← All articles

Estimation

The Planning Fallacy: Why Your Estimate Was Too Short Again

October 5, 2026

Sydney Opera House Under Construction in the mid-1960s 02

Image: Sydney Opera House Under Construction in the mid-1960s 02, My Grandfather. CC0, via Wikimedia Commons.

In 1957, the government of New South Wales picked a design for an opera house on Bennelong Point in Sydney Harbour. The plan said it would cost about 7 million Australian dollars and open in 1963. It opened in 1973. The final cost was about 102 million dollars.

The Sydney Opera House is now one of the most loved buildings in the world. It is also the textbook case of a mistake that you have probably made this month, on a much smaller scale.

The name for it

In 1979, the psychologists Daniel Kahneman and Amos Tversky gave this mistake a name: the planning fallacy. It describes our habit of predicting that a task will take less time than it does, even when we know that similar tasks took longer in the past.

The strange part is that last clause. We aren't ignorant of our history. We just don't use it when we plan.

The thesis study

In 1994, Roger Buehler, Dale Griffin and Michael Ross asked university students writing honors theses to predict when they would finish. On average, students predicted about 34 days. The average actual time was about 56 days. Even when asked for a worst-case estimate, most students still finished later than that.

The students weren't careless. When they thought about their own project, they pictured how the work would go: the chapters, the draft, the edits. They did not picture the flu, the lost file, the advisor who took two weeks to reply. Those things felt like exceptions. Taken together, the exceptions were the schedule.

Why it happens

The best explanation is that we plan from the inside. We build our estimate out of the steps we can imagine, and the steps we can imagine are the ones that go right. The surprises are, by definition, the ones we didn't see coming.

Think of it like packing for a trip by listing only what you plan to wear. You'll get the outfits right and still forget the charger, the umbrella and the thing you spilled coffee on. The list wasn't wrong, just incomplete in a predictable way.

We've written before about why deadlines feel like they arrive in dog years. The planning fallacy is the other half of that feeling. The deadline didn't move faster. The estimate was short from the start.

The fix: look outside

Kahneman later described the remedy as taking the "outside view." Instead of asking "how long will this take me," you ask "how long did tasks like this take, for me or for others?" The researcher Bent Flyvbjerg built that idea into a method called reference class forecasting, now used for large public projects. You find a group of similar past projects, look at how much they actually overran, and adjust your plan by that amount.

You can do a personal version in ten minutes:

  1. Find three past tasks that are similar to the one you're estimating. Same kind of work, roughly the same size.
  2. Write down how long each actually took, from start to done. Check your calendar or email if you're not sure.
  3. Compare to what you expected at the time, if you remember. Most people find a ratio somewhere between 1.3 and 2.
  4. Multiply your new estimate by your own ratio. That's your honest number.

You might be thinking this just makes every plan longer. Yes. Annoyingly so. But a long estimate you keep beats a short one you blow, and saying "not this month" in March is much easier than apologizing in April. Our guide to estimating work without lying to yourself covers the inside view, and buffer days covers where to put the extra time.

The limit

The outside view won't save a project whose scope keeps growing. If the task itself changes, your old ratio no longer applies. That's a different problem, and we cover it in scope creep.

Your next step: the next time somebody asks "when can you have it?", don't answer in the meeting. Go find three jobs like it from your own calendar, see how long they really took, and come back with that number. It'll be longer than you want. It'll also be the one you hit.

Sources

← All articles · Try the deadline converter