The planning fallacy, and how teams beat it
2026-10-21 · The Scrums team
In 1979, Kahneman and Tversky named our habit of underestimating how long things take, and Douglas Hofstadter joked that it always takes longer, even when you allow for that. Teams can do better than individuals.
In 1979, the psychologists Daniel Kahneman and Amos Tversky gave a name to something everyone who has planned a project already knew: we consistently underestimate how long tasks will take, even when we know that similar tasks took longer in the past. They called it the planning fallacy.
The same year, Douglas Hofstadter put it more memorably in Gödel, Escher, Bach: "It always takes longer than you expect, even when you take into account Hofstadter's Law."
Why individuals get it wrong
When we estimate alone, we imagine the task going to plan: the steps we know about, done one after another. We don't imagine the review that takes two days, the dependency that changes, or the bug in the code next door, because we can't. The surprises are, by definition, the parts we didn't see.
Why teams can do better
A team can't see the future either, but its members see different parts of it. The tester knows which areas always break. The person who built the payments code knows how fragile it is. Estimating together, the right way, puts those views side by side:
- Estimate independently, then compare. If everyone says their number out loud in turn, the first one anchors the rest. Planning poker keeps them hidden until they're all in.
- Treat disagreement as information. When one person says 3 and another says 13, they're picturing different work. Ask why before you settle.
- Estimate relative size, not hours. "Twice as big as that story we did last month" is easier to get right than "eleven hours".
- Estimate risk separately. Small doesn't mean safe.
- Look back at what happened. Compare the plan with what actually happened, and the next plan improves.
In Scrums
Sprint poker hides every vote until the reveal, then shows the spread, the average, the median and how much the team agrees, so the outliers get their minute. A split vote is one click from a Re-vote. Each story can be estimated on up to four dimensions, such as story points and risk, each with its own deck. And the Plan vs. Actual retro template is built for the look back: what you planned, what happened, and why they differed.
It will still take longer than you expect. But by less. Try sprint poker in your browser, no account needed.
More on planning poker in Scrums, and the Plan vs. Actual retro format.