Direct answer
Core relationship: Link the change to a mechanism and write what should happen if the mechanism is correct. Include a comparison or baseline, define the measure in advance, and decide what each result will change.
For design small experiments for problem solving, begin with name the uncertainty and finish with write decision rules. The sequence keeps the starting facts, intermediate result, and final limit visible instead of relying on memory.
Method you can audit
Name the uncertainty
A broad goal such as improve service is not a testable question. Capture the result before moving to make one interpretable change. A written output shows whether the next deduction follows from the evidence or quietly adds a new assumption.
Make one interpretable change
Several simultaneous interventions make attribution difficult. Compare this output with the warning about pilot as rollout. If that failure still fits, revise the representation before using the result in another deduction.
Predefine outcome and window
Choose the measure and observation period before results appear. Use predefine outcome and window as a checkpoint. Name the input, the change made, and the conclusion that becomes available so another reader can reproduce the move.
Write decision rules
State how positive, negative, and unclear results affect the next action. Finish by stating what this step establishes and what it leaves undecided. That boundary prevents a valid local conclusion from being stretched beyond the question.
Check against pilot as rollout
Questions for make one interpretable change
- Can another reader reproduce name the uncertainty from the stated information?
- Did you complete make one interpretable change without reversing or enlarging the claim?
- Does the result survive the failure mode pilot as rollout?
- Which part remains uncertain after write decision rules?
This check is specific to the method above. A clear answer should expose the evidence chain and the point where a different fact would produce a different conclusion.
Worked interpretation
Name the uncertainty in the case
If reminder timing causes missed appointments, test two timing windows on comparable groups and measure attendance. A result matters only if the team has decided what difference would justify changing the standard reminder.
Why write decision rules matters
The case starts with name the uncertainty, then uses make one interpretable change to control the next inference. The conclusion is acceptable only after write decision rules. Changing any fact that supports those moves can change the answer, while an unrelated detail should not.
Limits and traps
Pilot as rollout
A smaller launch without a comparison may reveal operations but not causal effect. Return to name the uncertainty and rebuild only from information the problem actually provides.
Success metric shopping
Selecting the best-looking measure afterward weakens learning. Compare the disputed move with predefine outcome and window; the two should use the same definitions and direction.
No safety boundary
Even reversible tests need limits and monitoring. Run the practice task while deliberately watching for no safety boundary. Record the first point where the reasoning begins to drift.
Use it on a real problem
Turn one proposed solution into a one-week test plan with hypothesis, change, comparison, primary measure, stop rule, and next decision.
Continue after write decision rules
Use the root cause analysis guide for the broader method, then apply the idea in the root cause map. Keep the written reasoning so the next attempt can improve a specific step rather than simply repeat the exercise.
Compare this method with Use an Evidence Log for Competing Hypotheses, then continue to Run a Pre-Mortem Before an Important Decision for a different reasoning task.