Methods, done simply

The planning fallacy

The planning fallacy is our habit of underestimating how long our own tasks will take, even when we know similar tasks ran late before. Daniel Kahneman and Amos Tversky named it in 1979, and studies since have found it in theses, tax returns and building projects. The fix with the best evidence is to plan from how long similar things took you before, not from the plan in your head. Breaking tasks into parts helps sometimes, not always.

Coming soon to iPhone and iPad

Last checked

On this page

Where the idea comes from

Daniel Kahneman and Amos Tversky, the psychologists whose work on judgement later earned Kahneman a Nobel prize, proposed the planning fallacy in 1979. Their point was that when we predict our own projects, we take the “inside view”: we imagine how this particular plan will unfold and ignore how projects like it have actually gone. In 2003, Dan Lovallo and Kahneman widened the idea to costs and benefits too, which is why it comes up in discussions of overrunning bridges and railways as well as late essays.

What the studies found

The best-known test is a 1994 study by Roger Buehler, Dale Griffin and Michael Ross. Psychology students predicted how long their senior theses would take. On average they said 33.9 days; even their “if everything goes as poorly as it possibly could” guess averaged 48.6 days. The theses actually took 55.5 days, and only about 30% of students finished in the time they predicted.

The same paper found three things that matter for your list:

  • We are optimistic about our own tasks, not other people’s. Observers predicting someone else’s completion time tended to overestimate it.
  • We plan from scenarios, not from history. Asked to think aloud, people talked mostly about how the task would go, rarely about how long similar tasks had taken.
  • We excuse the past. People explained their earlier late finishes as one-off bad luck, which let them ignore those lessons next time.

It is not only students. A survey of Canadian taxpayers, published in 1997, found they mailed their forms about a week later than they predicted, despite knowing exactly how late they usually were.

Fixes, and how well they work

Fix What it means Evidence
Look at past tasks Before estimating, recall how long similar tasks took you Strong: in the 1994 study, linking predictions to past experience removed the optimistic bias
Reference class forecasting Base the estimate on the actual outcomes of a group of comparable projects Developed by Bent Flyvbjerg for large projects; works where comparable cases exist
Unpacking List every part of the task before estimating Mixed: Kruger and Evans (2004) found it reduced bias; other studies found similar strategies did not help
Backward planning Plan from the end goal back to today Promising: a 2016 set of four experiments found longer, less biased predictions
A detailed step-by-step plan Visualise each step going well Can backfire: it sometimes makes people more optimistic

What the evidence says and where it is weak

The core finding, that people underestimate their own completion times, is one of the more robust results in judgement research. The weak spots are elsewhere. Much of the evidence comes from students and from one-off projects; everyday tasks you do often, such as a weekly shop, are estimated better, because you have a fresh record. The remedies are less settled than the problem: a 2016 review of interventions by Jessica Wiese, Roger Buehler and Dale Griffin describes the unpacking results as somewhat mixed, notes that unpacking works less well when a task has few parts or lies far in the future, and warns that a detailed plan can make the bias worse. And the planning fallacy is about calendar time, not effort; people can guess the working hours fairly well and still finish late because life gets in between.

How to try it this week, on paper

  1. Keep an estimate log. For five days, write each task’s estimated time before you start, and the real time when you finish. Two columns on a card is enough.
  2. Find your ratio. On Friday, divide the real total by the estimated total. If it is 1.5, your gut runs one and a half times short. That is your own reference class.
  3. Ask “when did I last do something like this?” Before estimating anything new, write the last similar task and how long it actually took. Plan from that, then adjust.
  4. Plan backwards for anything with a deadline. Start at the due date and write the steps in reverse until you reach today. If they do not fit, you find out now.
  5. Plan the day from what you finish. Do not fill a day to the brim. Plan your day shows how to plan from what you usually get done, and how long will it take? goes deeper into estimating one task.

The related idea that work expands to fill the time you give it is Parkinson’s law, and the reason a day of ten “quick” jobs never fits is in how many tasks a day.

Where Dozen helps, and where it doesn’t

The best-tested fix is your own past record, and Dozen keeps one for the day as a whole. Your usual day is the median number of to-dos you finished on your 14 most recent working days. A bar on Today shows what is done, what is planned and what your usual day holds, so you can see a heavy day before you are in it. When Today holds more than your usual day, Lighten offers to move the few things that can wait, and nothing moves until you say so.

Where it does not help: The usual day appears once you have finished something on at least 5 different days. Dozen counts to-dos, not minutes, so the estimate log above stays on paper or in a spreadsheet. Dozen is not out yet. And it does not block time on a calendar, so if you plan hour by hour, a calendar does that better.

Questions and answers

Why do I underestimate how long things take?

When you predict your own task, you picture the plan: the steps going as intended. You leave out the interruptions, the waiting and the parts you forgot, and you explain past delays as one-offs. Other people predicting your task do not do this, which is why they tend to guess longer.

How much extra time should I add?

There is no universal multiplier. The better answer is your own record: if your last three similar tasks took about one and a half times your estimate, plan for that. Keeping a simple log of estimates and actual times gives you the number.

Does breaking a task into steps help with estimates?

Sometimes. One study found that listing the parts of a task lengthened predictions and reduced the bias, but other studies found similar strategies did not help, and a very detailed plan can even make people more optimistic. Comparing with past tasks is more reliable.

Sources

What this page says about Dozen was checked against the app on .

Last checked:

Get Dozen

Dozen is free to download. Dozen Pro, a monthly or yearly subscription, adds the Assistant.

Coming soon to iPhone and iPad

Dozen: Tasks & Notes

Dozen is free to download. Dozen Pro, a monthly or yearly subscription, adds the Assistant.

Coming soon to iPhone and iPad