Why Everything Takes Longer Than You Think

If you have ever finished something exactly on the day you first thought you would, treasure the memory - it is rarer than it feels. Researchers gave the pattern a name back in the 1970s, and by now it has been documented in students finishing assignments, governments building railways and almost everyone estimating how long a work e-mail will take to answer.

A question about your own tasks

In a series of five studies published in 1994, the psychologists Roger Buehler, Dale Griffin and Michael Ross asked 465 university students to predict how long various tasks would take - finishing an academic assignment, and a range of everyday nonacademic tasks. Across every study, the same asymmetry showed up. Students were too optimistic about their own completion times, routinely finishing later than they had predicted. But when the same students were asked how long it would take someone else to do the identical task, their estimates were considerably more realistic, sometimes erring toward pessimism instead. People were not bad at predicting task duration in general; they were specifically bad at predicting their own.

Why looking forward misleads you

The explanation the researchers landed on was about where people look for information. Asked to predict their own future, participants reported building their estimate mainly from a mental simulation of the task going smoothly this particular time: this is what I plan to do, this is roughly how long each step should take, assuming nothing goes wrong. What they largely ignored was their own track record - how long the last few similar tasks had actually taken, including the delays and interruptions that showed up every time. Predicting someone else's task, by contrast, they had no plan to simulate, so they leaned more on general expectations of how such things go, which happened to be closer to reality.

The same idea, from the outside

This distinction had already been described a few years earlier by Daniel Kahneman and Amos Tversky, who separated forecasting into an "inside view" - reasoning from the specifics of the case in front of you, your particular plan and its particular steps - and an "outside view" - reasoning from a wider set of similar cases and how they actually turned out, regardless of how each one was planned. Their argument was that the inside view is where the optimism creeps in: a plan, by its nature, describes how things are supposed to go, and leaves out the unplanned parts that eat time in every real project. The outside view does not care about your plan at all; it only looks at outcomes.

Turning the outside view into a method

Decades later, the economic geographer Bent Flyvbjerg took this idea out of the psychology lab and into infrastructure policy, where planning-fallacy-sized errors show up as real overrun costs on bridges, railways and public buildings. His proposed fix, reference class forecasting, skips the specific project's plan almost entirely at the estimation stage. Instead, it identifies a "reference class" of genuinely comparable past projects, looks at how far their actual costs and timelines ended up from their original estimates, and adjusts the new project's forecast using that real distribution of past overruns - rather than using the details of the new project's own blueprint, however carefully drawn. It is, in effect, forcing an inside-view planner to answer an outside-view question: not "how will this go", but "how have things like this actually gone". Public infrastructure studied this way showed overruns were not just common but fell into a fairly consistent proportion of the original budget for a given type of project, which is itself useful: a predictable-sized error is one you can correct for in advance, even when it cannot be eliminated.

Knowing about it does not cure it

Awareness does not appear to fully close the gap on its own. People can recite the planning fallacy back to a researcher and then walk out and make exactly the same kind of overly optimistic estimate for their next task, because the bias is not a knowledge problem. It happens each time someone builds a forecast by imagining a specific, uninterrupted run at the task rather than pulling up a memory of how the last one actually went, and knowing the concept in the abstract does not automatically trigger pulling up that memory in the moment a deadline is being set. That is part of why reference class forecasting works better as a fix than a warning does: it changes the question being asked, rather than relying on someone remembering to be humble.

What this means for you

None of this means your plans are worthless or that you should stop making them - the fix is not less planning, it is adding one outside-view question to a process that is otherwise all inside view. Before committing to a deadline, it helps to ask a second question alongside "what does my plan say": how long did the last few similar things actually take me, including the ones that went badly? If those two numbers disagree, the honest move is usually to trust the second one, pad the estimate, and build in the kind of buffer your plan does not naturally include. If missed deadlines are a source of real distress rather than mild inconvenience, that is worth exploring with a professional rather than solving through better spreadsheets alone. If you want to look at your own patterns around deadlines, optimism and putting things off, our time management, optimism and procrastination tests each look at a different piece of this puzzle.