Stage Gates and Stopping Rules by the Numbers
Chapter 9 of 10 · 10 min read
Pivot, stop or switch to the competing investment
Five Gate 5 options under a A$1.8m capital limit: full launch A$116.8k, waiting a season A$650.2k, and a pivot to the retrofit kit with the welding cell funded now, A$893.7k.
A gate that stops a course of action has not finished its work. Once the full product is ruled out, the money and people it would have used are free, and the gate has to decide what to do with them. The options are rarely just “launch” or “stop”. There may be a smaller product that uses what was learned, a reason to wait for more evidence, and, in most businesses, another project that has been waiting for the same capital. A stop decision taken without looking at those alternatives can be as costly as a launch taken without looking at the evidence.
This chapter sets out the option set that Corella Plains Implements faced at Gate 5 on 21 April 2026, values the retrofit kit, the robotic welding cell and a one-season wait, and combines them into five options under the business’s capital limit. It shows why the pivot won, what the field signal had to do with it, and how an early comparison nearly dropped the best option by charging it with the past.
The option set and the capital limit
This is an illustration. All money in this chapter is in Australian dollars excluding GST, in thousands (A$k). Values are expected net present values at 10% a year, taken at the decision date.
Corella’s board had set a limit of A$1,800k on new capital commitments in the 2026–27 financial year. The figure was not chosen with SB36 in mind: it was what the business could fund from cash and its existing facility without raising its gearing above the bank’s covenant. It happened to equal the full boom’s launch budget, which meant that a full launch would use the whole of the year’s capital.
The other claim on that capital was a robotic welding cell for the frames of Corella’s chaser bins. The cell had been proposed in 2025 and deferred while SB36 was in development. It was a conventional process investment, with far less uncertainty than a new product, and it had been ready to order for months.
Four product options were open at Gate 5, after the Weak field signal of Chapter 7 had put the market’s probabilities at 0.0655 High, 0.3843 Medium and 0.5502 Low:
- Launch the full boom anyway, against the kill criteria: expected NPV −A$261.2k.
- Pivot to a retrofit kit: cameras, controller and nozzle valves fitted to growers’ existing booms.
- Wait a season: run a second field season before deciding.
- Stop and sell the two prototypes as conventional booms: A$90k.
Each could be paired with the welding cell now, or with the cell deferred a year if the product option used too much of the A$1,800k.
The retrofit kit
This is an illustration. The kit sells at A$38k with a variable cost of A$24k, a contribution of A$14k a kit. Its development would cost A$140k, because it reuses about A$430k of SB36’s detection software, camera mounts and controller, and its launch A$420k, an investment of A$560k. Valued over six seasons at 10% (annuity factor 4.3553), its expected sales in each market state give:
| Kit units a year | Kit NPV | Full boom NPV | |
|---|---|---|---|
| High | 30 | 1,269.2 | 7,258.9 |
| Medium | 22 | 781.4 | 11.8 |
| Low | 12 | 171.7 | −1,347.1 |
| Expected, prior at Gate 3 | | 690.0 | 1,348.0 | |
| Expected, after Strong | | 974.0 | 3,502.9 | |
| Expected, after Weak | | 477.9 | −261.2 |
The kit’s present value of contribution is units × 14 × 4.3553: A$1,829.2k, A$1,341.4k and A$731.7k in the three states, less the A$560k investment. Its NPV is positive in every state. Its expected value is A$690.0k on the Gate 3 prior, A$974.0k after a Strong signal and A$477.9k after a Weak one.
That table explains two things. First, why the kit was not the original plan: on the prior at Gate 3, the full boom was worth A$1,348.0k against the kit’s A$690.0k, nearly twice as much, because the full boom’s High-market payoff was so large. Second, why the kit became the best option after a Weak signal: once the High market became unlikely, what mattered was the downside, and the kit has none. The kit’s value moves much less with the market state, so the evidence that sank the full boom only dented the kit. It was also, in part, a product of the full programme: without the A$430k of reusable work, its development would have cost far more, and the 14 grower enquiries recorded in the field season were the first sign of demand for it.
The welding cell
This is an illustration. The welding cell costs A$850k and saves A$260k a year in labour, rework and floor time for seven years. At 10%, the seven-year annuity factor is 4.8684, so the present value of its savings is 260 × 4.8684 = A$1,265.8k, its NPV A$415.8k and its profitability index (present value of savings divided by cost) 1,265.8 ÷ 850 = 1.49. The Equipment Investment Appraisal (NPV, IRR, payback) calculator reproduces calculations of this kind for any equipment purchase.
If the cell is deferred a year, the same cash flows start a year later and its NPV becomes 415.8 ÷ 1.1 = A$378.0k. Deferral costs A$37.8k of value. That is the price of letting the full boom take the whole capital limit, and it is what makes the welding cell more than a footnote: every option that defers it pays A$37.8k for doing so, and every option that funds it now gains A$415.8k.
Waiting a season
This is an illustration. The fourth product option was to run a second field season through the summer of 2026–27, at a cost of A$200k, and decide in April 2027. The second season would use the same criteria as the first, so its likelihoods of a Strong result are the same: 0.85, 0.45 and 0.10 in the High, Medium and Low states. As in Chapter 5, a year’s delay was assumed to let the imported competitor establish itself, so a late full boom would sell only 28 booms a year in the High state instead of 40, and its High-market NPV would fall to A$4,541.3k. The kit, sold to growers who already own booms, was assumed not to lose share by waiting.
- Wait a season−A$200k now; decide April 2027
- Strong again, p 0.2836posterior 0.196 / 0.610 / 0.194
- Late full launchA$637.3k
- Retrofit kitA$758.9k
- Weak again, p 0.7164posterior 0.014 / 0.295 / 0.691
- Late full launch−A$865.4k
- Retrofit kitA$366.6k
- Strong again, p 0.2836posterior 0.196 / 0.610 / 0.194
Starting from the Weak posterior, a second Strong result has a probability of 0.2836. After it, a late full launch is worth A$637.3k and the kit A$758.9k. After a second Weak result, the launch is worth −A$865.4k and the kit A$366.6k. The kit wins on both branches. The expected value in April 2027 is 0.2836 × 758.9 + 0.7164 × 366.6 = A$477.9k, exactly the kit’s value today, and the option’s value at Gate 5 is −200 + 477.9 ÷ 1.1 = A$234.4k.
The coincidence is not an accident. When a test cannot change the decision that follows it, the expected value of the decision after the test equals the value of making it now; the test only adds its cost and its delay. In the language of Chapter 4, the second season has an expected value of sample information of zero, because the loss of High-market share has made a late full launch worse than the kit even after a Strong result. Waiting here is passive delay dressed as learning. It would cost A$200k and a year, and buy nothing.
Five options under the limit
This is an illustration. Combining the product options with the welding cell gives five options worth comparing, each checked against the A$1,800k limit:
| New capital, A$k | Within A$1,800k | Expected NPV, A$k | Made up of | |
|---|---|---|---|---|
| A Full launch; welding cell deferred a year | 1,800 | yes | 116.8 | −261.2 + 378.0 |
| B Retrofit kit; welding cell now | 1,410 | yes | 893.7 | 477.9 + 415.8 |
| C Stop, sell prototypes; welding cell now | 850 | yes | 505.8 | 90.0 + 415.8 |
| D Wait a season; welding cell now | 1,050 | yes | 650.2 | 234.4 + 415.8 |
| E Full launch and welding cell now | 2,650 | no | 154.6 | −261.2 + 415.8 |
Option B, the retrofit kit with the welding cell funded now, is worth A$893.7k, A$243.4k more than waiting a season and A$387.9k more than stopping. It uses A$1,410k of the A$1,800k limit, leaving A$390k uncommitted for the rest of the year. Option A, the full launch, is the worst of the options within the limit: it uses the whole limit, defers the cell, and combines a negative product value with a reduced cell value. Option E shows that the capital limit is not what rules out the full launch. Even with unlimited capital, launching the boom and the cell together would be worth A$154.6k, far below B.
The decision is a pivot: the business keeps the part of the programme that the evidence supports, the detection system, and drops the part it does not, the full boom. It is also a switch: the capital the full boom would have taken goes to the welding cell, which has been waiting for it. Saying no to good projects: opportunity cost, balance and choosing what to fund describes opportunity cost across a portfolio. Here it is a single number: a full launch would have cost the business the A$415.8k cell now, or A$37.8k of its value by deferral, on top of its own expected loss.
The comparison only works because the alternatives were still alive at Gate 5. The kit existed as an option because the detection system had been designed as a module rather than as part of one boom, and the cell because its quotation and layout had been kept current while it waited. Options derived from the outcome, and why the choice changed in Outcome-First Projects by the Numbers shows how options built from the outcome a business needs, rather than from the first solution proposed, can change a choice; Stages and gates: organising the front end in Front-End Planning by the Numbers shows how alternatives are compared and documented at an early gate so that they can still be picked up later.
The signal decided it
It is worth asking what would have happened with the other field result, because it shows what the gate was really deciding. The exercise at the end of this chapter works the same table under a Strong signal. The answer, briefly, is that the full launch then wins by a wide margin and the welding cell is deferred. The kit, the cell and the capital limit are the same in both worlds; only the field season’s result differs. That is the sense in which Chapter 4’s A$100k market programme bought the decision: a single result, written as a criterion in December 2025, separated two quite different uses of the business’s capital.
The error: the kit charged with the past
This is an illustration. The first comparison prepared for Gate 5, a week before the meeting, was laid out as a ranking of projects competing for the 2026–27 capital: each project’s NPV, with the rule that only projects with a positive NPV would be funded. In it, the welding cell appeared at its NPV of A$415.8k. The retrofit kit appeared at −A$767.1k: its expected NPV of A$477.9k less the A$1,245k already spent on SB36, on the reasoning that the kit was “the SB36 programme continued” and should carry its costs. On that table the kit failed the positive-NPV rule, and the draft recommended stopping SB36 altogether and funding the welding cell: option C.
The independent board adviser questioned the kit’s figure by asking what each option would cost from today. The kit needed A$560k of new money; the A$1,245k had been spent whether or not the kit went ahead, and was the same under every option, including stopping, so it could not be charged to one of them. With the kit at its forward value of A$477.9k, it passed the positive-NPV rule, and option B, the kit with the cell, became the best use of the capital. The same mistake as Chapter 8’s, made in the opposite direction, had nearly stopped the best option instead of launching the worst. Charged with the past, the kit looked worse than doing nothing with SB36 at all; on the forward basis, dropping it would have given away A$387.9k of expected value, the difference between options B and C.
The finance manager added a rule to the capital ranking: every project is entered at its value from today, and spend to date appears in a separate column that does not enter the total. A project that has spent heavily and one that has spent nothing compete on the same terms.
Exercise: the other signal
Suppose the field season had returned a Strong signal, so that the posterior probabilities were 0.4971 High, 0.4211 Medium and 0.0819 Low. The capital limit is still A$1,800k. The full boom’s expected NPV is A$3,502.9k and the kit’s A$974.0k; the welding cell is worth A$415.8k now or A$378.0k deferred a year, and the prototypes A$90k if sold. What are options A, B and C worth, and which should the gate choose?
Worked answer. Option A, the full launch with the cell deferred, is worth 3,502.9 + 378.0 = A$3,880.9k and uses the whole A$1,800k. Option B, the kit with the cell now, is worth 974.0 + 415.8 = A$1,389.8k. Option C, stopping with the cell now, is worth 90 + 415.8 = A$505.8k. The gate should launch the full boom and defer the welding cell: A beats B by A$2,491.1k. Deferring the cell costs A$37.8k, which is trivial against the difference. Even if the capital limit could be lifted to fund both at once, the gain would be only that A$37.8k (3,502.9 + 415.8 = A$3,918.7k), so the limit hardly matters. With a Strong signal the full launch is the best use of the capital; with a Weak signal it is the worst. The same option set, the same capital and the same people reach opposite decisions on one field result, which is the case for writing that result down as a criterion before the season starts.