Plan slowly, deliver quickly: where a project's money is exposed

A capital project has two phases that feel alike on a calendar and are nothing alike in a bank account. In the front end, a small team defines what is to be built, compares options, investigates the site and the services, prices the work and decides whether to go ahead. In delivery, contractors and suppliers build, install and commission what was defined. A week in the front end costs the salary of a few people and some studies. A week in delivery costs a site full of trades, equipment on order and money that cannot be taken back. The argument of this book is that weeks in the first phase are cheap, weeks in the second are dear, and a business that spends the cheap weeks well needs fewer of the dear ones.

This chapter sets out the running case: one fruit packer, one upgrade and two ways of starting it. It measures what a week costs in each phase, separates money spent from money committed, explains why the ability to change a project’s cost falls as spending rises, and summarises what published research has found about planning before delivery. It previews the two paths’ expected cost, which later chapters derive in full, and ends with a board paper that priced the front end as though the business were paying a delivery crew to wait.

Two paths for one upgrade

This is an illustration. Northbank Fruit Packers is a fictional family business in northern Victoria that packs apples and pears from its own orchards and from local growers. It handles about 18,000 bins a season at around 400 kg each, 7,200 t of fruit. Its six cold rooms hold 6,000 bins, so a large share of every harvest goes to hired cold storage, and its packing line is old and slow. All money in this book is in Australian dollars excluding GST, and “k” means thousands.

The 2028 upgrade adds four cold rooms of 1,000 bins each, a new refrigeration plant to serve the old rooms and the new ones, an imported optical grader for the packing line, and the electrical, civil, fire services and dock work they need. Once running, it is expected to earn a net A$1,400k a year: A$560k from the grader, A$640k from storage no longer hired and A$200k from energy. The managing director wants it in place for the 2028 harvest.

In early November 2026 the business has a concept: a site sketch, a list of equipment and an estimate of A$5,410k prepared from budget prices. Two ways forward are on the table.

• The rush path. Approve the concept estimate with 10% contingency, a budget of A$5,950k, in week 2 (Monday 16 November 2026), and award a design-and-construct contract at once. The contractor would design, seek approvals, order equipment and build, in a planned 52 weeks, finishing in week 54 (Monday 15 November 2027).
• The front-end path. Spend 22 weeks and A$248k defining the project first, through two further stages that Chapter 2 describes, and approve it at Gate 3 in week 22 (Monday 5 April 2027). Long-lead equipment is then ordered at approval and delivery runs a planned 36 weeks, to week 58 (Monday 13 December 2027).

Week numbers in this book count from Monday 2 November 2026 as week 0. The book follows the front-end path as what the business did, and uses the rush path as the comparison.

Figure 1Two ways to start Northbank's upgrade

	Rush path	Front-end path
Approval	Week 2, 16 November 2026	Week 22, 5 April 2027
Estimate at approval	A$5,410k, concept	A$6,428k, defined, including the A$248k front end
Planned delivery	52 weeks	36 weeks
Planned ready	Week 54, 15 November 2027	Week 58, 13 December 2027
Float to the 31 January 2028 deadline	11 weeks	7 weeks
Delivery spend per week	A$104k	A$171.7k
Money approved	A$5,950k budget	A$6,948k funding, A$6,700k for delivery

The rush path looks better on two lines of that table: it starts earlier and has more float. Most of this book is about why those two lines mislead.

What a week costs in each phase

The simplest measure of exposure is the burn rate: money spent per week in each phase. For the front-end path, the front end costs A$248k over 22 weeks, A$11.3k a week. Delivery costs the delivery base of A$6,180k over 36 weeks, A$171.7k a week. A week of delivery costs 15.2 times a week of front end.

The rush path’s delivery runs at A$5,410k over 52 weeks, about A$104k a week. It is slower because its 52 weeks include design, permits and supply applications that the front-end path completes before delivery begins, and because it runs on the concept estimate, which leaves out scope that Chapter 5 shows was found later.

Figure 2Spending per week by phase

Front end, FEL 2 and FEL 311.3 A$k a weekA$248k over 22 weeks

Rush delivery104 A$k a weekA$5,410k over 52 weeks

Planned delivery after Gate 3171.7 A$k a weekA$6,180k over 36 weeks

The ratio matters because the two phases are traded against each other. A week added to the front end costs about A$11.3k. A week of delivery avoided, because the work was defined well enough to run without stopping, saves the time-related costs that run every week a site is open, and a week of delay at the wrong time of year can cost far more than that, as Chapter 8 shows. A business that sees only the calendar treats both weeks as equal. A business that sees the burn rate does not.

Spent is not the same as committed

Burn rate measures money leaving the bank. It understates exposure, because the decisive moment in a project is not when money is paid but when the business becomes bound to pay it.

At Gate 3, the front-end path has spent A$248k, which is 3.9% of the A$6,428k total. On the same day, approval lets two fixed-price packages, the refrigeration plant and rooms at A$1,720k and the optical grader at A$1,480k. That commits A$3,200k, 51.8% of the delivery base, in one decision. More than half of delivery is bound before a sod is turned, which is exactly why the business wanted the front end done first: by the time the money is committed, the business knows what it is buying, what the site needs and when each item must arrive.

The rush path commits at week 2. A design-and-construct contract priced on a concept binds the business to the contractor’s price for the concept scope, and leaves every item the concept missed to be priced later as a variation, when the business has no competition to keep the price honest. By week 22, when the front-end path is just approving, the rush path would have spent about A$2,080k (20 weeks at about A$104k) and committed nearly all of its A$5,950k budget.

Figure 3Northbank's position at Gate 3, week 22

A$248kSpent on the front end3.9% of the A$6,428k total

A$3,200kCommitted at approval51.8% of the A$6,180k delivery base

A$270kSpent if the board said noincluding the A$22k concept stage

The last figure is the price of the option to stop. If the definition work had shown that the project did not pay, the business would have lost A$270k, including A$22k spent on the concept before week 0. On the rush path the same discovery would arrive after the contract was signed, when stopping means paying the contractor for work done and costs incurred. Committing before you know: spending, committing and the gate plan in Stage Gates and Stopping Rules by the Numbers develops the difference between spent and committed as a decision rule for gates.

Influence falls as spending rises

The ability to change a project’s cost is greatest when little has been spent. Paulson (1976), writing in the Journal of the Construction Division, described the level of influence over a project’s cost as falling steeply as the project moves from concept through design into construction: influence is greatest during design, when expenditure is still relatively small, and by the time construction is under way most of the cost has been fixed by decisions already taken. The curve is often drawn, but the point does not need numbers. It is the burn-rate table turned upside down.

Northbank’s case shows why. The decisions that set most of the upgrade’s cost are how many rooms to build, whether to replace the old refrigeration plant or add a second plant beside it, how to get enough electricity to the site, and how to arrange the grader in the packhouse. Each is made in the front end, for the price of studies and meetings. Once panels are on order and the slab is poured, the same decisions can still be changed, but only by paying for what has been built and building it again. A front end is the period in which the business buys its decisions at the lowest price they will ever have.

What the research says

Research on owners’ front-end planning has been led in the United States by the Construction Industry Institute (CII) and its university partners. Its findings come mostly from large industrial and building projects, and they are quoted here for their direction, not as numbers a small business should expect.

• Gibson, Kaczmarowski and Lore (1995), in the Journal of Construction Engineering and Management, described pre-project planning, also called feasibility analysis, conceptual planning or front-end planning, as the work at the interface between business and engineering. The owner is responsible for it, even though consultants perform many of its functions.
• Hamilton and Gibson (1996), in the Journal of Management in Engineering, used data from 62 industrial projects to test whether project success is related to the effort put into pre-project planning, and found that it is.
• Gibson, Wang, Cho and Pappas (2006) drew together five research projects over about 14 years, covering more than 200 capital projects, and reported a positive relationship between thorough pre-project planning and project performance.
• Merrow (2011), in Industrial Megaprojects, drawing on Independent Project Analysis’s project records, argues that front-end loading is the most reliable predictor of how large industrial projects turn out, and that cutting it short to save time or money is a leading cause of failure.

The Gibson definition puts the owner in charge, which is where a small business stands whether it likes it or not. Northbank cannot hand the front end to a contractor and expect the contractor to decide what the business needs. That decision starts from the outcome the upgrade must deliver, a point developed in From a request for new plant to the outcome behind it in Outcome-First Projects by the Numbers. The estimate the front end produces is still the team’s own view of its own project, however well defined, and The inside view and the plan on the table in Reference Class Forecasting by the Numbers explains why such estimates lean optimistic and how an outside view checks them.

The front end also has a place in the wider timeline. Time to value: the clock starts before the project does treats the front end mainly as an interval to be shortened, because much of it is waiting for decisions and approvals. That is right for waiting. This book is about the working part of the front end, which is worth buying deliberately, and about measuring what it buys.

The calendar that sets the deadline

Northbank’s upgrade has a date it cannot move. The 2028 harvest begins with first fruit intake on Monday 7 February 2028, week 66, and intake runs for 12 weeks. The packing line runs from intake until stored fruit runs out in late October, so the line can be changed only between 1 November and 31 January. The new rooms must be at temperature and the grader trialled before intake, which sets the ready-for-season deadline at Monday 31 January 2028, week 65.

Figure 4Northbank's calendar from concept to harvest

1. Week 0Front end beginsMonday 2 November 2026; concept stage complete, A$22k spent
2. Week 2Rush path approval16 November 2026; design-and-construct contract on the concept
3. Week 10Gate 211 January 2027; option chosen, Class 4 estimate
4. Week 22Gate 3, approval5 April 2027; long-lead packages let, A$3,200k committed
5. Week 54Rush path planned ready15 November 2027; 11 weeks of float
6. Week 58Front-end path planned ready13 December 2027; 7 weeks of float
7. Week 65Ready-for-season deadline31 January 2028; rooms at temperature, grader trialled
8. Week 66First fruit intake7 February 2028; benefits begin on either path

Two facts on that calendar shape every later chapter. First, benefits begin at week 66 on either path. Finishing in November rather than December earns nothing extra, because no fruit arrives until February. Second, the cost of missing the deadline is not a gentle rate per week. A grader that is not ready by the end of January cannot be installed until the following November, so its whole season’s benefit is lost at once. Chapter 8 prices that cliff.

A first look at what each path is expected to cost

Chapters 4 to 8 build up the full comparison. The result is previewed here so that the reader knows where the argument goes. Counting the base estimate, the scope found during delivery, the extra time-related costs of a longer delivery and the expected cost of missing the harvest, the rush path is expected to cost about A$7,451k, 1.252 times its A$5,950k budget. The front-end path is expected to cost about A$6,562k, including its A$248k front end, 0.944 of its A$6,948k funding. The expected saving is A$889k for A$248k of front-end work, about 3.6 times what it cost. A schedule simulation in Chapter 8 gives the rush path a 44.2% chance of missing the deadline and the front-end path 5.3%, despite the rush path’s larger float.

These are expectations from a model with stated assumptions, not promises. Chapter 7 explains the delivery window behind them and Chapter 8 the simulation.

The error: pricing the front end as a cost of waiting

This is an illustration. When the managing director proposed the rush path, the board paper set out the case against a front end in one line: “22 weeks of planning delays the start by 22 weeks, at A$171.7k a week, a cost of waiting of A$3,776.7k.” The figure was the delivery base, A$6,180k, multiplied by 22 ÷ 36, and it made the front end look as though it cost more than half the project.

The finance manager asked the question that a cost of waiting has to answer: what benefit is lost while we wait? The manager traced it through the calendar. The business earns nothing from the upgrade until fruit arrives in week 66, and both paths plan to be ready before then. A later start therefore loses no benefit at all. The A$171.7k a week is not a cost of waiting either: it is the price of delivery, which is paid once whichever week it starts. Starting sooner moves the same spending earlier, which, if anything, costs more in interest.

The real cost of the front end was A$248k of studies, design and engineering time, about one-fifteenth of the figure in the paper, plus one thing the paper had not mentioned: float. The front-end path plans to finish 4 weeks later than the rush path, so its float to the deadline falls from 11 weeks to 7. Whether 7 weeks of float on a defined project is safer than 11 weeks on an undefined one is a real question, and Chapter 8 answers it with a simulation.

The error was detected because the finance manager asked for the lost benefit to be named. It was corrected by replacing the line in the board paper with three: the cash the front end would spend, the benefit lost by a later start (none, because both paths aim at the same harvest), and the change in the chance of missing the harvest, to be estimated before approval. The general rule the business adopted is that a cost of waiting is the benefit lost, plus any costs that continue during the wait, and never the cost of the work that will be done afterwards.

Exercise: an amenities block

A business plans a A$900k staff amenities building. It proposes an 8-week front end costing A$36k, followed by a 20-week delivery costing A$864k. What is the burn rate in each phase, how many times dearer is a week of delivery, and what share of the total does the front end take?

Worked answer. The front end burns A$36k ÷ 8 = A$4.5k a week and delivery A$864k ÷ 20 = A$43.2k a week, so a week of delivery costs 43.2 ÷ 4.5 = 9.6 times a week of front end. The front end is A$36k ÷ A$900k = 4.0% of the total. The ratio is lower than Northbank’s 15.2 because amenities have a smaller share of their cost in equipment and specialist trades, but the conclusion holds: an extra week spent defining the building costs about a tenth of a week of building it, and is worth spending if it is likely to save more than a few days of delivery or a variation of more than A$4.5k.