When to Stop Spending on Large Capital Projects

Across the world, large capital projects accumulate commitments long before they produce benefits. Foundations are poured, equipment is ordered, contracts are signed and public promises acquire an audience. When the case weakens, the next funding decision is rarely made in an empty room. It arrives carrying years of effort, organisational credibility and expenditure that nobody wants to describe as wasted. The question of what should happen next can become an argument about whether the original decision deserves defending.

That is a dangerous substitution. Past investment may have created useful assets, new knowledge and unavoidable obligations. Those consequences matter. But money that cannot be recovered does not become a reason to spend more simply because the amount is large. Equally, a cost overrun does not automatically establish that stopping is the best remaining choice. A troubled investment can still have a credible continuation case; a nearly completed asset can still fail to justify its next commitment.

This comprehensive blog will explore why sunk costs distort continuation decisions on large capital projects, how lock-in becomes escalation, what a genuine point of no return means, and how leaders distinguish justified persistence from further avoidable loss. The central discipline is forward-looking judgement: assess the benefits, costs, risks and obligations still ahead without asking past expenditure to vote for the next spend.

1. Four Ideas That Must Not Be Confused

Sunk costs are money, time and effort that cannot be recovered. The sunk-cost fallacy occurs when those past investments are used to justify additional commitment. The distinction is simple but consequential. Expenditure being sunk is a fact about recoverability; allowing its historical amount to determine the next decision is a reasoning error.

Escalation of commitment is a broader continuation pattern. Resources keep being committed despite adverse evidence, with cumulative prior investment used to defend the course. It can involve individuals, groups or whole organisations. Each additional payment can make the next request feel harder to refuse, even when the original purpose is becoming less attainable.

Lock-in describes the narrowing of actual or perceived alternatives. Premature commitment can create it: a route, technology or promise is fixed before alternatives have been adequately examined. Later escalation can reinforce it. These are related stages, not interchangeable labels. The mistake of committing too early is different from the mistake of continuing for the wrong reasons after circumstances have changed.

The financial distinction also needs care. A partly completed asset may have salvage value or support another use. Work already performed may have reduced the cost of completing a valuable service. A contract may leave future obligations whether the investment continues or stops. None of those consequences should be dismissed by saying that everything already spent is irrelevant.

The historical payment is different from the options it has created. A usable asset belongs in today's assessment because of what can still be done with it, not because the organisation paid heavily for it. An exit obligation belongs in the assessment because it will affect future consequences, not because it restores the original business case.

On large capital projects, this distinction prevents two opposite errors. One is continuing because too much has been spent. The other is recommending cancellation while ignoring real costs of closure or benefits still available from completion. Both avoid the difficult comparison. The right question is what each feasible course now requires and provides, not whether yesterday's expenditure feels too painful to write off.

Figure 1: Unrecoverable past expenditure does not justify the next commitment. Remaining benefits, expenditure, usable assets and future obligations belong in the comparison of today's feasible choices. The historical price of an asset is not the same as its present usefulness.

A useful practical check is to separate the statement “this has cost a great deal” from the statement “this still has value.” The first describes history. The second requires evidence about future use. An expensive item can retain little value; an item acquired cheaply can remain essential. The comparison should also avoid counting a future obligation twice, once as an argument for completion and again as a separate penalty for stopping. The particular appraisal method requires professional judgement, but the reasoning boundary is straightforward: compare the consequences of the available decisions, not the emotional weight of their histories.

2. How Early Commitment Becomes Later Escalation

The path into escalation can begin before a credible delivery promise exists. A preferred solution gains momentum, a public date is announced, or a funding approval is secured after shallow analysis. Alternatives that were once available become politically awkward or organisationally invisible. Commitment starts to be treated as evidence that the chosen course must be right.

Delivery exposes the unresolved assumptions. A difficult interface requires a patch. The patch changes another sequence. Additional spending is justified as the amount needed to protect the investment already made. A pattern of problems, fixes and further problems develops, while each corrective package makes the original direction harder to question.

The economic and political stories can then diverge. A continuation case may be weakening while the case for defending the original approval becomes stronger. More people have endorsed it, more contracts have been placed and more public explanations depend on eventual success. The organisation can become committed not just to the intended benefit but to demonstrating that its earlier commitment was justified.

This sequence is not inevitable. Nor does every request for additional money represent escalation. Uncertainty is real, and a justified recovery package can require more resources than originally expected. The warning is a change in the basis of reasoning: additional commitment is defended mainly through the amount already committed rather than the value still achievable.

The distinction between early and late errors matters. A poorly examined initial decision can be followed by a rational choice to complete because completion is now the best feasible option. A well-examined initial decision can later require stopping because new conditions have undermined its remaining value. Present judgement must not be reduced to a verdict on the competence of the original team.

For large capital projects, the practical danger is a sequence in which every new decision is described as merely protecting the previous one. Each package looks smaller than the accumulated loss it promises to rescue. The cumulative result is a commitment that has never received a fresh, forward-looking justification. Breaking that sequence requires a review of the current alternatives, not another attempt to make the original approval look unchanged.

There may also be a gradual shift in the object being defended. The organisation starts with a service or strategic result, then begins defending the construction programme, and finally defends completion of whatever scope can still be afforded. That change can be legitimate if a reduced outcome has a credible case. It becomes dangerous when a smaller benefit inherits the justification of a larger one without being assessed afresh. Protecting the name of the investment is not the same as protecting its purpose. A revised course needs its own forward explanation, not borrowed credibility from an outcome that is no longer available.

Figure 2: Premature commitment can lead through adverse evidence, defensive patches and additional investment to greater attachment and pressure to continue. A deliberate reassessment provides an exit from the loop. This is a synthesized pathway, not an inevitable sequence for large capital projects.

3. Why the Room Keeps Saying “Just a Little More”

Past investment is difficult to separate from the people who made it possible. Teams have committed energy, careers and relationships as well as money. Sponsors have explained the investment to boards, partners or public audiences. A recommendation to stop can be heard as a judgement that all of that effort was foolish, even when the actual question concerns only the next commitment.

That emotional weight helps explain why continuation can feel responsible while closure feels wasteful. The argument “after all this effort” recognises a real human investment but supplies no evidence about the remaining benefits. The phrase “almost there” may describe physical completion while saying nothing about whether the final expenditure will produce a useful operating asset.

Political pressure adds another layer. Audiences may interpret cancellation as throwing away money already spent. Decision-makers can therefore face incentives to continue even when they understand the forward economic argument. Defending another package can appear safer than explaining why the historical expenditure is unrecoverable whichever course is chosen.

Prior optimistic expectations can also make the recovery narrative attractive. The original outcome remains familiar; the downside alternative requires a new and uncomfortable explanation. A pack that describes another increment of funding as the final bridge to success can preserve the desired story without demonstrating that the bridge reaches the benefit originally promised.

These are diagnostic possibilities, not grounds for assuming bad faith. Phrases such as “just a little more,” “stopping admits failure” or “too much has been spent” are illustrative warning signs. They do not prove that the current recommendation is wrong. They indicate that the room should ask for the forward case rather than accept the emotional force of the statement as the case itself.

Genuine recovery evidence is specific. It identifies the remaining work, dependencies, resources and benefits, and explains why those expectations are credible now. Reassurance repeats the desired outcome or the team's determination without resolving the reasons the outcome is in doubt. The difference is particularly important when a series of supposedly final recovery packages has already failed to settle the problem.

The room also needs a way to acknowledge effort without making it an investment criterion. People can have worked competently and honourably on a course that is no longer justified. Respect for their contribution does not require further spending. Protecting that distinction makes a truthful decision easier than forcing everyone to choose between defending colleagues and examining the remaining case.

A useful challenge is to ask what the recommendation would be if the same assets, obligations and remaining choices had arrived under different leadership. The question does not erase history or assume that a replacement team would be unbiased. It tests whether the recommendation depends on personal ownership of the earlier promise. If the answer changes mainly because different people would be explaining the loss, the review needs a clearer separation between investment judgement and accountability for past decisions. Both deserve attention, but they should not determine one another without an explicit reason.

4. Is There Really a Point of No Return?

The phrase “point of no return” can refer to very different conditions. Sometimes it means that political embarrassment makes retreat unattractive. Sometimes it means that a specific physical investment cannot be recovered or that another concept can no longer be adopted without major cost. Treating those meanings as identical obscures the actual decision.

Large capital projects can contain genuinely irreversible commitments. A tunnel cannot simply be returned to its pre-construction state. Equipment designed for one purpose may have limited value elsewhere. An asset-specific investment may not be saleable for anything close to its original cost. Those are real constraints, not merely cognitive errors.

But irreversibility of the past investment does not establish the value of every future payment. The fact that yesterday cannot be restored leaves a decision about tomorrow. The feasible choice may no longer be the one available before authorization, yet it can still be a choice between completing, changing scope, preserving work temporarily or closing it safely.

Not every branch remains available. A reduced scope may fail to provide any useful service. A pause may create expensive preservation obligations or lose a critical operating opportunity. Closure may involve contractual liabilities and continuing responsibilities. Completion may require further capital and expose the organisation to operating losses. The comparison must reflect those conditions rather than assume a costless escape.

A credible review therefore asks what “no return” means in practical terms. Is an earlier alternative physically impossible, economically unattractive or simply politically uncomfortable? Which consequences follow under every option, and which can still be changed? What advice is needed on safety, contractual commitments, environmental responsibilities and other project-specific obligations?

Those questions do not supply universal legal answers. They prevent a slogan from substituting for appraisal. Some obligations must be honoured regardless of whether discretionary investment stops. Some assets can still support value that would be lost through abrupt closure. Both need visibility.

The resulting judgement can be uncomfortable: the original opportunity has gone, the past loss cannot be undone, and none of the remaining choices is attractive. Even then, a choice remains between their future consequences. The least damaging feasible course is not made better by pretending that further expenditure will recover money that is already beyond recovery.

Figure 3: Irreversible investment can prevent a return to the original starting point without removing every forward choice. Assess only feasible routes, including remaining benefits, costs and obligations. Recovery of all past expenditure is not an available branch merely because the organisation wants it to be.

5. Rebuild the Continuation Case from Today

The continuation question should start with the purpose still achievable, not with the amount already spent. The review needs a current account of the benefits, the conditions required to realise them and the time over which they remain relevant. A delivery plan that reaches completion is not sufficient if the intended service, commercial return or strategic outcome has become implausible.

Next comes the remaining commitment. That includes expenditure still required, delivery uncertainty, dependencies and operating requirements beyond physical completion. The original estimate provides a record of the decision made then; it is not automatically a reliable forecast of what remains. A revised case should explain what has changed rather than quietly replace the old assumptions.

Feasible alternatives then need a consistent comparison. Completion, a redesigned or smaller service, a justified pause and closure should be assessed only where they can actually be carried out. Future obligations, preservation expenditure, salvage or reuse opportunities and consequences for benefits must be considered on each relevant path. The comparison concerns differences that decisions can still affect.

Consider a fictional review of an industrial facility. Most construction expenditure has already been incurred. A recovery package promises completion, but a critical operating dependency remains uncertain. The initial argument for funding is that stopping now would waste everything already built. That argument identifies the painful history but does not demonstrate the package's value.

A forward review asks whether the proposed additional work can produce a useful facility, what further operating commitments are required and whether the dependency is credible. It compares that course with feasible modification, preservation or safe closure, including the costs and residual value of each. The work already completed affects those options through its present usefulness. Its historical price does not count again as a benefit of completion.

The result could favour proceeding. If the remaining benefits credibly exceed the remaining commitments and risks relative to alternatives, continuation may be justified even though the original decision was poor. The result could favour stopping. Physical completion may be close while the operating outcome remains unattainable. A completion percentage cannot settle the decision in either direction.

This comparison is a practical synthesis, not a prescribed financial model or validated termination formula. Detailed appraisal still requires appropriate expertise and evidence. Its purpose is to keep the decision anchored in the future rather than in an emotional contest over the past.

Uncertainty should remain explicit. The review should identify assumptions that matter, evidence that would change the recommendation and the consequences of being wrong. An attractive central scenario should not conceal a dependency that could remove most of the remaining benefit. Equally, uncertainty alone is not proof that waiting or stopping has greater value. The alternatives also carry uncertainty, cost and consequences that require examination.

6. Make Review Conditions Harder to Rewrite

Recognition of sunk-cost reasoning is useful, but recognition alone does not ensure that a difficult continuation decision will be made well. Large capital projects need review arrangements that make the current case visible before another material commitment becomes effectively irreversible. The following safeguards are practical applications of the combined evidence, not a proven anti-escalation protocol.

First, establish material review conditions and clear decision ownership. A condition might concern whether a necessary benefit remains achievable, whether a critical dependency has been resolved or whether a recovery commitment is still supported by evidence. The conditions should reflect the investment rather than a universal overrun percentage. Their purpose is to identify when a decision must be reconsidered.

A breached condition should prompt judgement, not automatic cancellation. Otherwise the organisation may hide the breach to protect a course that is still defensible, or stop mechanically without considering the consequences. The review should explain what the breach means and whether continuing, changing, pausing or stopping is now justified.

Second, bring challenge that does not depend entirely on defending the original approval. Delivery leaders have an important account of what can be achieved, but investment review should not rely solely on advocacy for the existing plan. Challenge should examine the assumptions and alternatives without turning into an accusation that the team lacks commitment.

Third, preserve the decision record. The original case, subsequent changes and current forecast serve different purposes. Replacing them with a succession of cleaner baselines can make deterioration difficult to recognise. A useful record shows why previous expectations changed, what resources were then committed and what evidence now supports the recommendation.

Fourth, make the outcome of review explicit. A request for another recovery package should not pass as an investment decision by default. The record should name the chosen course, its material assumptions, who is accountable for the decision and the evidence checkpoint that follows. If conditions are attached, the organisation needs to understand what happens when they are not met.

These safeguards do not guarantee independence, eliminate politics or make all alternatives comparable with precision. They make reasoning inspectable. A sponsor can then explain why more spending is justified without relying on the amount already spent, or why expenditure should stop without pretending that exit has no cost.

The discipline is especially valuable where a succession of small commitments can evade a larger decision. Each increment may look manageable while the combined exposure changes substantially. A current, explicit view of the remaining commitment helps prevent the organisation from crossing another practical boundary without acknowledging that a choice has been made.

A checkpoint should also have a defined scope. It may concern the next funding tranche, a design change or a decision that materially narrows future alternatives. Its scope should not be quietly expanded into approval for every subsequent commitment. Conversely, a review cannot assume that withholding one payment stops all exposure. Work already authorised may continue generating obligations. Decision-makers need visibility of that difference so that the record describes what is genuinely being controlled. This is an illustrative governance practice, not a universal rule about contract payments or a substitute for project-specific commercial advice.

Figure 4: A synthesized review aid asks whether benefits remain achievable, what commitments remain, which alternatives are feasible, what evidence would change the case and who owns the decision. The output is a recorded choice with conditions, not an automatic cancellation instruction.

7. Persistence Is Not the Enemy

Delivery requires determination. Large capital projects encounter ordinary difficulties that competent teams can overcome through disciplined action. Reopening every execution choice or constantly revising the odds of success can undermine focus without improving the investment decision. Commitment is not automatically a bias, and resilience should not be mistaken for denial.

The distinction lies in the basis for persistence. A credible recovery plan identifies achievable benefits, realistic remaining work and the conditions on which success depends. Escalating persistence relies on the need to protect past expenditure or vindicate the original decision when the current case no longer supports it. Both can use the language of confidence; only one supplies a forward justification.

Post-commit optimism can help people act under difficulty. It cannot establish that an additional investment is worthwhile. The functions should remain separate: teams need enough clarity and confidence to deliver an authorised course, while accountable decision-makers must examine material evidence that changes the case for that course.

Independent review should therefore be neither a campaign for cancellation nor a demand for permanent reassurance. It should test the current recommendation consistently, including the claims made for stopping. A closure proposal that ignores future liabilities or usable assets is no more disciplined than a continuation proposal defended through historical expenditure.

Respect for previous effort remains important. A decision to change course need not imply that the team acted carelessly, just as a decision to continue does not erase the mistakes that produced the current position. The organisation should be able to honour contribution, learn from earlier decisions and choose the best remaining course without forcing those tasks into one verdict.

The goal is not weaker commitment. It is commitment directed toward a useful purpose, supported by current evidence and bounded by responsibility for what will happen next.

The same distinction matters in communication with teams and stakeholders. A continuing recommendation should explain the achievable outcome and the evidence supporting it, not merely celebrate courage. A stopping recommendation should explain the remaining consequences and how responsibilities will be discharged, not merely announce that the past was a mistake. Clear reasoning cannot remove disappointment, but it can prevent disappointment from becoming the only basis for the next choice. Confidence in leadership is better served by an honest account of present options than by a succession of promises that depend on avoiding the central question.

The Past Cannot Vote for the Next Spend

The question of when to stop spending on large capital projects cannot be answered by the amount already spent, the embarrassment of retreat or a simple completion percentage. Nor can an overrun alone settle it. The relevant comparison concerns remaining benefits, expenditure, risks and obligations under the choices still available.

Premature lock-in can make that comparison harder. Emotional and political investment can turn another funding request into a defence of yesterday's judgement. Irreversible assets can remove attractive alternatives. None of these conditions permits the past to supply benefits that the future no longer offers.

The practical discipline is to make the next commitment explicit, examine feasible alternatives, preserve uncertainty and record who owns the decision. Continue when the current case justifies continuation. Change, pause or stop when another feasible course has the stronger forward justification. Some responsibilities will survive whichever option is chosen, and those responsibilities must remain visible.

Strong leadership combines determination with the ability to recognise when determination is serving the original promise rather than the intended purpose. The money already spent cannot be recovered through conviction. The resources not yet committed can still be directed by judgement.

References

• Flyvbjerg, B., & Gardner, D. (2023). How Big Things Get Done: The Surprising Factors That Determine the Fate of Every Project, from Home Renovations to Space Exploration and Everything in Between. New York: Currency.
• Project Management Institute. (2014). Navigating Complexity: A Practice Guide. Newtown Square, PA: PMI.
• Williams, T. M., Samset, K., & Sunnevåg, K. J. (Eds.). (2009). Making Essential Choices with Scant Information: Front-End Decision Making in Major Projects. Basingstoke: Palgrave Macmillan.
• Lovallo, D., & Kahneman, D. (2003). "Delusions of Success: How Optimism Undermines Executives' Decisions." Harvard Business Review, 81(7), 56–63.
• Project Management Institute. (2024). The Standard for Program Management (5th ed.). Newtown Square, PA: PMI.