Estimate in ranges, not point estimates

Instead of "my estimate is 500," say "I think it is between 200 and 2000."

Why it works

Point estimates communicate false precision and hide the uncertainty that is genuinely present. A range estimate communicates calibrated confidence: the width of the range reflects how much you actually know. Experts in forecasting — including superforecasters — consistently outperform novices partly by expressing uncertainty as ranges rather than collapsing to point guesses.

How to do it

1. After computing a central estimate, ask: under what conditions would the real answer be 10x higher? 10x lower?
2. State a range covering "reasonably plausible outcomes" — not the absolute worst case but the plausible spread.
3. Use geometric rather than arithmetic ranges for quantities spanning orders of magnitude (e.g., 100–10,000, not 100–900).
4. Communicate the range and your confidence in it, not just the point.

Evidence

Calibration research shows that expressing estimates as ranges and tracking their accuracy over time substantially improves calibration. Superforecasters routinely use confidence intervals and narrow them with evidence. (observational)

People tend to produce overconfident ranges (too narrow); training can widen them to appropriate calibration but requires feedback over many predictions.

Sources

• Tetlock & Gardner (2015), Superforecasting — range-based probabilistic estimates and calibration

Common mistake

Giving a range so wide it is uninformative (between 1 and a billion), which defeats the purpose of estimation by avoiding commitment.