Not Just Optimism: Why the First Numbers Are Wrong on Large Capital Projects

Around the world, large capital projects are authorised on numbers that look tidy in the pack and kinder than the asset later turns out to deserve. Costs sit low. Schedules sit short. Benefits sit high. Years later the same projects arrive late, over budget and short on the returns that unlocked the money. Postmortems then split into familiar camps. One camp says the people who built the case were too optimistic — sincere, talented and systematically wrong. The other says the case was sold: the numbers were shaped to win approval, not to describe the work. Both camps can point to real patterns. Neither camp, on its own, explains why early numbers on large capital projects go wrong so often, in the same direction, across industries and jurisdictions. This comprehensive blog will explore why the first numbers are wrong, what optimism bias and the planning fallacy explain, what strategic misrepresentation explains, why the either-or debate misleads, and what leaders should ask at the gate when both psychology and incentives are in play.

1. The Business Case Nobody Believes Later

Consider a familiar scene. A steering committee opens an approval pack for a major facility, a transit corridor, a processing plant or a hospital expansion. There is a single cost. There is a single date. There is a single return. Contingency looks thin but “manageable.” The champion walks the room through a clean most-likely story. Challenges are sparse. Detail looks thorough enough that questioning the pack feels like obstruction rather than diligence. Money moves.

Years later the asset is late. The cost has climbed. Benefits that were sold as certain have softened. Interfaces that were “under control” in the pack have become claims. The room that once celebrated the green case now argues about blame. Some voices say the planners were naïve. Others say the pack was always a sales document. The argument is energetic. It rarely improves the next gate, because the next gate is built on the same habits: a kind story, a competitive capital process, and weak early accountability.

The pattern is not confined to one sector or one country. Large capital projects of every kind — plants, terminals, rail systems, data centres, ports, public buildings — show the same directional story. Costs and time tend to rise relative to the case that unlocked funding. Benefits tend to fall short of what was promised. Occasional honest error does not explain a directional pattern that repeats. Something in how early numbers are produced, selected and locked in is systematically kind to the case and unkind to the asset.

That is the object of this blog. The question is not how to rebuild a full outside-view forecasting method, how to catalogue every decision bias in a room, or how to score front-end definition completeness. Those jobs matter and belong in their own treatments. The question here is sharper: when the first numbers are wrong, is the useful diagnosis optimism, misrepresentation, or both — and what should leaders do at the gate once they stop forcing a single villain?

2. The Pattern: Early Misinformation, Late Pain

The damage begins early. In the concept and early development phases of large capital projects, misinformation about costs, benefits and risks is pervasive. Accountability measures are weakest precisely when the numbers that will later be treated as commitments are softest. The pack that looks decisive is often the pack that has been least tested against what similar completed efforts actually did. By the time independent scrutiny is intense, the narrative is already locked and the political cost of reopening it is high.

The typical outcomes are familiar. Cost overruns. Schedule slips. Benefit shortfalls. When large sets of completed efforts are examined together, the share that land both cost and time as originally forecast is low, and the right-hand tail of bad outcomes is fat. The point for this essay is not to reprint those tables. The point is the direction. Errors are not random around a truthful mean. They lean the same way: the case that unlocked the money was kinder than reality.

That systematic lean is what makes the postmortem debate so sticky. If early numbers were merely noisy, one would expect as many pleasant surprises as unpleasant ones. Large capital projects do not behave that way at the portfolio level. “We took a rational risk that sometimes pays off” is therefore a weak default explanation for a pattern that keeps pointing one way. Something upstream is producing early misinformation — about cost, about time, about benefits, and about risk.

Three neighbouring questions must stay out of the way so this cut stays clean. How to forecast better with class data is a method question. How rooms amplify cognitive bias under pressure is a hygiene question. What happens after money and reputation are sunk — and how stop rules fail — is an escalation question. This essay owns a different cut: why the early numbers were already wrong, and whether the cause is best told as psychology, as incentives, or as both acting together.

Figure 1: Early numbers, late reality. The approval pack shows low cost, short schedule and high benefits; delivery shows higher cost, longer schedule and lower benefits. The arrow between them is early misinformation — systematic in direction, not occasional honest error.

3. The Optimism Story: Bias Without Villains

One powerful story says the people who built the case believed it. They were not villains. They were subject to a well-documented forecasting pathology: the planning fallacy. Under the planning fallacy, planners overestimate benefits, underestimate costs and time, spin success scenarios and overlook failure modes. The result looks like confidence. It is delusional optimism dressed as a business case. High rates of initiatives that miss budget, time or expected returns are then misread as rational risk-taking that will pay off in the long run. Often they are not. They are the downstream trace of flawed forecasting.

The cognitive sources are ordinary and stubborn. People exaggerate their relative talent — the above-average effect — and bring that habit into project rooms where almost every champion believes their team is better than the average team that delivered the last similar asset. They take credit for positive outcomes and attribute negatives to external factors, so past luck is coded as skill and past failure is coded as weather. They overstate personal influence relative to chance — an illusion of control that tempts executives to treat risk as a skill challenge and to downplay uncontrollable events in the forecast. Imagination cannot foresee every failure mode. Treating a single “most likely” scenario as the probable outcome understates the combined weight of many small-chance unfavourable paths that together are not small at all.

Amplifiers make the bias worse on large capital projects. Analysis anchors on the champion’s optimistic preliminary plan; subsequent work adjusts around that plan rather than replacing it. Contingencies that look prudent inside the anchored story are often still insufficient once the asset meets reality. Competitor and overcapacity scenarios are neglected because the room is focused on its own capabilities and its own schedule. Organisational competition for scarce capital rewards rosy forecasts and selects the most overoptimistic projects into the portfolio. Stretch goals and compensation targets further skew the pack. Pessimism is often treated as disloyalty, so the group reinforces an unrealistic view even when individuals privately doubt it.

None of this requires intentional deceit. People can be sincere and still wrong in a predictable direction. “They believed it” does not clear the forecast. Senior expectations that systematically understate cost and duration and overstate benefits put delivery teams at a disadvantage before detailed planning even begins. The optimism story is therefore not a soft excuse. It is a hard diagnosis of how honest rooms produce dishonest-looking numbers.

The remedy implied by the optimism story is not a culture that roots out all optimism. Optimism generates enthusiasm, resilience and motivation. Aggressive goals can still motivate after commitment. What must stay realistic are the forecasts that decide whether to commit. Decision and support roles need realism; an optimistic controller is a different problem from an optimistic delivery lead. Action roles often benefit from optimism once the go decision is made. Confusing those jobs is how stretch ambitions get smuggled into the authorization case and then defended as “leadership.”

Uniqueness bias often travels with the optimism package: the claim that “this project is different” so class experience does not apply. A short beat is enough here. The uniqueness claim is one of the ways an inside view protects itself from correction. Depth on uniqueness and on how to answer it with a named class belongs with the outside-view method, not with this debate. What matters for delusion versus deception is simpler: uniqueness is a favourite shelter for both sincere optimism and strategic packaging, because it disables the comparison that would expose either.

4. The Misrepresentation Story: Incentives Without Naïveté

A second powerful story says belief is beside the point. Strategic misrepresentation is the deliberate or systematic distortion of costs, benefits or risks to win approval or contracts. It is a political-economic story, not a personality lecture. Incentives to present an overly favourable view are strongest early, when accountability measures are weakest and when the prize — funding, political credit, a career milestone, a firm’s backlog — is still in play. The same early window that makes optimism bias hardest to catch is the window in which distortion pays the most.

Power drivers sit beside psychology. Political lock-in rewards getting the project through the gate before opposition organises. Career and firm incentives reward the pack that looks greenest under competitive capital allocation. Consultants and contractors who know that only the optimistic case will be funded learn what the process selects. Once a favourable narrative is locked, later evidence that would reopen the case is costly to the people who sold the original story. Premature commitment and distorted early numbers reinforce each other: the misrepresentation helps lock the decision; the lock then protects the misrepresentation.

The practical point is blunt. Not every overrun on a large capital project is honest bias. Approval systems that reward the most optimistic pack select for spin as well as for delusion. A governance response that only trains people to “be less optimistic” will fail when the process still pays for the greenest case. Hunting for liars alone will also fail, because sincere planning fallacy is often enough to sink the numbers. But pretending that incentives never distort the pack is naïve — and it leaves the incentive design untouched while the next champion prepares the next kind story.

Misrepresentation must also be kept distinct from ordinary commercial shaping. Negotiating how value and risk are allocated among parties is not the same as distorting the forecast that justifies the asset. Parties can share value honestly under a realistic case. They can also negotiate hard under a case that was never true. This essay is about forecast distortion at the gate, not about contracting strategy as such.

Figure 2: Two stories of the same wrong number. The optimism path runs from a sincere inside view through the planning fallacy to a rosy case people believe. The misrepresentation path runs from approval incentives through favourable presentation to spin that wins the gate. Both paths can produce the same wrong number.

5. Both, Not Either/Or

The debate that wastes time is the demand for a single diagnosis. Were they stupid, or were they lying? On large capital projects the useful answer is usually: stop forcing the choice.

The same early misinformation can be produced by bias, by incentives, or by both in sequence. A room begins with a sincere inside view and a planning-fallacy forecast. Competitive packing then selects the rosiest version of that forecast among rivals for capital. Political lock-in freezes the narrative. By the time the asset is late, participants can honestly disagree about whether the original belief was sincere — because sincerity and selection both operated, and because memory itself is kind to the story that once won.

Organisational pressure already blurs the line inside the optimism package. Systems that reward rosy forecasts select overoptimism whether or not individuals “intend” to deceive. A person who shades a contingency downward to keep the pack competitive may feel loyal rather than dishonest. A team that drops a downside scenario to avoid being labelled negative may feel professional rather than manipulative. The governance problem does not wait for a courtroom standard of intent. It waits for a process that stops selecting the kindest story and starts demanding numbers that can survive contact with a class of completed work.

The practical implication is symmetrical. A response that only trains people to be less optimistic fails when incentives still pay for the greenest pack. A response that only hunts liars fails when sincere planning fallacy is enough to produce the overrun. Framing the drivers as universal — psychology and power, at all scales — is more useful than a personality contest about the champion. The champion is often both a believer and a player in a game that rewards belief.

Diagnose both. Fix both. That is the climax this essay owns. Method depth for class-based forecasting, multi-bias room hygiene, front-end definition scorecards and post-commit stop rules remain neighbouring jobs. What belongs here is the refusal to pick one villain when the gate is fed by both — and the recognition that either/or thinking produces either/or fixes that leave half the problem untouched.

A useful test in a real gate is to ask which fix the room is reaching for first. If the only proposed fix is training or a pep talk about realism, incentives are probably still untouched. If the only proposed fix is finding who “lied,” the planning fallacy is probably still untouched. Healthy gates can name both: how the forecast was built, and how the process selects among forecasts. Rooms that can only hold one of those sentences will keep shipping kind numbers.

Figure 3: Both drivers, one gate. Psychology and power both feed the approval gate on early numbers; the outbound path is overrun and benefit shortfall. Diagnose both; fix both. Either/or is the distraction.

6. What Does Not Settle the Debate Alone

Several familiar counters help — and none of them, alone, settles the delusion-versus-deception debate.

Taking an outside view — disciplining the forecast with what similar completed efforts actually did — makes both sincere optimism and opportunistic spin harder. A class distribution is more difficult to wish away than a bottom-up story built only from the champion’s plan. Independent appraisal that must show where the specific case sits in a named class raises the cost of both delusion and deception. That is the logic that belongs in this essay. The full procedure for naming a reference class, building a distribution and positioning the specific project is a method spine owned elsewhere and must not be rebuilt here as a substitute for facing the dual diagnosis.

Optimism in its place remains essential. Aggressive goals can motivate after commitment. Forecasts that decide whether to commit must stay realistic. Rooting out all optimism is the wrong campaign; it drains energy from delivery without fixing the gate. Separating goal-setting from capital-approval forecasts is the right one. When stretch ambitions are labelled as goals, they can energise execution. When they are smuggled into the authorization number, they become early misinformation with a motivational alibi — and they invite later “we were being ambitious” excuses for a case that was never a forecast.

Generic optimism-bias uplifts — adding a standard percentage when class data are thin — are a fallback, not a full answer. They can reduce the kindness of a pack when no better evidence exists. They do not replace named class work when class data exist, and they do not redesign the incentives that select the rosiest pack. Treating an uplift as a cure for misrepresentation is a category error: a fixed add-on does not change who wins when the greenest case gets funded.

Definition completeness before authorization reduces late change that comes from thin scope. That practice is complementary and valuable. It is not the same as curing delusion or deception in the benefit–cost story. A well-defined wrong case is still a wrong case. Scope clarity does not automatically produce honest benefits or realistic contingency.

Heuristics for deciding when information is still scant help leaders choose tools under uncertainty. They help a room stop pretending that three-decimal precision equals truth when the concept is still open. They do not explain why the thin numbers were already skewed in a favourable direction. This essay owns that skew — the directional kindness of early numbers — not the full toolkit for deciding under scarcity of fact.

Owner teams that validate authorization estimates independently — rather than accepting the champion’s bottom-up story as the only input — reduce the room for both bias and spin. That validation habit is adjacent to front-end discipline and to class-based forecasting; it is not a licence to merge industrial-project databases with other base-rate tables casually. The governance idea is what travels: someone who does not win when the pack is approved must be able to challenge the kindness of the numbers before lock-in.

7. What Leaders Should Ask at the Gate

Leaders who own the delusion-versus-deception debate do not need another forecasting manual at the moment of approval. They need a short set of questions that make both psychology and incentives visible before money moves. The questions below are diagnostic, not a substitute for class-based method work.

Direction check. Is the case optimistic in the usual direction — cost and time down, benefits up — relative to analogous completed efforts? If the pack is kinder than the class in every dimension, optimism or selection is already on the table. A pack that beats every analogue on cost, time and benefit without an extraordinary, tested reason is not a triumph of planning; it is a warning.

Who wins? Who gains if this pack is approved? Who loses if a more cautious pack is tabled? Does the process systematically select the rosiest forecast among competing bids for capital? If yes, misrepresentation pressure is structural even when individuals feel sincere. Changing the selection rule matters more than lecturing the current champion.

Accountability timing. Are early-stage numbers weakly accountable by design? What independent appraisal sits before lock-in? If the people who benefit from approval also control the only analysis, both bias and spin have an open door. Independence that arrives after political commitment is theatre.

Goals versus forecasts. Are stretch ambitions labelled as goals, or smuggled into the authorization forecast? Aggressive targets after commit can be healthy. The same numbers used to decide whether to commit are not goals; they are claims about reality. Rooms that refuse to separate the two will keep calling optimism “leadership” at the gate and “unforeseen complexity” after the overrun.

Class check. Is the champion’s plan the only analysis, or is a class comparison required before money moves? An inside-view monopoly is the natural habitat of both planning fallacy and strategic packaging. Requiring a class comparison does not settle every argument; it changes the burden of proof.

After commit. Once the project is approved, do not confuse delivery energy with rewriting the forecast that justified the go decision. Post-commit optimism can motivate execution. Quietly revising the case that unlocked the money to stay green is a different act — and it feeds later escalation when reality refuses to cooperate and the organisation has already spent reputation defending the original story.

Sponsor behaviour that rewards overpromise deserves one clear beat here: if sponsors celebrate only the greenest packs and punish bearers of cautious numbers, they are not victims of optimism bias — they are producers of the selection pressure that makes bias and misrepresentation rational. Fuller catalogues of sponsor failure modes belong elsewhere. The gate questions above are enough to make the dual diagnosis operational without stealing those catalogues.

These six questions are deliberately short. They are meant to be asked in the room, not filed as a checklist nobody reads. A chair who asks only one of them — usually the direction check — and skips who wins and accountability timing will catch some optimism and miss most misrepresentation. A chair who only hunts motive and never asks for a class comparison will catch some spin and miss most planning fallacy. The set works as a set.

Figure 4: Gate questions for bias and incentives. Six checks — direction, who wins, accountability, goals versus forecasts, class comparison, and after-commit discipline — make both psychology and incentives visible before money moves on large capital projects.

8. Stop Choosing One Villain

Large capital projects overrun for psychological and political-economic reasons that often co-occur. Early numbers are wrong in a systematic direction because optimism bias and the planning fallacy distort sincere forecasts, because approval incentives reward favourable presentation, and because organisational systems select the kindest story whether or not anyone sets out to deceive.

The either-or debate is a distraction. It produces either-or fixes: training without incentive redesign, or witch-hunts without forecasting discipline. Diagnose both. Separate goals from forecasts. Put class discipline and incentive design at the gate. Leave full forecasting methods, bias-hygiene catalogues, definition scorecards and stop rules to the places that own them. The first numbers on large capital projects will keep being wrong until leaders stop asking which single villain to blame — and start asking which of both drivers their gate still feeds.

References

• 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.
• 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.
• Lovallo, D., & Kahneman, D. (2003). "Delusions of Success: How Optimism Undermines Executives' Decisions." Harvard Business Review, 81(7), 56–63.
• Project Management Institute. (2014). Navigating Complexity: A Practice Guide. Newtown Square, PA: PMI.
• Merrow, E. W. (2018). Leading Complex Projects: A Data-Driven Approach to Mastering the Human Side of Project Management. Hoboken, NJ: Wiley.