The Planning Fallacy: Why Everything Takes Longer Than You Think

The persistence is the interesting part. Knowing about the planning fallacy does not fix it, and neither does personal experience of being wrong in exactly the same way.

The finding

Daniel Kahneman and Amos Tversky named it in 1979. The observation is that predictions about a specific task are systematically optimistic, while general knowledge about how such tasks usually go is more accurate — and the two are held simultaneously without contradiction being noticed.

The clearest demonstration comes from Roger Buehler and colleagues, who asked students to estimate when they would finish their thesis, both under normal expectations and in a worst-case scenario where everything went wrong.

On average, students finished later than even their worst-case estimate. Fewer than half were done by the date they had given as an outer bound assuming disaster.

That result is what separates this from ordinary optimism. The worst-case estimate was supposed to be the pessimistic one, and it was still too optimistic.

Why experience does not help

The explanation Kahneman later formalised distinguishes two ways of estimating.

The inside view builds a forecast from the specifics of this task: the steps involved, how long each should take, what the plan is. It produces a coherent narrative of the task proceeding as intended.

The outside view ignores the specifics and asks how long comparable tasks have historically taken. It requires treating the current task as one instance of a class rather than as a unique case.

People overwhelmingly use the inside view, and it fails predictably. A plan is a description of things going right; the reasons things take longer are individually unlikely and collectively near-certain. You cannot enumerate them in advance, but the historical data has already absorbed all of them.

This is also why an incomplete task keeps its grip differently than a completed one, and why writing a concrete plan changes what an unfinished goal does to you — see the Zeigarnik effect.

The asymmetry that keeps it alive

One detail explains why the bias never corrects itself through feedback.

When a task overruns, the reason is always available and always specific: a supplier was late, someone was ill, a requirement changed. Each delay comes with its own explanation, and each explanation feels like a one-off.

So the lesson learned is not that estimates are systematically low; it is that this particular project had bad luck. The next estimate is made the same way, on the assumption that this time nothing unusual will happen.

The bias survives on the availability of excuses. Every instance is genuinely explicable, and the pattern across instances goes unexamined.

There is a further wrinkle that makes the bias harder to see in organisations: the incentives point the same way as the bias. An optimistic estimate is what gets a project approved, so the people producing forecasts are rarely rewarded for accuracy. Whether that is bias or strategic misrepresentation is a live question, and the two are difficult to separate from the outside.

Reference class forecasting

The remedy that follows from the diagnosis is to force the outside view — to find comparable past cases and use their actual durations as the starting point, adjusting only where there is specific reason to.

Bent Flyvbjerg has applied this to large infrastructure projects, where cost and schedule overruns are endemic, and reference class forecasting has been adopted in some public-sector planning guidance as a result. The method is unglamorous: rather than modelling your project, you look up what happened to similar projects.

The obstacle is not technical. It is that the outside view feels like it ignores everything you know about your own situation, and that its answer is usually unwelcome. A forecast that says this will take twice as long as you think, based on projects you consider less well run than yours, is easy to reject on plausible-sounding grounds.

Practical versions are cruder and still work. Track how long your own recurring tasks actually take rather than how long you expect them to. And when you cannot resist an inside-view estimate, apply a multiplier derived from your own history rather than from optimism about this occasion.

The effect is also asymmetric in an important way: it applies to your own future tasks far more strongly than to other people’s. Asked how long a colleague’s project will take, people give noticeably more accurate estimates, because they are forced into something closer to the outside view by not having access to the plan.

Why it gets worse with more planning

A counterintuitive corollary: detailed planning can make the estimate worse rather than better.

The more granular the plan, the more coherent and plausible the narrative of the task going as intended — and the more confident the resulting forecast. Breaking work into steps and estimating each one produces a total built entirely from best cases, with no allowance for the interactions and interruptions that consume most of the overrun.

This is why organisations with elaborate planning processes are not obviously better at hitting schedules than those with crude ones. The process improves the inside view; the inside view is the problem.

What it is not

Two clarifications, because the term gets stretched.

It is not laziness or poor discipline. The effect appears in careful, motivated people estimating tasks they care about, and it appears in organisations with formal planning processes.

And it is not general optimism about everything. The planning fallacy is specific to duration and cost estimates for tasks one intends to carry out. It sits alongside other systematic distortions in how time is judged, including the compression of long periods in retrospect — different mechanism, same broad theme that temporal judgement is built rather than measured.