The smallest meaningful test

• test
• experiment
• validate
• try
• prove
• pilot
• hypothesis

• What single belief is this test meant to check?
• What is the cheapest artefact that can check it?
• What result would change the decision?
• How long before the answer is readable?

• the belief stated as a sentence
• the measure that would confirm or deny it
• the sample needed for the difference to be visible

• A test too small to read
• A test that cannot fail

• If
  No result would change the decision
  Then
  Do not run the test. Make the decision.
• If
  The test requires building first
  Then
  Reduce it: a conversation, a page or a priced proposal usually suffices.
• If
  The sample is too small to detect the difference sought
  Then
  Enlarge the sample or accept the test only detects a large effect.

Write the belief, the measure, the sample, the success rule and the stop rule on one page before doing anything.

Read after 21 days.

• All five are written before execution
• The result is unambiguous when it arrives

• The success rule cannot be written — the belief is too vague to test

• The success rule is met → Scale one step and re-measure.
• The stop rule is met → Stop, and record the finding as a company fact.

Aury can carry out: prepare document, create task, file memory. Specialists involved: strategy, research.

Your approval: Success and stop rules are agreed with the founder before execution begins.

Applies to any decision where evidence is obtainable cheaply.

Small tests detect large effects. They cannot settle marginal questions, and a null result is not proof of absence.

• Sourced operator experience: Experiment-design practice, standard across product and growth work
• Aurygine synthesis: Aurygine synthesis across founder journeys

Confidence: high. Last reviewed 2026-08-25. Your company's own recorded results override this playbook whenever the two disagree.