Superforecasting — A 5-Step Prediction Framework

Philip Tetlock’s superforecasting research distilled into a repeatable process for making better predictions about uncertain outcomes.

The 5-step framework

1. Break the question down — decompose big fuzzy questions into smaller, answerable sub-questions
2. Separate known from unknown — for each sub-question, identify what you can estimate vs. what is genuinely uncertain. Even “impossible” questions become tractable when decomposed far enough
3. Adopt the outside view — “nothing is 100% unique.” Find base rates and historical priors that anchor your estimate. What has happened in similar situations before?
4. Take the inside view — now examine what is genuinely unique about this situation. What factors make it different from the reference class?
5. Synthesize with weights — combine outside and inside views, weighting different pieces of information by reliability and relevance, to produce a well-calibrated prediction

Why this matters

The framework fights two failure modes: anchoring too heavily on a single narrative (pure inside view) and ignoring context that makes a situation genuinely different (pure outside view). The synthesis step is where judgment lives.

Connects to decision-making, mental models, analytics craft.

Open questions

• How do you calibrate the weighting between inside and outside views in practice?
• Could this framework be formalized into an analytical checklist for business decisions?