The principle
Find the one thing that must be true. Test that first.
State the uncertainty
Write down the assumption that could invalidate the project: data access, decision quality, user reliance, or value.
Make a small test
Use a representative sample and a baseline. Keep the experiment narrow enough to learn from its failure.
Decide in advance
Specify what evidence would support proceeding, changing direction, or stopping.
Take it to your team
Three questions to start with.
- What must be true for this to work?
- What is the smallest honest test?
- What result would change your mind?