 16.20 the owner pays 16.200, 0.675 more, because the NOPs’ pain stopped at 1.080. The target-cost contract pushes harder on the contractor in the tail; the alliance protects the NOPs in the tail and leaves it with the owner.

The reason for the cap is behavioural. A firm whose own direct costs are at risk on a failing project will defend its position, and the joint decision making that justifies the alliance breaks down exactly when it is needed. Capping pain at limb 2 keeps every NOP solvent and at the table whatever happens. The owner pays for that with the tail of the cost distribution, which is why it must be confident that the TOC is realistic and that the alliance’s practices will keep the outcome away from the cap.

A target that is too generous is the other side of the same risk. Chapter 5 showed that a target set 6% high, at 14.31, pays the contractor 0.405 of gainshare at a 50% share for a saving that never happened. The alliance shares at the same 50%, so the same fat target would cost the owner the same amount. That is why owners pay for an independent estimator to test the TOC before it is committed, and why that estimator appears in the alliance’s running cost below.

Key result areas

Cost is not the only result the manufacturer cares about. The KRAs tie part of the NOPs’ reward to four outcomes that the shared scorecard of Chapter 9 also tracks. Each KRA is scored from −1 to +1, multiplied by its weight and by the ±0.30 pool.

Figure 3Maximum KRA payment or deduction by result area

Safety: injury frequency and critical-risk audit score, 25%75 A$k

Quality: defects at practical completion, target 40 or fewer, 20%60 A$k

Schedule: DC operating date, 30%90 A$k

Operational readiness: order lines per hour in week 4, target 1,150, 25%75 A$k

A KRA is worth having only if it rewards performance beyond what the TOC already pays for. The TOC is built on a programme, a quality plan and a safety plan, and the NOPs are paid their fee for delivering those. A KRA score of zero should therefore mean “delivered what the TOC assumed”, a positive score something better, and a negative score something worse. The scale must be measurable by someone independent of the participants, and the measures should not reward one result at the expense of another, such as a fast operating date bought by handing over with 90 defects. The operational-readiness KRA matters most to this owner, because a DC that opens on time but cannot pick 1,150 order lines an hour in its fourth week still leaves the third-party stores in use.

What the alliance costs to run

An alliance needs more management than a conventional contract. In the case, the owner’s external cost of running it is A$400k over 18 months, a teaching assumption that covers the alliance leadership team’s facilitation, the independent estimator who tests the TOC, open-book audits of limb 1, workshops and the owner’s share of a joint office. That is A$180k more than the A$220k of external administration for the target-cost contract; Chapter 10 sets out the administration costs of every model together.

The alliance must earn that overhead through lower cost or better results. The case assumes, as a teaching assumption and not a research finding, that the alliance’s joint planning and open problem solving bring direct cost 2% below the neutral estimate: a behaviour factor of 0.98, or 13.23 at the median. Chapter 10 states the factors for all five models; the target-cost contract’s is 1.01.

At the median, with no owner changes and no integration event, the arithmetic is close. The alliance’s limb 1 of 13.23 is 0.270 under the TOC, so the NOPs gain 0.135 and the owner pays 13.23 + 1.080 + 0.135 = 14.445, or 14.845 with the alliance overhead. The target-cost contract’s direct cost of 13.635 is 0.135 over target, the contractor bears 0.0675, and the owner pays 14.6475, or 14.868 with its administration. The alliance is ahead by about A$22.5k, a margin that rests entirely on the behaviour assumption. Chapter 10 tests the comparison across the full range of outcomes, where the cap and the integration event change the answer.

What the evidence says

Wood and Duffield (2009) studied 14 Australian public alliance projects over five years for the Victorian Department of Treasury and Finance. They found that alliances performed well on non-cost outcomes and allowed early starts, but that outturn costs were well above the estimates in the business cases on which the projects had been approved. The finding led to closer scrutiny of how alliance targets were set.

The study’s comparison was with the business case, not with the TOC. An alliance can finish close to its TOC and still be far above the budget the board approved, if the TOC itself grew during development. Which baseline an overrun is measured against changes the answer completely; Measuring overrun: baselines, real dollars and scope sets out why the baseline should be the budget at the decision to proceed. For the manufacturer, the test is the A$15.60m works allowance, not the TOC.

Bresnen and Marshall (2000) offered a broader caution: collaborative arrangements do not change behaviour by themselves. A charter, a multi-party agreement or a shared office achieves little unless incentives, decision rights and daily practice change with it. Chapters 7 to 9 deal with those practices, and they apply under any contract.

The error: a schedule KRA that paid for the plan

This is an illustration. In the draft alliance agreement, the schedule KRA scored the DC operating date on a linear scale: zero at operation at the end of month 18, +1 at 3 weeks early and −1 at 3 weeks late. Each week early was worth one third of the schedule KRA’s A$90k, or A$30k.

The TOC’s own programme, however, was not a month-18 programme. When the alliance developed the TOC it had re-sequenced the racking and conveyor installation, and the P50 of its programme showed operation 2 weeks early. The cost of that programme was already in the TOC. Delivering the TOC’s own P50 would score 2 ÷ 3 = +0.67 on schedule and pay 0.30 × 0.67 × A$300k = A$60.3k for business as usual.

The error was found by the independent reviewer engaged to assure the TOC, who checked every KRA threshold against the basis of the TOC and found that the schedule scale was anchored on the contract date while the cost was anchored on the alliance’s own programme. The other three KRAs were already anchored on the TOC basis.

The scale was corrected to score zero at the TOC programme’s P50, +1 at 4 weeks better and −1 at 4 weeks worse. Under the corrected scale, operation 3 weeks early, one week better than the TOC programme, scores +0.25 and pays 0.30 × 0.25 × A$300k = A$22.5k. Each week better than the P50 now earns the NOPs A$22.5k against the owner’s A$80k a week cost of delay, so the owner keeps about 72% of the value of each week saved and the NOPs have a clear reason to pursue it.

Figure 4Checks on a KRA before the agreement is signed

• Do: Zero score equals the performance the TOC already pays forprogramme, defect allowance and safety plan in the TOC basis
• Do: Each measure can be scored by someone independent of the participantsdata source, timing and method named in the agreement
• Do: Positive and negative scores are possible and symmetrical±1 at a stated distance from the basis
• Do: The value of one point is compared with the owner's value of the resultA$22.5k a week against A$80k a week of delay
• Avoid: Measures that can be met by trading one result against anotheran early date bought with 90 defects at practical completion
• Ask: Is the schedule scale anchored on the TOC programme or on the contract dateanchored on month 18 while the TOC programme showed 2 weeks early

Exercise: a moderate overrun with good results

The alliance finishes with an actual limb 1 of 14.2 and an overall KRA score of +0.4. What is the NOPs’ painshare, their KRA payment and their total profit, and what does the owner pay for the works? What would the owner pay, KRA aside, if limb 1 had reached 16.5?

Worked answer. The overrun is 14.2 − 13.50 = 0.700, inside the cap, so the NOPs bear half: a painshare of 0.350. The KRA payment is 0.4 × 0.30 = +0.120. NOP profit is limb 2 less pain plus KRA: 1.080 − 0.350 + 0.120 = 0.850. The owner pays limb 1 plus that profit: 14.2 + 0.850 = 15.050, or 15.450 with the A$400k alliance overhead, inside the 15.60 works allowance. At 16.5, the overrun of 3.000 would take the NOPs’ pain to 1.500 without the cap; the cap holds it at 1.080, so the owner pays 16.500, carrying 1.920 of the overrun. The A$120k of KRA payment in the first case is money well spent only if each scale was anchored on the TOC basis, so that +0.4 means results genuinely better than the TOC assumed.