Agree before seeing the result

Set the evidence standard before the experiment begins. Otherwise enthusiasm and sunk cost will move the goalposts.

The standard can be a range. It should account for quality, adoption, and risk, not only speed. A process that becomes faster while producing more expensive mistakes did not improve.

Write what happens under each result:

• Below the lower boundary: stop or redesign.
• Inside the uncertain band: continue only to answer a named uncertainty.
• Above the decision boundary: prepare the investment case.
• Strong result plus repeatable adoption: consider industrialization.