Set decision thresholds before you start

The short answer

Before an experiment or commitment begins, define the minimum sample, success threshold, failure threshold, review date and exceptions. A threshold protects you from stopping after one bad day or continuing indefinitely because the evidence is uncomfortable.

Write the five-part threshold

1. 1. Decision Name the choice the evidence will inform.
2. 2. Minimum sample Specify how much exposure or time is needed before the signal is interpretable.
3. 3. Continue line Define the result strong enough to keep the current mechanism.
4. 4. Change or stop line Define the result weak enough to revise the mechanism or end the bet.
5. 5. Exception List any safety, cash or reputation event that overrides waiting for the full sample.

Avoid false precision

A threshold does not make weak data strong. Use ranges when sample sizes are small and document important differences in audience, timing or execution quality.

Do not move the threshold after seeing the result unless new information genuinely changes the cost of being wrong. If you revise it, keep the original visible.

Questions

What if I cannot estimate a minimum sample?

Choose the smallest sample that includes the normal sources of variation and label the decision provisional.

Should every goal have a stop rule?

Use stop or change rules for uncertain, costly or recurring approaches, not for simple one-off tasks.

Can I act before the review date?

Yes when a predefined exception occurs or the evidence becomes overwhelmingly clear.