Founders need conviction about a direction and honesty about the route. Without a testable claim, almost any result can justify continuing. Equally, endless testing can postpone a decision for which you already have enough evidence. Useful learning changes what you will do.
Use this principle when: the team disagrees about a bet, a roadmap expands, a result is being presented as validation or further research keeps delaying action.
What to do
- List what must be true for the next commitment to work. Include customer behaviour, technical performance, access, cost and timing. Prioritise by decision impact, uncertainty and the cost of obtaining useful evidence.
- State the claim and an alternative explanation. “Users like our demo” could mean they value the outcome, enjoy novelty or want access to your team. Design the test to distinguish explanations that matter.
- Write the expected result, acceptance criteria and action branches before observing the outcome. Use realistic conditions, retain failures and ask someone to challenge the interpretation.
Evidence that changes the decision
Proceed: the critical claim survives a relevant test strongly enough to justify the proposed commitment. State what remains unproven; passing one test does not validate the whole company.
Change the plan: the outcome is mixed or the test did not measure the claim. Repair the test, narrow the use case or investigate an alternative explanation. Give the follow-up its own budget and deadline.
Stop this route: a necessary assumption fails a credible test and no affordable, plausible change can repair it. Sunk effort is not new evidence.
Where founders misapply it
Do not optimise easy metrics while avoiding a difficult existential question. In small B2B samples, precise percentages can disguise weak evidence. Combine observed behaviour, technical results and explicit commitments without pretending they are a large statistical sample. Some questions require expensive trials; “smallest useful test” does not mean “cheapest activity”.
Camuffo et al.’s randomised studies support more disciplined search and idea termination. Research on A/B testing informs digital experimentation; transfer to other settings requires judgement. Neither validates every template in this guide. [1][3]
Connect it to the whole venture
BOM’s Plan building block connects direction, goals, strategy, investment and measurement. Make the investigation feed a specific roadmap decision: identify which customer, product or team milestone would change, which resources would move and who updates the plan after the result. Keep the plan responsive to evidence. [18]
Next decision: complete: “If we learned that …, we would change …”. If the second blank stays empty, reconsider the test.