State Each Hypothesis So It Can Come Out Wrong

Acceptance Hypothesis template and examples

The chapter “Design Every Test to Have a Chance of Failing” explains the shape.

For [ICP or actor] experiencing [trigger or problem], if [promise or system condition] is true, then under [realistic conditions] we expect to observe [behavior or outcome] without [unacceptable rescue or distortion].

Fill-in

Part	Your answer
ICP or actor
Trigger or problem
Promise or system condition
Realistic conditions
Behavior or outcome, with a threshold
Sample size: how many participants
Refutation rule: the result you will treat as false, written before the test
Rescue or distortion you will not accept
The observation that would prove this wrong

Part

Your answer

ICP or actor

Trigger or problem

Promise or system condition

Realistic conditions

Behavior or outcome, with a threshold

Sample size: how many participants

Refutation rule: the result you will treat as false, written before the test

Rescue or distortion you will not accept

The observation that would prove this wrong

Weak and testable versions

Weak	Testable
Users will like the new onboarding.	Practice managers at groups of ten to thirty physicians will reach their first completed intake within one working day, with no call to support.
The product is easy to use.	Ten intended users, given no instruction, will complete the core task with no more than one navigation hint each.
Customers will renew.	Accounts that reached the first useful action in week one will still use it weekly in month four, and at least half will add a second user unprompted.

Weak

Testable

Users will like the new onboarding.

Practice managers at groups of ten to thirty physicians will reach their first completed intake within one working day, with no call to support.

The product is easy to use.

Ten intended users, given no instruction, will complete the core task with no more than one navigation hint each.

Customers will renew.

Accounts that reached the first useful action in week one will still use it weekly in month four, and at least half will add a second user unprompted.

Checks

Does it name an actor, a condition, a behavior, a threshold, a sample size, and a refutation rule?
Could a result you can imagine make it false? Set the threshold before the test from three questions: what is the smallest result that would change the decision, how far from your expectation would surprise you, and which mistake costs more, keeping a false claim or dropping a true one.
Have you written the competing explanations before the test?