The Planning Fallacy and the Outside View

Definition

The "inside view" forecasts an outcome by reasoning from the specifics of the plan at hand — its steps, its team, the obstacles currently visible. The "outside view" instead asks: how did similar projects, undertaken by comparable teams, actually turn out? Kahneman and Tversky found the inside view reliably produces the planning fallacy: forecasts "closer to a best-case scenario than to a realistic assessment," because the inside view can only extrapolate from a plan going right and has no way to imagine "the strange sequence of unlikely events" that derail most real projects.

In the Book

Kahneman recounts leading a team to write an Israeli high-school judgment-and-decision-making curriculum. A year in, with good visible progress, he privately polled the team on how much longer completion would take — estimates clustered around two years. He then asked Seymour Fox, the team's curriculum-development expert, to think of comparable teams starting from a similar stage: Fox realized, to his own visible surprise, that roughly 40% of such teams never finished at all, and that the ones that did took seven to ten years — despite Fox's own prior estimate having been inside the optimistic two-year consensus. The team, confronted with this outside-view data, nonetheless "carried on as if nothing had happened" (Kahneman calls this third finding "irrational perseverance") — and the book was in fact finished eight years later, by which point Kahneman had left the team and the ministry's interest had evaporated so the finished curriculum was never used.

Why It Matters

The planning fallacy explains why cost and time overruns are the norm rather than the exception across construction, software, publishing, and personal projects — the inside view feels rigorous because it's detailed, but detail about a specific plan tells you nothing about the base rate of plans like it failing or running long. The outside view is a generalizable corrective for any forecast: before trusting a plan's own logic, find the reference class of comparable past cases and anchor on how those actually went, not on how this one is supposed to go.