A project manager puts a 2.4-million-euro estimate on the table for a system replacement that ends up costing 3.4 million — an overrun of forty percent that is more predictable than most sponsors realise. Ask any project team for an estimate and you almost always get the same thing: a number too low for the costs and too high for the benefits. That is not chance and not an isolated error. It is a structural pattern, measured across thousands of projects, and one of the best-documented pitfalls in the entire field. For anyone judging a plan of approach, or drafting one, that pattern is not trivial knowledge: it determines whether you can trust the figures at face value.
2,062 projects, one clear pattern
In 2021 Flyvbjerg analysed a dataset of 2,062 capital projects across eight types. The finding is sobering: actual costs were on average 1.39 to 1.43 times the estimate, and actual benefits only 0.83 to 0.94 times what was predicted. Both with a p-value below 0.0001: this is not noise, this is a regularity.
It becomes even sharper in a subset: in 53 of 62 rail projects demand had been overestimated, on average 41 percent lower than predicted. Flyvbjerg calls it "the planning fallacy writ large". Estimates do not fail randomly up and down. They fail consistently in the direction that makes the project look more attractive than it is.
Two sources: error and interest
Two mechanisms lie behind this, and the difference between them determines how you should respond.
The first is cognitive bias: honest optimism. Kahneman showed that people naturally take the inside view: they reason from the specific details of their plan and systematically underestimate how often things go wrong. This is sincere; no one is lying.
The second is political bias: strategic misrepresentation. Wachs described in 1989, and Flyvbjerg, Holm and Buhl in their study with the telling title "Error or Lie?", that estimates are sometimes deliberately presented in rosy terms to obtain approval. Whoever knows that an honest estimate would sink the project has an incentive to low-ball.
For those assessing a plan, the lesson is hard but necessary: you may not assume that an over-optimistic estimate is merely an error. Independent support counts, precisely because from the outside you cannot see whether the optimism is honest or strategic.
The remedy: the outside view
The good news is that there is a remedy, and that it is surprisingly simple: the outside view, also known as reference class forecasting. Instead of building up your estimate from the details of your own plan, you calibrate it against the actual outcomes of a class of comparable, completed projects.
Not: "how much do we think this will cost?" But: "what did projects like this ultimately really cost?" That second question sidesteps the optimism, because it starts not from hope but from history.
Kahneman called reference class forecasting "the single most important piece of advice regarding how to increase accuracy in forecasting". It is not a marginal technique; according to the founder of the bias literature it is the single most important piece of advice that exists. In the United Kingdom it has been institutionalised in the UK Green Book, the guidance for public investment.
The question is not how carefully you have calculated your project, but whether you have compared your figure with what actually happened on projects like yours.
An honest nuance
The literature is not unanimous on the causal primacy of optimism bias. Love and Ahiaga-Dagbui (2018) point out that part of the cost overruns has to do with baseline artefacts and later scope changes, not with bias in the original estimate. The picture is thus more nuanced than "everyone is too optimistic".
But, and this is decisive, even the critics embrace empirical benchmarking. Whether the cause is bias or scope growth, the answer remains the same: test your estimate against experience figures and build in room for setbacks. On the remedy there is broad consensus, even where the cause is debated.
What the highest level requires
APM states that business cases must be "supported by relevant and realistic information": realism is an explicit quality criterion, not merely the presence of figures. The Court of Audit requires that ambition be "demonstrably in line with people, resources and time".
Concretely that means: a high-level estimate is not a neatly built-up point estimate. It is an estimate with a range, with a contingency for setbacks, and with a test against experience figures from comparable projects. A single, precise number without margin is in fact a warning signal: precision without support is false certainty.
What this means for you
Take the most important estimate in your plan and find one reference project for it: something comparable that has already been completed. Is your number in the range of what that project actually cost, or is there no benchmark under your figure at all?
Whoever can answer that one question has already dismantled the most predictable pitfall in project management, before the start.
Sources
- Flyvbjerg, B. (2021), Top Ten Behavioral Biases in Project Management, Project Management Journal.
- Wachs, M. (1989); Flyvbjerg, B., Holm, M.K.S. & Buhl, S.L. (2002), Underestimating Costs in Public Works Projects: Error or Lie?, Journal of the American Planning Association.
- Flyvbjerg, B. (2013), Quality control and due diligence in project management, International Journal of Project Management.
- Love, P.E.D. & Ahiaga-Dagbui, D.D. (2018), zoals aangehaald in de PlanScore Body of Knowledge (kritische kanttekening bij optimism bias).
- APM — eis dat business cases "supported by relevant and realistic information" zijn.
- Algemene Rekenkamer — eis dat ambitie "aantoonbaar in overeenstemming is met mensen, middelen en tijd".