Six methods for deciding under uncertainty — and when each one is worth it

Most bad decisions do not come from a lack of methods. They come from using the wrong one — or using one after the answer is already known. Here is the toolkit we use, with the trigger that justifies each method and what must be recorded 在 the outcome.

Horizon scanning

When: something may be shifting in the audience, technology, rules or the market. Record: the signal, when it was observed, its source, the possible consequence and the date you will check it.

Interviews about the user's job

When: it is unclear who needs the product and why. Record: the real job, today's workaround, what success looks like — and the person's consent to use the material.

Contradiction solving

When: a confirmed contradiction blocks the result (you need both speed and safety, say). Record: the contradiction, the options, and the cheapest test that tells them apart. It is not a ritual for every task.

Expert forecasts in rounds

When: a checkable question about the future that current data cannot answer. Record: independent participants, probabilities written before the outcome, two rounds, a resolution date — and afterwards a score against the base rate.

Experiments with a review

When: you must decide whether to keep a change. Record in advance: the metric, its denominator, the baseline, the time window and the stop rule. Afterwards: the decision.

Alternative checks

When: meaningful ambiguity remains after the deterministic checks. Record: evidence for and against, and where opinions diverge. Several reviewers agreeing is a signal to look closer, not a proof — reviewers who see the same inputs tend to make the same mistake.

The common rule: data first, method second, and anything that judges the result is written down before the result exists.