Reference-Class Forecasting

Starting an estimate from the outside view: find a class of similar cases and use its base rate as the anchor, then adjust for what is genuinely specific about this case. It deliberately reverses the instinct to reason from the vivid details of the case in front of you, which is where forecasts go wrong.

Why it Matters

The inside view, built from the specifics of a situation, systematically ignores how comparable efforts actually turned out, producing optimistic and overconfident estimates. Grounding the forecast in a reference class first puts the realistic distribution of outcomes on the table before the details pull the number away from it.

Signals

• Estimates built entirely from this case's particulars, with no comparison to similar ones.
• Plans that assume the best case while ignoring the failure rate of comparable plans.
• Surprise that a project landed near the historical average it was always likely to.

Benefits

A realistic anchor drawn from how things usually go, a defense against the optimism of the inside view, and a starting point that detailed reasoning then refines rather than replaces.

Risks

A poorly chosen reference class anchors on the wrong base rate; over-adjusting back toward the inside view erases the correction; rare cases may have no usable class.

Tensions

Every case feels unique, and the specifics are real, so the outside view can feel like ignoring relevant information; the discipline is to start outside and adjust inward, not the reverse.

Examples

Estimating a project's duration from how long similar projects actually took before adjusting for this team's situation; forecasting a venture's odds from the base rate of comparable ventures, not its pitch.