Planning Fallacy: Why Everything Takes Longer Than You Planned

Kahneman and Tversky named the planning fallacy in a 1979 paper on intuitive prediction, and their account of it rests on a distinction between two ways of forecasting. The inside view builds the forecast from the case itself: the steps the task will need, how long each should take, what could go wrong. The outside view asks how long tasks of this kind have taken before, and starts from that distribution. People planning their own work take the inside view almost exclusively, and the inside view is systematically optimistic, because it is a plan, and a plan is a story about things going as intended.

Buehler, Griffin and Ross showed in 1994 how the two views coexist in one head. Their students knew that past projects had run late, and the detail of their own next task was the problem: they could see exactly how it would go. A scenario of the future, however detailed, is one path among many, and the paths with a flat tire, a sick child or a corrupted file are each improbable while being, together, where most of the delay lives.

Three features keep the fallacy in place. People do not spontaneously consult their past when planning; when they are asked to, they explain the past delays away as circumstances that will not recur. Plans are made at the level of the task and not of the week, so the other demands on the same hours are left out. And optimism about completion is partly motivated: the student who predicts an early finish gets the pleasure of it now, and the deadline sometimes moves to meet the prediction. Buehler, Griffin and Peetz's 2010 review sets out these cognitive, motivational and social origins and finds the pattern across dozens of studies.

The fallacy scales. Flyvbjerg, Holm and Buhl's 2002 study of 258 transport projects over seventy years found costs underestimated in nearly nine of ten, with no improvement over the period. They argued that at that scale the optimism is strategic as well as cognitive, since the forecast that gets a project approved is the hopeful one.

Roger Buehler, Dale Griffin and Michael Ross published "Exploring the planning fallacy: Why people underestimate their task completion times" in the Journal of Personality and Social Psychology in 1994. The first of its five studies gave the fallacy its best-known numbers.

Thirty-seven psychology students in the final semester of the honors thesis course at the University of Waterloo were telephoned and asked to predict as accurately as possible when they would submit their finished thesis. They were also asked for the dates if everything went as well as it possibly could and if everything went as poorly as it possibly could. The course coordinator recorded the actual submission dates. Four students had not finished when the records closed, so the completion-time means below are for the other 33, while the percentages count all 37, including the four who had not finished.

The students' best estimates averaged 33.9 days from the survey. Their actual completion averaged 55.5 days. Only 29.7 percent of them finished by the date they had predicted. Their optimistic estimates averaged 27.4 days, and 10.8 percent finished by that date. The result that made the study famous is the pessimistic one: asked to imagine everything going wrong, the students gave an average of 48.6 days, still a week short of reality, and fewer than half of them, 48.7 percent, finished even by their worst-case date. The predictions were not uninformative: students who predicted a longer time did take longer, a correlation of .77. The whole group was simply shifted toward optimism.

The later studies in the paper looked for the cause. Students asked to think aloud while predicting focused on plans for the future and rarely mentioned past experience; when they did recall past delays, they attributed them to circumstances that would not apply this time. Recalling past completion times did not by itself fix the predictions. What helped was requiring them to connect the past to the present task, saying how relevant past experiences bore on the one in front of them. That is the outside view made into a procedure.

The fallacy appears wherever a person predicts their own completion time from a plan. The rows below are illustrations, not findings from the studies cited on this page.

The counter with the most evidence is the outside view: starting from how long tasks of this kind have taken rather than from how this task should go.

• Find the reference class. Before estimating, ask what category this task belongs to and how long members of that category have actually taken, in your own record or in others'.
• Connect the past to the present out loud. In the 1994 studies, recalling past delays was not enough; the correction came when students were made to say how those past experiences applied to the task at hand.
• Plan in weeks, not tasks. Put the task into a calendar that already contains everything else, and the hours it competes with become visible.
• Unpack the bad paths together. No single mishap is likely; the chance that none of them happens is small. Add a contingency for the category of mishap rather than for each one.
• Keep a record. A simple log of estimate and actual, kept for a few months, gives you a personal correction factor.
• Distrust the worst case. In the thesis study, more than half of the students missed their own worst-case date. A safe date starts from the record, not from imagination.

The planning fallacy is a specific optimism about one's own time and cost estimates. Its neighbors share the optimism or the forecasting error without being the same thing.

• It is not optimism bias in general, the tendency to expect better outcomes than the evidence supports. The planning fallacy is optimism applied to completion times and costs, and it has the specific mechanism of the inside view; a person can be a pessimist about life and still plan a renovation in six weeks.
• It is not overconfidence, which is excess certainty about one's own judgments. The thesis students were not necessarily sure of their dates; they were systematically early, which is a bias in the estimate rather than in confidence about it.
• It is not hindsight bias, the sense after the fact that an outcome was predictable. Hindsight distorts memory of what one expected; the planning fallacy distorts the expectation itself, and the two together explain why the last late project does not make the next plan more cautious.
• It is not the sunk cost fallacy, though the two often meet in the same project: the optimistic plan starts it and the sunk costs keep it going.