The exposure window: delivering quickly once committed

Once the major money is committed, a project is exposed. Contracts are signed, equipment is on order, the site is open, and every week that passes adds cost that cannot be taken back: site overheads, the owner’s team, interest on money that is not yet earning, and the chance that something goes wrong while the work is half done. The time from that commitment to the day the asset is ready is the exposure window. The case for planning slowly is, in large part, a case for making that window short and predictable.

This chapter measures the window for both of Northbank’s paths, prices what accumulates during it, and shows why starting delivery earlier does not bring the benefits forward when they wait for a season. It describes how the front end shortens the window, notes the safety duties that run through it, and ends with an early start that was meant to protect float and spent A$130k and two weeks of it instead.

The window from commitment to readiness

The window opens when the business can no longer walk away cheaply. On the rush path that is week 2, Monday 16 November 2026, when a design-and-construct contract is signed on the Class 5 concept. On the front-end path it is Gate 3 at week 22, Monday 5 April 2027, when the refrigeration and grader packages are let at fixed prices. The window closes when the rooms are at temperature and the grader has been trialled.

This is an illustration. The rush path plans 52 weeks of delivery, to week 54; the front-end path plans 36 weeks, to week 58. Those are the plans. The schedule simulation explained in Chapter 8 gives the expected lengths: a mean of 62.2 weeks on the rush path and 37.9 weeks on the front-end path. The rush’s expected overrun is 10.2 weeks; the front-end path’s is 1.9 weeks. A window that starts earlier but is less well defined is not only longer on paper; it stretches further in practice, because the work still to be designed, approved and connected is done inside it rather than before it.

Figure 1The two exposure windows

Measure	Rush path	Front-end path
Window opens	week 2, 16 Nov 2026	week 22, 5 Apr 2027
Basis of commitment	Class 5 concept, A$5,410k	Class 3 design, A$6,180k delivery base
Planned length	52 weeks	36 weeks
Mean length in simulation	62.2 weeks	37.9 weeks
Mean overrun	10.2 weeks	1.9 weeks
Expected time-related cost at A$14k a week	A$152k	A$34k
Holding cost before the harvest	A$355.8k	A$301.7k

The idea of an exposure window is not new to this series. Modular Projects by the Numbers applies it inside a single installation of a few days on a working production line, and models disruptive events arriving at random while the line is open, in Short exposure windows: installing in days, not weeks. This chapter takes the wider view: the whole delivery phase, measured in weeks, and the money and time that accumulate across it.

What accumulates while the window is open

Four things grow with the length of the window.

Time-related cost. Preliminaries, site supervision, temporary facilities, hired plant and the owner’s project team are paid by the week. Northbank’s case puts the cost of each week beyond the planned duration at A$14k excluding GST. On the rush path’s expected overrun that is A$152k; on the front-end path’s, A$34k.

Committed money. Once a fixed-price package is let, most of its value is committed whether or not the project continues. At Gate 3 the front-end path commits A$3,200k, 51.8% of its delivery base, but it does so on a design that has been scored, estimated to Class 3 and reviewed. The rush path commits its whole contract at week 2, on a concept that has not yet met the electrical distributor, the geotechnical engineer or the fire engineer.

Capital tied up before benefits. Money spent on a cold store earns nothing until fruit goes into it. Until then it has a cost, the return the business could have earned on it elsewhere or the interest it pays on the loan facility.

Exposure to events. Each week the site is open is a week for wet weather, a late delivery, a supplier’s failure, a request for change or an injury. The front-end path does not remove these, but it gives them fewer weeks in which to happen.

A short window means a fast burn. The front-end path’s planned delivery spends about A$171.7k a week against the rush path’s A$104k, because the same work, and more of it, is packed into 36 weeks instead of 52. That is not a sign of extravagance. The weekly costs that depend on time, such as supervision, site facilities and the owner’s team, are paid for fewer weeks, and the costs that depend on quantity, such as panels, plant and concrete, are the same whenever they are bought. A high delivery burn rate on a defined scope is what a short window looks like from the accounts.

Holding cost before the harvest

The third item is the easiest to price and the most often ignored. If money is spent evenly across a period, on average it has been spent at the midpoint, and it waits from there until benefits start. The holding cost is the amount, multiplied by the average number of weeks it waits, multiplied by the weekly cost of capital. Northbank uses 9% a year, so a week costs 0.09 ÷ 52 = 0.173% of the amount held.

Figure 2Holding cost of money spent before benefits begin

holding cost = amount × average weeks before benefit × 0.09 ÷ 52

amount

Money spent in the period, A$k

average weeks before benefit

Week 66 less the midpoint of the period

0.09 ÷ 52

Northbank's cost of capital, 9% a year, per week

1. Rush: 5,410 × (66 − 28) × 0.09 ÷ 52 = 5,410 × 38 × 0.09 ÷ 52 = A$355.8k
2. Front-end path: 248 × (66 − 11) × 0.09 ÷ 52 = A$23.6k for the front end, plus 6,180 × (66 − 40) × 0.09 ÷ 52 = A$278.1k for delivery
3. Difference: 355.8 − (23.6 + 278.1) = 355.8 − 301.7 = A$54.1k in favour of the front-end path

This is an illustration. The front-end path spends more money, A$6,428k against the rush’s A$5,410k base, yet carries A$54.1k less holding cost, because almost all of its money goes in later and waits less. The comparison also flatters the rush: its A$1,425k of scope found during delivery would add more holding cost on top, and so would its expected overrun, which pushes spending later but not benefits earlier.

A$54.1k is not the main argument for the front end. Chapter 5’s late premium of A$435k and Chapter 8’s expected cost of delay are larger. It matters because it runs against intuition: a business that thinks of early starts as cheap insurance rarely notices that money spent early on a seasonal asset is money that waits.

Starting earlier does not earn earlier

The rush path’s case rests on a reasonable-sounding idea: start now, and there is more time. The trouble is that the extra time arrives at the wrong end. Benefits on both paths begin with the 2028 harvest, at week 66. The packing line can be changed only between 1 November and 31 January, and fruit arrives on 7 February 2028 whether or not the rooms are ready. A rush delivery that finished on plan, at week 54 on 15 November 2027, would then wait 12 weeks for fruit. Of the rush’s 38 average weeks of holding, 12 are weeks when the finished plant waits; at 9% a year they cost 5,410 × 12 × 0.09 ÷ 52 = A$112.4k.

The early start does buy float: 11 weeks against 7. But float is only as useful as the delivery is predictable. The rush’s expected overrun of 10.2 weeks consumes almost all of its 11 weeks of float on average, and much more in the bad cases, because a concept-based delivery still has to find its supply upgrade, its ground conditions and its fire services while the clock runs. The front-end path’s 7 weeks of float face an expected overrun of only 1.9 weeks. Chapter 8 turns this into probabilities; here it is enough to see that more calendar does not mean more margin when the work inside it is less known.

Reinertsen (2009) argues that the cost of delay should be quantified rather than assumed, because it drives most decisions about time. On a seasonal project it has an unusual shape: zero for every week before the deadline, then a large step. Delivering quickly, in that setting, means delivering in a short and predictable window that ends before the deadline. It does not mean starting as soon as possible.

How definition shortens the window

The front-end path’s window is 16 weeks shorter on plan and about 24 weeks shorter on average. Most of that comes from work moved out of the window and done before it:

• Long-lead items ordered at Gate 3. The refrigeration plant (26 weeks) and the imported optical grader (30 weeks) are ordered on the day of approval against specifications already written, so they arrive at weeks 48 and 52, in time for installation in the November to January line-change window.
• Approvals and connection already in train. The planning permit was lodged at week 8 and is granted at week 24. The distributor connection was applied for at week 12, offered at week 20 and accepted at week 22; its 30 weeks of works finish at week 52. On the rush path, both start from scratch inside the window.
• Fixed-price packages on a developed design. The two largest packages are let on a defined scope, so the contractors build rather than design, and changes are fewer.
• Known site conditions. The ground improvement, the fire services upgrade and the supply upgrade are in the design and the estimate, not waiting to be discovered by an excavator or a building surveyor.

How the packages are let matters as much as when. A fixed price on a defined scope gives cost certainty and lets work start the day it is approved; on an undefined scope it buys a stream of variations. How contracts shape behaviour: who keeps a dollar saved in Project Teams and Contracts by the Numbers sets out the trade between a contract’s incentive to cut cost and its flexibility to change the work, which is why a fixed price suits a scope that has been defined first.

Figure 3Delivery windows against the harvest deadline

1. Week 2, 16 Nov 2026Rush path window opensdesign-and-construct contract signed on the Class 5 concept
2. Week 22, 5 Apr 2027Front-end path window opensGate 3: A$3,200k of fixed-price packages; plant and grader ordered
3. Week 54, 15 Nov 2027Rush planned ready11 weeks of float; mean ready week 64.2
4. Week 58, 13 Dec 2027Front-end path planned ready7 weeks of float; mean ready week 59.9
5. Week 65, 31 Jan 2028Ready-for-season deadlinerooms at temperature, grader trialled
6. Week 66, 7 Feb 2028First fruit intakebenefits begin on both paths

Safety while the window is open

Northbank’s delivery runs from April 2027 to January 2028 beside a packhouse that keeps packing stored fruit until late October, with forklifts, bins and staff moving past an open construction site. A short window reduces the weeks of that exposure; it does not reduce the duties that apply in each of them. Victoria is the one Australian jurisdiction that has not implemented the model work health and safety laws: construction work there is governed by the Occupational Health and Safety Act 2004 (Vic) and its regulations, administered by WorkSafe Victoria, and that is where Northbank and its contractors look for their duties and guidance. Safe Work Australia’s Model Code of Practice: Construction work (November 2024) is national model guidance for the jurisdictions that have adopted the model laws; it does not apply in Victoria. The relevant point for planning is that pressure to recover lost weeks is itself a hazard: a programme with no float invites the shortcuts that cause injuries, which is one more reason to protect float by definition rather than by haste.

The error: the slab poured early

This is an illustration. In week 24, with the planning permit just granted, the civil contractor was ready to pour the plant room slab. The refrigeration supplier’s layout drawings, which would show the plinth positions and any pits, were due in week 27. The operations manager, keen to protect the seven weeks of float, authorised the pour to the concept plinth layout. The slab cost A$86k.

The supplier’s drawings arrived on time in week 27. They showed larger compressor plinths than the concept had allowed and a drainage pit under the condenser manifold. The problem was found at the routine drawing review, when the project engineer overlaid the supplier’s plinth and pit positions on the as-poured slab and found that neither could be cut in without breaking out the reinforcement. The slab had to be demolished for A$38k and poured again for A$92k: A$130k on top of the original A$86k, and 2 weeks lost on the plant room path.

The time cost more than it first appeared. At A$14k a week, two weeks of time-related cost add A$28k. In the schedule model, losing two weeks early in delivery has about the same effect as starting two weeks later: the chance of missing the 31 January 2028 deadline rises from 5.3% to 15.6%, and the expected cost of delay from A$36k to A$110k. Three weeks of waiting for drawings would have cost nothing; not waiting cost A$158k of direct and time-related money, plus a higher risk of missing the harvest.

The correction was a rule, written into the project’s procedures and later into Northbank’s checklist: no early works start on any element before that element’s design basis is frozen, and the person authorising an early start must name the drawing or specification that freezes it. The early start had looked like the opposite of rushing; in fact it was the rush path’s error in miniature, committing money on a concept to gain time that the plan did not need.

Figure 4Before authorising early works

• Do: Design basis for the element frozenthe supplier's layout or the engineer's drawing named and issued
• Do: Interfaces with ordered equipment checkedplinths, pits, penetrations, loads and services
• Do: Schedule benefit stated in weeksand checked against the float the plan actually needs
• Do: Cost of being wrong estimateddemolition, rework and weeks lost
• Avoid: Pouring or building to a concept layout to protect floatthe slab: A$130k and 2 weeks

Exercise: the price of waiting money

Using a cost of capital of 9% a year and spending spread evenly across each period, compute the holding cost of both paths up to first intake at week 66, and the difference. Then compute how much of the rush path’s holding cost arises in the 12 weeks its finished plant would wait for fruit.

Worked answer. Rush: A$5,410k spent from week 2 to week 54 has its midpoint at week 28 and waits 38 weeks, so 5,410 × 38 × 0.09 ÷ 52 = A$355.8k. Front-end path: A$248k of front end spent from week 0 to week 22 waits 55 weeks, giving A$23.6k; A$6,180k of delivery spent from week 22 to week 58 waits 26 weeks, giving A$278.1k; together A$301.7k. The difference is 355.8 − 301.7 = A$54.1k in favour of the front-end path, although it spends A$1,018k more. The rush’s 12 idle weeks hold the full A$5,410k: 5,410 × 12 × 0.09 ÷ 52 = A$112.4k, almost a third of its holding cost, paid for float that its expected overrun of 10.2 weeks would mostly use up.