The outside view asks how things like this usually go

Two ways to answer the same question

Asked how long something will take, a person can reason in one of two directions. The inside view starts from the specifics: the steps involved, the people doing them, the obstacles that can be foreseen from here. It builds an estimate upwards out of a plan. The outside view ignores nearly all of that and asks a different question — of the cases resembling this one, how did they actually turn out?

The two methods do not merely differ; they disagree in a consistent direction. Inside-view estimates come in shorter, cheaper and smoother than what follows. The tendency is durable enough to have earned a name, the planning fallacy, and it has been demonstrated both in the laboratory with student assignments and in field records of construction, software and public works. What makes it unusual among famous psychological findings is that it survives being pointed out, which most do not.

Why the detailed method fails

A plan is a scenario, and a scenario has to be coherent to be imagined at all. Coherence is exactly what filters out the things that will actually cost you: the illness nobody scheduled, the supplier who stops replying, the approval that arrives late for a reason particular to that week. Each of those is individually improbable, which is why none of them appears in the plan, and collectively they are close to certain.

So the inside view produces something closer to a best case dressed as an expectation. It is not laziness. The more carefully somebody builds the plan, the more convincing the underestimate becomes, because the effort goes into the branches that can be foreseen rather than into the aggregate weight of the ones that cannot.

The reference class is where the judgement lives

Everything the outside view offers depends on choosing the right group of comparable cases, and this is not a mechanical step. Too narrow a class and you have three examples and no signal. Too broad and the average describes something that isn’t your situation at all. Kitchen renovations in general are a reasonable class for a kitchen renovation; building projects in general are not.

The honest procedure is to name the class out loud, state why those cases are comparable, and then resist adjusting the answer downwards on the grounds that your case is better run. That last move is where the outside view usually dies. The adjustment is always available, always feels justified, and reintroduces the entire bias in one step.

It is fair to adjust when you can point to a structural difference rather than an attitude — a smaller scope, an experienced team that has done this exact thing repeatedly, a dependency that has already been removed. Optimism isn’t a structural difference.

What the evidence supports

The underestimation of durations and costs in large projects is well documented in archival data, and some public-sector appraisal guidance now instructs analysts to apply an explicit uplift to early estimates precisely because the historical record shows they run low. That is a practical endorsement of the method, though it is worth being clear that archival evidence of this kind is observational: it establishes the pattern reliably and the causes only loosely.

Experimental work supports the basic asymmetry too, with people predicting their own completion times optimistically while predicting other people’s more accurately — which is itself a clue about the mechanism, since an observer has no privileged access to your plan and falls back on how such things usually go.

Where the outside view is the wrong tool

It has real limits. When the case genuinely is novel, no reference class exists, and manufacturing one produces false precision rather than humility. When the distribution of past outcomes is very wide, the historical average tells you almost nothing useful about any single instance. And when the past cases were all produced under conditions that no longer hold, the base you are borrowing has quietly expired.

There is also a failure mode where the outside view becomes an excuse. Told that projects of this type usually overrun, a team can stop trying to be the exception. The point of the method is to set the expectation correctly, not to license the outcome.

Using both, in the right order

The workable sequence is outside first, inside second. Get the reference-class number before you have fallen in love with the plan, because once the plan exists, every subsequent number is anchored to it. Then use the inside view to identify what specifically about this case would move you off that baseline, and write those reasons down where somebody else can read them.

The written record is the part that makes it a method rather than a mood. If the adjustments were wrong, you will only find out by comparing them later against what happened, and you cannot compare against a reason you never wrote down.