Documented Project Failures and the Estimation Problem

Documented project failures reveal a recurring estimation problem. This post examines the Sydney Opera House, Channel Tunnel, Boston Big Dig, NHS National Programme for IT, Denver International Airport Baggage System, Scottish Parliament Building, Crossrail, and the Apollo Program to show how psychological and political mechanisms can produce systematic cost and schedule problems. The central mechanisms are optimism bias, the planning fallacy, strategic misrepresentation, escalation of commitment, and the megaproject paradox.

By Kateule Sydney — Researcher and Writer | Last Verified: 2026-10-05 | Published by Kat-Syd Resources Hub

Documented project failures are not merely unique disasters; they are data points in a recurring pattern in which psychological biases, political incentives, weak governance, and unrealistic assumptions can produce systematically optimistic estimates.

Table of Contents

• Chapter 1 — The Pattern: When Projects That Should Not Fail Do
• Chapter 2 — Optimism Bias and the Planning Fallacy: Why Estimates Go Wrong
• Case Study — Eight Documented Projects
• Chapter 3 — Strategic Misrepresentation: When Estimates Are Wrong on Purpose
• Chapter 4 — Escalation of Commitment and the Megaproject Paradox
• Conclusion
• References

Chapter 1 — The Pattern: When Projects That Should Not Fail Do

The projects examined in this post share three characteristics: they were large, strategically important, and affected by substantial cost or schedule problems. The pattern is not random.

What counts as a documented project failure? In this post, documented project failure refers to projects that exceeded their approved or publicly stated budgets or schedules by margins large enough to attract formal government audit, public inquiry, or substantial academic study. The term is deliberately broad because the cases differ in cause and severity.

Cost and schedule overruns are among the most visible indicators of project distress, although they do not by themselves prove that a project has failed in every respect. A project can exceed its budget and still produce substantial social or economic value.

The cases illustrate a wide range of problems:

• Sydney Opera House: The project was originally estimated at A$7 million (£3.5 million) and was expected to open in 1963. It opened in 1973 at a cost of about A$102 million — roughly 29 times the original estimate. Construction began before critical design problems had been resolved. The Sydney Opera House itself acknowledges that the decision to begin construction early created an “irreversible” momentum that contributed to subsequent problems.
• Channel Tunnel: Construction costs more than doubled from the initial prediction. Revenue forecasts were also substantially too optimistic. Eurotunnel reported a loss of £925 million in its first year of operation, and the company subsequently underwent major debt restructuring before reporting its first annual profit in 2007.
• Boston Big Dig: The project began with an estimated cost of approximately $2.6 billion and ultimately cost about $14.6 billion, making it one of the best-known examples of large-scale public-project cost escalation in the United States.
• NHS National Programme for IT: Contracts with CSC originally totalled £3.1 billion. By 2013, the Department of Health’s estimated total cost of programmes associated with the National Programme had reached £9.8 billion, while the benefits reported to March 2012 were £3.7 billion — about half the costs incurred to that point.
• Denver International Airport Baggage System: The automated baggage system was an ambitious technological undertaking involving approximately 17 miles of track and thousands of automated carts. Problems with the system contributed to delays and additional costs. A U.S. General Accounting Office report recorded $200 million in initial automated-baggage-system costs and approximately $300 million in capitalised interest associated with delays caused by the baggage-system problems.
• Scottish Parliament Building: The project was originally associated with an early estimate of £40 million, but costs eventually reached approximately £431 million. It opened in 2004, about three years later than originally anticipated. Investigations identified weaknesses in procurement, cost control, responsibility, risk management, and project governance.
• Crossrail: The original funding package agreed in 2010 was £14.8 billion. Later estimates placed the overall cost at £18.9 billion, excluding trains and depots. The central section also opened later than planned. The National Audit Office found that the compressed schedule, contractual arrangements, and absence of a realistic plan were compounded by a culture in which “can do” became unrealistic.
• Apollo Program: Apollo should not be described simply as a failed project. It ultimately achieved its primary objective, but it also demonstrates the risks of ambitious targets under severe technical and schedule constraints. NASA reported a total programme cost estimate of $22.718 billion in 1966 and $23.190 billion in 1967. The Apollo 1 fire and subsequent redesign work caused major schedule disruption, while the Lunar Module remained a critical technical pacing item.

The statistical pattern. The problem is larger than any one project. A study of 430 large public construction projects found an average cost overrun of 24.1% and an average schedule overrun of 46.1%. Approximately 87.4% of the projects experienced costs at or above their original estimates, while 95.3% took at least as long as initially planned. The study estimated that an 80th-percentile reference-class adjustment would require an uplift of approximately 35.1% for cost and 80.4% for schedule.

These figures do not mean that every project will overrun or that overruns are unavoidable. They mean that assuming an ordinary project will perform exactly as initially forecast is inconsistent with the historical evidence.

The key lesson is therefore simple: documented project failures are not merely a collection of isolated disasters. They are data points in a recurring pattern. For the structural causes that connect these failures, see my post on Why Large Projects Fail: The Structural Causes.

Chapter 2 — Optimism Bias and the Planning Fallacy: Why Estimates Go Wrong

The planning fallacy is not simply a character flaw. It is a well-documented cognitive tendency. People routinely underestimate how long tasks will take, even when similar tasks have previously taken longer than expected. Kahneman and Tversky identified the planning fallacy in 1979, and subsequent research has examined why individuals and organisations continue to produce optimistic forecasts despite contrary historical evidence.

Planning fallacy and optimism bias. The two concepts are related but not identical. Optimism bias is the broader tendency to expect favourable outcomes and underestimate negative events. The planning fallacy is a specific forecasting error in which people underestimate the time, cost, or difficulty of a future task.

One important explanation is that planners adopt an inside view. They concentrate on the particular characteristics of the project in front of them: What will we build? What resources will we use? What does our schedule look like? What can our team accomplish? The problem is that the inside view can cause planners to neglect the historical performance of comparable projects.

The alternative is the outside view. Instead of asking only, “How long will this project take?”, the planner asks: How long have comparable projects actually taken? This is the logic behind reference-class forecasting.

Reference-class forecasting. Reference-class forecasting estimates future outcomes from the actual distribution of outcomes achieved by comparable projects. The method is important because historical distributions contain information that a project-specific forecast may ignore.

Research using 430 large public construction projects found an 80th-percentile cost uplift of 35.1% and an 80th-percentile schedule uplift of 80.4%. Transport reference classes developed through the work of Bent Flyvbjerg and colleagues have also produced substantial recommended cost uplifts. For an 80% certainty level, the reported uplifts were approximately 32% for roads, 57% for rail, and 55% for fixed links such as tunnels and bridges. These are not arbitrary contingency percentages. They are derived from observed outcomes in comparable projects.

Why this matters. The practitioner who understands the planning fallacy accepts that a project-specific estimate may be systematically optimistic and tests it against historical evidence. The practitioner who assumes the estimate is accurate discovers the error later, during execution, when corrective action is more expensive and options are more limited. The lesson is not that optimism should be eliminated. It is that optimism should not be mistaken for evidence. For the decomposition discipline that supports more accurate estimation, see my post on the Work Breakdown Structure and the Art of Decomposition.

Case Study — Eight Documented Projects

The cases below illustrate how estimation interacts with design uncertainty, political pressure, governance, technology, incentives, and decision-making.

Sydney Opera House. The Sydney Opera House demonstrates the danger of starting construction before critical design problems have been resolved. Construction began in 1959 even though important structural questions remained unanswered. The Sydney Opera House’s historical account states that the site had not been adequately surveyed and that the weight and structural requirements of the roof were not yet known. It also notes that Premier Joseph Cahill was in a hurry to begin construction because of concerns that the project might be delayed by political or bureaucratic opposition.

The result was a chain of technical and management problems. Heritage NSW identifies the early commencement of construction, before the design for the shells and their supports had been resolved, as a major factor in the project’s difficulties. The project ultimately opened in 1973 at approximately A$102 million, compared with the original A$7 million estimate.

Lesson: political urgency can convert unresolved design uncertainty into future cost and schedule exposure.

Channel Tunnel. The Channel Tunnel illustrates the interaction between construction risk and demand forecasting. Construction costs more than doubled from the initial estimate. The project also suffered from substantially overestimated passenger and freight forecasts. Competition from ferries reduced the tunnel’s expected market share and put pressure on tariffs.

Eurotunnel reported a loss of £925 million in its first year of operation. The company’s heavy debt burden subsequently led to major financial restructuring. In 2007, the restructured group reported its first annual profit.

Lesson: a technically successful asset can still create severe financial problems if its cost and demand assumptions are unrealistic.

Boston Big Dig. The Boston Central Artery/Tunnel project, commonly known as the Big Dig, began with an estimated cost of about $2.6 billion and ultimately cost approximately $14.6 billion. The project illustrates the interaction between scope, complexity, governance, and cost control. Its final outcome cannot be explained by a single forecasting error. Design changes, construction difficulties, project complexity, and management and oversight problems compounded the original underestimate.

Lesson: an unrealistic baseline leaves little room to absorb the inevitable problems of complex execution.

NHS National Programme for IT. The NHS National Programme for IT demonstrates that technology projects can suffer from the same structural problems as physical megaprojects. The programme was launched to transform information systems across the National Health Service. The Department of Health entered major contracts with technology suppliers, including CSC. However, the programme encountered serious problems with delivery, implementation, contractual obligations, and benefits realisation.

The Public Accounts Committee reported that the original CSC contracts totalled £3.1 billion. The Department’s estimated total cost of the programmes associated with the former National Programme was £9.8 billion, while the benefits reported to March 2012 were £3.7 billion. The programme was formally dismantled, but the Public Accounts Committee noted that component programmes continued and continued to incur costs.

Lesson: ending a programme on paper does not necessarily end the institutional commitments created by it.

Denver International Airport Baggage System. The Denver baggage system is a classic example of technological ambition combined with schedule pressure. The automated system was intended to move baggage rapidly across the airport using thousands of automated carts travelling over an extensive network of track. The technology was highly ambitious for its scale and operating environment.

Testing revealed serious problems, including misrouted and damaged bags, mechanical failures, and difficulties operating the system reliably. The U.S. General Accounting Office reported that the initial automated baggage system cost approximately $200 million and that delays attributable to baggage-system problems generated about $300 million in capitalised interest. Additional costs were incurred to modify the system and establish backup arrangements.

Lesson: technological novelty increases estimation uncertainty, especially when an unproven system is placed on the critical path to opening.

Scottish Parliament Building. The Scottish Parliament Building at Holyrood illustrates problems associated with governance, procurement, scope development, and unrealistic early estimates. The project became associated with an initial figure of approximately £40 million, but costs eventually reached about £431 million. The building opened in 2004, roughly three years later than originally expected.

The project underwent repeated changes in cost expectations. An Auditor General report found that the £195 million cost plan established in 2000 was not a reliable prediction and that identified risks had not been adequately reflected in the budget. A contemporary PMI case study highlighted the absence of effective baselining, diffuse responsibility, and the use of parallel working that bypassed controls.

Lesson: a budget cannot function as an effective control baseline if the scope, risks, responsibilities, and design are not sufficiently defined.

Crossrail. Crossrail demonstrates how schedule optimism can become embedded in programme governance. The original 2010 funding package was £14.8 billion. By 2022, the estimated cost had risen to £18.9 billion, excluding trains and depots.

The National Audit Office’s 2019 investigation identified a compressed schedule, contractual problems, loss of downward pressure on costs, and the absence of a realistic plan. It concluded that an atmosphere in which “can do” had become unrealistic contributed to the programme’s difficulties. The problem was not simply that managers were optimistic. The original completion date became a constraint that influenced subsequent decisions, even as evidence accumulated that the programme was not on track.

Lesson: once a politically important deadline becomes embedded in governance, acknowledging that the date is unrealistic can become institutionally difficult.

Apollo Program. Apollo requires a different interpretation. It was not ultimately a failed project. The programme achieved its defining objective when Apollo 11 landed astronauts on the Moon in 1969. Nevertheless, Apollo illustrates the relationship between ambitious objectives, technical uncertainty, schedule pressure, and cost forecasting.

NASA’s historical records show that the total programme cost estimate rose from $22.718 billion in 1966 to $23.190 billion in 1967. The Apollo programme also experienced significant schedule slippage. NASA records indicate that the first human Apollo launch moved from the middle of the 1960s into 1967, while the lunar-landing schedule moved into late 1968. The Apollo 1 fire in January 1967 caused further disruption. NASA also identified the Lunar Module as the ultimate pacing item because fundamental technical problems had to be solved before the vehicle could be ready.

Lesson: ambitious targets can mobilise extraordinary performance, but they do not eliminate technical uncertainty. Money and political urgency cannot always compress fundamental engineering development.

Chapter 3 — Strategic Misrepresentation: When Estimates Are Wrong on Purpose

Optimism bias explains why estimates can be systematically optimistic even when nobody intends to deceive anyone. Strategic misrepresentation is different. It refers to deliberate distortion of costs, benefits, risks, or schedules in order to improve the likelihood that a project will receive approval or funding.

Bent Flyvbjerg’s research argues that political and economic incentives can produce systematic misrepresentation in major infrastructure planning. The Inter-American Development Bank similarly distinguishes between optimism bias, which can operate without conscious deception, and strategic misrepresentation, which involves deliberately presenting a more favourable picture to decision-makers.

How strategic misrepresentation works. Costs are understated. A lower cost estimate can make a project appear more attractive during the approval process. The problem is that the initial estimate can become politically difficult to challenge after approval.

Benefits are overstated. Demand forecasts can be particularly vulnerable. The Channel Tunnel provides a clear example. Passenger and freight demand was substantially overestimated, while competition from ferries reduced the project’s expected market share.

Political deadlines distort technical decisions. The Sydney Opera House illustrates how political timing can influence construction decisions. The project began before important design problems had been resolved because the government wanted to establish momentum.

Incentives become misaligned. The person who benefits from project approval may not be the person who bears the cost of an overrun. That creates a fundamental governance problem. If a project promoter receives political credit for announcing a new railway, hospital, airport, or technology programme, while future governments and taxpayers absorb the overruns, the incentive structure does not naturally produce conservative estimates.

Reference-class forecasting as a countermeasure. The countermeasure is not simply to tell planners to be less optimistic. It is to change the forecasting process. Reference-class forecasting forces decision-makers to compare the proposed project with the actual historical distribution of comparable projects.

The proposed Zuiderzee Line in the Netherlands provides a useful example. Its promoter estimated the cost at €4.1 billion. A reference class of 68 comparable rail projects suggested that the estimate should have been approximately 39% higher, at €5.7 billion, before applying a more conservative uplift for a lower probability of cost overrun.

The broader lesson is that estimates are not purely technical documents. They are also products of institutional incentives. For the governance structures that can counteract misrepresentation, see my post on The PMO, Project Governance, and Benefits Realization.

Chapter 4 — Escalation of Commitment and the Megaproject Paradox

Even when problems become visible, projects do not necessarily stop. This is where escalation of commitment becomes important.

Escalation of commitment is the tendency of decision-makers to persist with a failing course of action. Research in the Academy of Management Review identifies self-justification as one important explanation: decision-makers may resist admitting that resources already committed to a chosen course were spent in vain.

The psychology becomes particularly powerful when the decision has already been made publicly. Cancelling a project may mean admitting that the original estimate was wrong; that the business case was weak; that previous expenditure cannot be recovered; that political promises cannot be fulfilled; or that senior decision-makers made a poor decision. As a result, the question can subtly change from “Should we continue this project?” to “How can we rescue the project we have already committed to?” That is the sunk-cost problem.

The megaproject paradox. The megaproject paradox describes the persistence of major projects despite extensive evidence that megaprojects frequently experience cost overruns, delays, and benefit shortfalls. Research has identified this paradox as a recurring feature of megaproject decision-making.

Megaprojects can create powerful political incentives because their benefits are visible while many of their costs are dispersed over time. The pattern can be seen in several cases:

• Boston Big Dig: escalating costs did not lead to cancellation because the project had accumulated political, economic, and constituency support.
• Crossrail: the original timetable remained an important organising assumption even as the evidence increasingly indicated that the programme was not on track.
• NHS National Programme for IT: although the programme was formally dismantled, component programmes continued and continued to incur costs.
• Scottish Parliament Building: audit and management warnings accumulated while the project continued through successive stages of development.

The Channel Tunnel exception. The Channel Tunnel provides an important qualification. The project came close to financial collapse but ultimately survived through restructuring. Eurotunnel’s 2007 results showed the first annual profit after a major restructuring of its debt.

The lesson is not that continuing a troubled project is always irrational. A project may have strong underlying value even after its original business case has deteriorated. The correct question is therefore not “How much have we already spent?” It is: “Given today’s information, is the expected future value of continuing greater than the future cost and risk of doing so?” Past expenditure is a sunk cost. It should not determine the future decision.

The governance response. The practical response to escalation of commitment is structural. Organisations should establish independent stage-gate reviews; separate the authority to approve continuation from the team responsible for the original decision; require updated cost and benefit estimates; test the project against reference-class evidence; establish explicit cancellation criteria; and treat sunk costs as sunk costs rather than as investments that must be protected.

The discipline of project management is therefore not only about knowing how to continue. It is also about knowing when to stop. For the governance structures that can enforce independent review, see my post on Risk, Procurement, and Stakeholder Management.

Conclusion

The documented cases examined here reveal a recurring pattern. Optimism bias helps explain why estimates can be systematically optimistic. The planning fallacy explains why planners can focus on the current project while discounting the historical performance of comparable projects. Strategic misrepresentation explains why some estimates may be deliberately distorted to secure approval or funding. Escalation of commitment helps explain why projects can continue even after evidence of serious problems becomes available. The megaproject paradox explains why these patterns persist despite decades of evidence.

My central argument is that the estimation problem is not primarily technical. It is also psychological, organisational, and political. Better estimating techniques exist. Reference-class forecasting, independent review, staged approval, risk analysis, and transparent assumptions can all improve decision-making. But methods alone are insufficient if the surrounding incentives reward optimistic forecasts, fixed political commitments, and continued spending.

Five recommendations follow from the analysis:

• Use reference-class forecasting. Estimate from the outside as well as the inside. Compare the proposed project with the actual performance of similar projects.
• Separate the decision to proceed from the decision to continue. Independent review should have sufficient authority to recommend cancellation when the evidence changes.
• Treat estimates as governance artefacts. An estimate is not merely a spreadsheet. Ask who benefits from optimism, who bears the consequences of overruns, and whether those incentives are aligned.
• Allow the plan to change. A project baseline should be a control mechanism, not a political promise that cannot be revised. When evidence changes, the plan should change.
• Turn lessons learned into institutional action. Audits, inquiries, and post-project reviews are valuable only when their findings change future behaviour.

The argument in one sentence is this: documented project failures are not merely unique disasters; they are data points in a recurring pattern in which psychological biases, political incentives, weak governance, and unrealistic assumptions can produce systematically optimistic estimates and make correction increasingly difficult once a project is under way.

Accurate estimation does not guarantee project success. Execution can still fail. Technology can still disappoint. Scope can still change. External events can still intervene. But a realistic baseline gives a project something that an unrealistic baseline cannot provide: room to absorb uncertainty without turning every execution problem into a crisis.

References

• Buehler, R., Griffin, D., & Ross, M. (2012). “Inside the Planning Fallacy: The Causes and Consequences of Optimistic Time Predictions.” In Heuristics and Biases: The Psychology of Intuitive Judgment, pp. 250–270. Cambridge University Press.
• Brockner, J. (1992). “The Escalation of Commitment to a Failing Course of Action: Toward Theoretical Progress.” Academy of Management Review, 17(1), 39–61.
• Flyvbjerg, B. (2007). “Policy and Planning for Large-Infrastructure Projects: Problems, Causes, Cures.” Environment and Planning B, 34(4), 578–597.
• Flyvbjerg, B., Bruzelius, N., & Rothengatter, W. (2003). Megaprojects and Risk: An Anatomy of Ambition. Cambridge University Press.
• Inter-American Development Bank. (2017). Managing Megaprojects: Conceptual Framework and International Experience.
• Kim, K. (2017). “Effects of the Optimism Bias on Large Public Construction Projects: An Investigation Using Reference Class Forecasting Technique.” Journal of Policy Analysis and Evaluation, 27(3), 97–131.
• National Audit Office. (2019). Completing Crossrail.
• National Aeronautics and Space Administration. (2004). Effects of the Apollo 204 Accident on Schedule and Cost of the Apollo Program. NASA History Office.
• National Aeronautics and Space Administration. Background and Analysis: Apollo Programme History. NASA History Office.
• Project Management Institute. (2004). “Digging Deep: The Big Dig.” PM Network.
• Project Management Institute. (2004). “Plan Has Merit—It Cost: Scottish Parliament Building.” PMI Learning Library.
• U.S. General Accounting Office. (1995). Denver International Airport: Impact of the Delayed Baggage System. GAO/AIMD-95-230.
• UK Parliament, Public Accounts Committee. (2013). The Dismantled National Programme for IT in the NHS.
• Sydney Opera House. Construction Begins.
• Heritage NSW. Sydney Opera House.
• Global Infrastructure Hub. The Channel Tunnel Case Study.

Read Also on Kat-Syd Resources Hub

The PMO, Project Governance, and Benefits Realization — The PMO, governance structures, benefits realization, and project closure as organizational mechanisms connecting delivery to strategy.

Risk, Procurement, and Stakeholder Management — Risk management, procurement, stakeholder management, and change control as interconnected governance functions.

Project Management Across Sectors — Project management practices adapted across construction, IT, healthcare, government, and regulated sectors, with principles that remain constant.

Written by Kateule Sydney — Researcher and Writer

© 2026 Kat-Syd Resources Hub. All rights reserved.

Kat-Syd Resources Hub — Educational case studies and analytical reference