Committing before you know: spending, committing and the gate plan

Most of the money a development programme loses is not lost on the day it is spent. It is lost on the day it is committed: when a purchase order is signed, a supplier’s slot is booked, a product is announced to dealers, or a launch date is promised. After that day, stopping still saves the unspent cash only if the commitment can be undone, and the terms for undoing it usually get worse every week. A gate process exists to keep commitments in step with knowledge, so that the business commits to each stage only when it knows enough to judge it.

This chapter separates three amounts that budgets usually blur: money spent, money committed but not yet spent, and money still uncommitted. It introduces the running case of the book, an agricultural-equipment maker developing a camera spot-spray boom, looks at what its own product history says about committing early, sets out the programme’s gate plan, and ends with a tooling order that was placed ten days after a gate approved only development work. Chapter 2 builds the decision tree that puts a value on that gate, and Chapter 3 values the right to stop that the early order nearly gave away.

Why programmes lock in early

Programmes commit before they know for reasons that look sensible at the time. Long-lead items such as tooling, castings, electronic modules and production slots have lead times of months, and someone works backwards from the launch date and concludes that the order must go in now. Suppliers offer a better price or a guaranteed slot for an early order. Sales teams want something to show dealers and customers, and once a product has been shown at a field day or listed in a dealer’s price book, stopping it has a visible cost in reputation. And the people closest to the work are, quite properly, enthusiastic about it: a champion who did not believe in the product would not be leading it.

None of these reasons is wrong in itself. Lead times are real, and a late launch can cost a season of sales. The difficulty is that each early commitment converts a decision that could have waited for evidence into one that has already been made. The business loses the right to stop cheaply, and it usually does so without anyone deciding to give that right away. The commitment is made by a purchase order or a promise, under someone’s delegated authority, rather than at a gate where the people accountable for the money can weigh it.

Spent, committed and uncommitted

Every dollar in a programme budget is in one of three states at any moment.

• Spent: paid or accrued for work done and goods received. Spent money is gone whatever the business decides next. It may have left behind something with resale or reuse value, but the spending itself cannot be undone.
• Committed but unspent: promised by a signed order, contract or binding undertaking, but not yet paid. Whether this money can still be saved depends on the cancellation terms, which often change on known dates, for example when a toolmaker cuts steel or a supplier starts a production run.
• Uncommitted: in the budget, but not yet promised to anyone. This money is still fully available to the business if it stops.

The quantity that matters at a gate is the money that stopping would still save: the uncommitted budget plus whatever committed money can be recovered under the cancellation terms. A programme can be well within its spending budget and yet have most of its future money committed, so that a gate held next month has little left to decide. That is why reporting spend against budget, which most accounting systems do well, is not enough. A gate needs a commitment register: a list of every order and undertaking, its amount, how much is spent, how much is still cancellable, on what terms, and the date on which those terms next change.

Plan slowly, deliver quickly: where a project’s money is exposed in Front-End Planning by the Numbers traces the same distinction through a construction project, where the money becomes exposed when the main packages are let.

The running case: Corella Plains Implements

This is an illustration. Corella Plains Implements is a family-owned maker of trailed boomsprays and chaser bins in the Wimmera region of western Victoria, with about 70 staff. All money in the case is in Australian dollars excluding GST, and the business and every number in it are invented.

In February 2025 Corella starts the SB36 programme: a 36 m trailed boomspray with cameras and individual nozzle control that detects weeds in fallow paddocks and sprays only the weeds. Five people carry the decisions, and the book refers to them by role:

• the managing director, who owns the business, chairs Gates 3 to 5 and makes the decisions;
• the engineering manager, programme lead and its champion, who presents at gates and may approve spending requests up to A$75,000 each under delegated authority;
• the finance manager, who keeps the monthly commitment register and is a gatekeeper;
• the sales and dealer manager, who runs dealer relations and grower demonstrations and is a gatekeeper;
• an independent board adviser, a gatekeeper from Gate 3.

What Corella’s own record says

Before approving the SB36, the finance manager looked back at the eight products Corella had developed between 2014 and 2024. Four of the eight failed or were stopped, and the business spent A$2,813,000 on those four before they ended.

Figure 1Money spent on Corella's four failed or stopped products, 2014–2024

H5 air-seeder cart, 2018A$1,640kdevelopment 790 + launch 850; withdrawn after 2 seasons

H2 round-bale handler, 2015A$470kdevelopment 230 + launch 240; withdrawn after 2 seasons

H3 rock picker, 2016A$455kstopped after prototype; 87.5% of its development budget

H7 hydraulic boom fold kit, 2021A$248kstopped before launch; 107.8% of its development budget

The pattern matters more than the total. The two products that were stopped before launch, the rock picker and the boom fold kit, cost A$455,000 and A$248,000. They were stopped late in development, at 87.5% and 107.8% of their development budgets, but before the launch money was committed. The two that were withdrawn after two seasons on the market, the round-bale handler and the air-seeder cart, cost A$470,000 and A$1,640,000, because by the time the market had given its answer the business had paid for tooling, inventory and dealer stock as well as development. A$1,090,000 of the A$2,813,000 was launch spending on products that the market then rejected.

Two other facts in the record point the same way. Six of the eight products overran their development budgets, with actual cost averaging 1.157 times budget. And Corella had held 11 gate reviews between 2019 and 2024 on the three most recent products, every one of which ended in “go”. A gate that has never said anything other than “go” has not yet shown that it can stop anything. Chapter 3 puts a dollar figure on what that costs.

The SB36 gate plan

The SB36 programme is planned as five stages separated by five gates, following the stage-and-gate structure described by Cooper (1990): each stage is a defined block of work that buys information, and each gate is a meeting at which the people who control the money decide whether to go on to the next stage or to kill the project. Cooper’s later work (Cooper, Winning at New Products, 5th edition, 2017) adds the must-meet and should-meet criteria and the gatekeepers who apply them. The arithmetic in this book assumes such a process exists and asks what each gate should decide.

Figure 2The SB36 stages, budgets and gates

	Budget	What it buys	Gate after it
Stage 1, scoping	A$40k	market and technical scoping	Gate 2, 15 April 2025
Stage 2, concept	A$180k	camera and nozzle-control concept; bench detection trial	Gate 3, 4 August 2025
Stage 3, development	A$560k	two prototype booms	Gate 4, planned 26 November 2025
Stage 4, field validation	A$320k	summer fallow season on 6 growers' farms, 9,000 ha: technical A$220k, market programme A$100k	Gate 5, planned 15 April 2026
Stage 5, launch	A$1,800k	tooling and fixtures A$600k; initial build inventory A$700k; dealer training and demonstration units A$300k; compliance, manuals and launch marketing A$200k	none: in production

The money rises steeply through the stages. The first two stages together cost A$220,000. Stage 3 alone costs A$560,000 and the launch A$1,800,000, more than all the development stages combined. The plan is built so that each expensive stage is approved only after a cheaper one has answered a question: the bench trial answers whether the cameras can detect weeds at all, the prototypes whether a full boom can do it reliably, and the field season whether it works on growers’ farms and whether growers and dealers will pay for it.

Stage gates that make real decisions: go, hold, recycle or stop describes how such a process is set up: deliverables, criteria, gatekeepers and the four decisions a gate can make. The Project Gate Review tool scores a project at a gate on value, fit, urgency, risk and readiness. Neither tells the gate what the next stage is worth in dollars; that is the job of Chapter 2.

Figure 3SB36 gates and the dates that matter for the tooling

1. 11 Feb 2025Gate 1: goprogramme admitted to Stage 1, scoping
2. 15 Apr 2025Gate 2: goStage 2 approved, A$180k
3. 4 Aug 2025Gate 3: goStage 3 approved, A$560k
4. 14 Aug 2025Tooling order placedA$600k, ten days after Gate 3
5. 31 Aug 2025Commitment registercommitted but unspent A$1,065k against a Stage 3 balance of A$465k
6. 19 Sep 2025Order cancelledA$40k slot reservation instead of the A$180k fee
7. 9 Oct 2025Steel cut, avoidedcancellation fee would have become 100%
8. 26 Nov 2025Gate 4, plannedStage 4 field validation, A$320k
9. 15 Apr 2026Gate 5, plannedlaunch decision, A$1,800k
10. 2 Sep 2026Tooling ready if ordered at Gate 520-week lead time
11. 1 Nov 2026First production deliveries neededfor the 2026–27 season

Keeping a commitment register

Corella’s finance manager keeps the commitment register monthly, and it is a simple spreadsheet. Each row is one order or undertaking: supplier, what it is for, the stage and budget line it is charged to, the amount, the amount invoiced so far, the cancellation terms, and the date the terms next change. The summary at the top gives three totals for the programme: spent, committed but unspent, and uncommitted, which together always equal the approved plan. A fourth number, cancellable now, is the part of the committed money that could be recovered if the business stopped this week.

Two checks make the register useful at a gate rather than only as a record. First, every commitment must sit within a stage that a gate has already approved; a commitment charged to a later stage is a red flag whatever its size. Second, the register lists the coming cliff dates, the days on which cancellation terms get worse, so that the gatekeepers can see what waiting a month would cost before they decide whether a decision can wait.

The error: a tooling order ten days after Gate 3

This is an illustration. Gate 3, held on 4 August 2025, approved Stage 3: A$560,000 to build two prototype booms. The minutes recorded “SB36 programme approved to proceed” and attached the full five-stage plan, including the A$1,800,000 launch budget, for information.

On 14 August 2025 the engineering manager placed a A$600,000 order for the SB36’s tooling and fixtures: the boom-section jigs, frame fixtures and press tools. The reasoning was the lead time. The toolmaker quoted 20 weeks, and the engineering manager wanted the tooling in hand well before the first production builds. The order was well above the engineering manager’s A$75,000 delegated limit, but the purchasing system allowed an order against any budget line in an approved plan, and the minutes said the programme was approved. The toolmaker’s terms were common ones: cancellation before steel was cut cost 30%, A$180,000; after steel was cut, scheduled for week 8 (9 October 2025), it cost 100%.

Seen from Gate 3, the chance that the SB36 would not reach a full launch was 0.6794. Chapter 2 derives it: a 0.75 chance that the prototypes pass at Gate 4, multiplied by a 0.4275 chance that the field season gives a strong market signal, makes 0.3206 for a launch, and 1 − 0.3206 = 0.6794 for no launch. If the order had stood past steel-cut, the expected write-off would have been 0.6794 × A$600,000 = A$407.6k, before any value the tooling might have had for another product.

Figure 4The end-August commitment register, before and after the fix

	Spent	Committed but unspent	Uncommitted
As reported, 31 August	A$315k	A$1,065k	A$1,520k
With the 19 September fix	A$355k	A$465k	A$2,080k

The order was found by the finance manager while preparing the end-August register. Spent to date was A$315,000: Stage 1 A$40,000, Stage 2 A$180,000 and Stage 3 A$95,000 so far. Committed but unspent was A$1,065,000, made up of the A$465,000 balance of Stage 3 and the A$600,000 tooling order. The first check failed at once: A$600,000 was committed against Stage 5, which no gate had approved.

The finance manager then ran the lead-time check. If the tooling were ordered at the planned Gate 5 date, 15 April 2026, a 20-week lead time would deliver it on 2 September 2026, 60 days before the first production deliveries were needed on 1 November 2026. Ordered on 14 August 2025, it would have arrived about 1 January 2026 and waited ten months for a launch decision that had not been made. The early order bought no time that the plan needed. All it did was move A$600,000 from uncommitted to committed, and, after 9 October, from cancellable to lost.

The managing director met the engineering manager and the toolmaker on 19 September 2025, three weeks before steel was due to be cut. The order was cancelled. The toolmaker agreed to waive the A$180,000 cancellation fee in exchange for a non-refundable A$40,000 reservation of a design-and-production slot and first refusal on the tooling job if the launch was approved. The episode cost A$40,000 instead of an exposure of A$600,000, and the reservation kept most of what the engineering manager had wanted: a toolmaker ready to start promptly after Gate 5.

Corella made three changes. Gate minutes now state the amount and the stage approved (“Stage 3 approved, A$560,000; no Stage 5 commitments”), not “programme approved”. The purchasing system accepts orders only against the current stage. And any long-lead order for a later stage comes to the gatekeepers with its lead-time check and cancellation terms, so that they can buy a cheap option, such as a slot reservation, rather than the full commitment. How the tooling contract’s terms divide the cost of a change between buyer and supplier is the subject of How contracts shape behaviour: who keeps a dollar saved in Project Teams and Contracts by the Numbers.

What a gate approves

The lesson of the tooling order is not that long-lead items should never be ordered early. Sometimes the lead-time check shows that waiting for the gate would delay the launch, and then the question becomes a calculation: the expected cost of the commitment if the programme stops, against the expected cost of the delay if it goes ahead. The point is that the calculation belongs at a gate, with the probabilities and the cancellation terms on the table, rather than in a purchase order raised under a broad reading of the minutes.

Figure 5Before committing money for a later stage

• Do: The commitment sits within a stage a gate has approvedotherwise it is a gate decision, whatever its size
• Do: Lead-time check done from the planned gate datedoes ordering at the gate still meet the need date
• Do: Cancellation terms and cliff dates recordedwhat the order costs to undo, and when that rises
• Do: Expected cost if the programme stops is statedprobability of no launch × amount not recoverable
• Do: Cheaper ways to hold the option are comparedslot reservation, design-only order, price held for a fee
• Avoid: Approval read from "programme approved" minutesthe minutes must state the stage and the amount

Gate 3 approved A$560,000 because, as Chapter 2 shows, the programme was worth about A$323,100 in expected terms with the gates in place. Chapter 3 shows that the same programme approved all at once, with the launch commitments made at Gate 3, would be worth less than nothing. The tooling order was a step towards that all-at-once programme, taken without anyone deciding to take it. Locking in the plan: long-lead orders, readiness and change after approval in Front-End Planning by the Numbers covers long-lead orders at the point where commitment is intended, after the final approval.

Exercise: camera modules ordered at Gate 4

At Gate 4, held on 17 December 2025, the prototypes pass their technical tests and Stage 4, the field season, is approved. Gate 5 is scheduled for 21 April 2026. The camera supplier proposes that Corella order the camera modules for the first production run, A$240,000, now, with a 50% cancellation fee, and offers a 5% discount for the early order. The supplier’s lead time is 12 weeks and production starts on 1 August 2026. At Gate 4 the technical question has been answered, so the chance of no full launch is the chance that the field season gives a weak market signal, 0.5725. Should Corella place the order at Gate 4?

Worked answer. If Corella orders now and the field signal is weak, it cancels and pays 50% of A$240,000. The expected cancellation cost is 0.5725 × 0.5 × A$240,000 = A$68.7k. The early-order discount is worth 0.05 × A$240,000 = A$12,000, well below the expected cancellation cost. The lead-time check settles the rest: ordered at Gate 5 on 21 April 2026, the modules arrive 12 weeks later, on 14 July 2026, before production starts on 1 August 2026. Ordering at Gate 4 saves no time and costs about A$56,700 in expected terms net of the discount, so the gatekeepers should reject it. If the supplier’s lead time were longer than the 14 weeks and 4 days between Gate 5 and production start, the comparison would change and the gate should then price a slot reservation, as Corella did for the tooling.