The Planning Fallacy: Why Everything Takes Longer Than You Think

Everything takes longer than you think because when you estimate, you build a scenario for how this particular project will unfold, and you do not consult the record of how similar projects actually went. Daniel Kahneman and Amos Tversky named this the planning fallacy in 1979 and named the two modes it depends on: the inside view, which reasons from the specifics of the case in front of you, and the outside view, which reasons from the distribution of outcomes for cases like it. The inside view produces a plausible story. The outside view produces an accurate number. People overwhelmingly use the first.

What makes this more than a productivity complaint is that the bias is unusually resistant. It survives experience, expertise, and being told about it, which is a rare property for a cognitive bias and the part that gets left out.

Where it was named

The term comes from "Intuitive Prediction: Biases and Corrective Procedures," in TIMS Studies in Management Science. The illustration Kahneman used for the rest of his career was his own.

He was on a team writing a curriculum textbook. A year in, he asked everyone to privately estimate how long the project would take. The estimates clustered between eighteen months and two and a half years. He then asked a colleague, an expert in curriculum development, to think of comparable projects by other teams. After a long pause, the colleague reported that roughly forty percent had never finished at all, and none that did had taken less than seven years.

The team heard that. They finished in eight years, and the book was never used. That is the phenomenon in miniature: the inside view produced a confident estimate, the outside view produced an accurate one, and the outside view lost anyway.

The study that measured it

Roger Buehler, Dale Griffin and Michael Ross put numbers on it in 1994 in the Journal of Personality and Social Psychology, in "Exploring the planning fallacy: Why people underestimate their task completion times."

Their cleanest study followed psychology students working on their honours theses. Each predicted when they would finish. The average prediction was about 34 days; the average actual completion was about 55 days. Roughly a third finished by the date they had named.

The detail that makes this hard to explain away came next. Students were also asked for a worst case, assuming everything went as badly as possible. On average, they still finished later than their own worst case. This is not overconfidence that a nudge toward caution would fix. Asked explicitly to be pessimistic, they produced a number reality beat anyway.

Buehler and colleagues also found the asymmetry that gives the mechanism away. People estimating for someone else are markedly less optimistic. The bias is not about the task but about being inside it, holding a plan that feels like information and works as a distraction from the record.

Why experience does not fix it

Everyone has been late before, so why does memory not correct the estimate. Buehler's answer is that it does not get used. People readily recall past overruns and then explain them: that time there was a family emergency, that time the client changed the spec. Each failure is attributed to a disruption that will not recur, and the base rate is dismantled one anecdote at a time.

A second account is less flattering to the motivational story. Michael Roy, Nicholas Christenfeld and Craig McKenzie argued in 2005 in Psychological Bulletin that much of the effect may be memory rather than optimism: people misremember how long past tasks took, recalling them as shorter than they were. If your stored durations are too short, an unbiased prediction drawn from them is still too short. That makes this partly a data problem rather than purely a wishfulness problem.

The correction: reference class forecasting

Bent Flyvbjerg turned Kahneman's outside view into a procedure for infrastructure, where the stakes are large enough that the bias shows up in national accounts. His evidence base is the reason to listen. Flyvbjerg, Mette Skamris Holm and Soren Buhl studied 258 transport infrastructure projects across 20 nations and published the results in the Journal of the American Planning Association in 2002, under the title "Underestimating Costs in Public Works Projects: Error or Lie?" Costs were underestimated in roughly nine out of ten projects. Average overruns ran to about 45 percent for rail, 34 percent for bridges and tunnels, and 20 percent for roads, and the bias showed no sign of improving across the seventy years the data covered.

His fix appears in "From Nobel Prize to Project Management: Getting Risks Right," in Project Management Journal in 2006. Reference class forecasting has three steps, and the method is in the ordering.

1. Identify a reference class of past, completed projects that resemble yours. Not your best cases. Not the ones you remember. The class, including the failures.
2. Establish the distribution of actual outcomes for that class. Not the average of what those teams predicted. The average of what happened.
3. Place your project within that distribution. Start from the class outcome and adjust only for differences you can defend, which will be fewer than you expect.

The critical feature is what step three excludes. You are not permitted to begin from your own plan and add a buffer. You begin from the base rate and justify departures from it. Every instinct pushes the other way, because the plan is vivid and the base rate is a number about strangers.

This is not a productivity tip. The American Planning Association endorsed the method in 2005, and the UK Treasury's Green Book requires explicit optimism bias uplifts drawn from historical data on comparable projects, because the alternative was demonstrably worse.

The honest note: knowing does not help much

Debiasing research here is not encouraging. Informing people about the bias produces little change. Asking them to recall similar past tasks produces little change unless they are explicitly required to connect that recollection to the current estimate. Kahneman was told the base rate for his own project, by an expert, in the room, and the team proceeded on the inside view anyway.

Daniel Lovallo and Kahneman made the organizational version of this argument in Harvard Business Review in 2003, in "Delusions of Success." Optimistic forecasts are not merely tolerated inside organizations but selected for, because the project that gets funded is the one with the attractive projection. Under that incentive, the outside view is professionally costly to supply.

So the practical version is structural rather than mental. You do not beat the planning fallacy by resolving to be realistic. You beat it, partially, by making the outside view a required input someone must produce before the estimate is accepted. Write the base rate down before you write the plan, and pre-commit to the rule that departures from it need a stated reason, which is the same if-then logic covered in the piece on implementation intentions.

Where this fits

The planning fallacy is the operational face of a broader pattern. Believing things will go well is not free; it is paid for in the preparation you skip, which is the distinction drawn in the piece on when positive thinking makes you less prepared.

And it explains why the people who annoy planning meetings are sometimes the useful ones. Detailed rehearsal of what could go wrong is the same move that improves performance for a specific group, described in the piece on when expecting the worst actually helps. The person listing failure modes is doing forecasting work, and the room's instinct to reassure them is the room losing information.

The reason it never feels like a bias from the inside is that your plan is not wrong. Everything in it is achievable. What it omits is the interruption that has not been invented yet, and the only record of how often those arrive is every similar thing you have already done.