Experiment reviews
Write the review rule before the result
A reusable review card for deciding when to continue, change, pause or stop a small experiment without moving the goalposts after seeing the result.
A small experiment needs an ending you can recognize. Otherwise, an encouraging moment can become a reason to keep going, and every disappointing moment can become a reason to try just one more version. Write a short review rule while the result is still unknown.
The card below is an editorial planning method for a small, reversible project. It connects an observation to the next commitment. It is not a statistical stopping procedure or a claim that a few successful attempts establish general effectiveness. Use it after defining a question and observation method in the Small Steps Planner.
Name the commitment under review
Begin with the next action you might approve. A fictional club is considering a printed inventory card for a shared craft drawer. Its next commitment is one more supervised trial with a revised card, not installing the arrangement permanently or claiming it works for every club.
Write a maximum commitment beside that action. In this example, the club sets aside two preparation hours across no more than two card versions. That limit is a choice made for the invented situation. For your project, use the time, materials and attention you can actually spare.
GOV.UK's alpha guidance explicitly allows teams to proceed, repeat an earlier phase or stop when the evidence does not justify the next stage. This worksheet adapts that decision point to a much smaller project; it does not reproduce the government's service assessment process.
Write three branches in advance
Continue: state what would justify the bounded next action. The invented club chooses: at least five of six agreed task attempts locate the requested item without a spoken explanation, and the card remains accurate during the session. This is a local planning threshold, not a universal benchmark.
Change: name a specific problem a revision could address. If the inventory stays accurate but people misunderstand a category label, revise that label once and repeat the same task. Describe the suspected mechanism so a new version has a reason to exist.
Stop: name a condition that defeats the approach or exhausts the agreed commitment. If normal use makes the inventory inaccurate before it can reasonably be maintained, stop testing this card format. Also stop when the preparation limit is reached, unless a separate decision explicitly funds another round.
Give missing evidence its own outcome
A cancelled observation or an incomplete note is not a successful task and does not automatically mean the idea failed. Add an insufficient-evidence branch. Record what is missing, whether obtaining it still fits the commitment, and who will decide what happens next.
For the fictional card, three complete attempts out of six planned attempts leave the review incomplete. The club could finish the remaining observations within the same limit, or pause the project. It should not silently shrink the agreed denominator until the result qualifies.
Keep a practical stop condition available throughout the test. A problem that prevents the task from being completed calls for attention when it occurs; the review date is not a reason to carry on with an unusable arrangement.
Copy the review card
Question and version: What uncertainty are we examining, and what exact arrangement will be shown? Next commitment: What limited action could this review authorize?
Evidence required: Which attempts, conditions and notes must be present before we decide? Continue rule: What observation supports the next commitment? Change rule: What specific pattern supports one named revision?
Stop rule: Which condition ends this approach? Commitment limit: What is the maximum preparation or material budget? Missing-evidence rule: When do we finish the observation, pause or close the project?
Review record: What happened, which branch applies, what remains uncertain, and what action follows? Rule changes: If a definition changed, preserve its old wording, the reason and the point at which the change was made.
Work through an invented review
Suppose the fictional record contains six complete attempts: four locate the item independently, while two need help interpreting the same category. The inventory remains accurate. Preparation has used seventy minutes of the two-hour allowance. These are invented numbers for practicing the decision.
The continue branch does not apply because only four attempts meet its definition. The change branch does apply: revise the confusing category once, identify the new card as version two, and repeat the agreed task within the remaining allowance. Keep the first record alongside the second.
Now change one fictional fact: normal borrowing repeatedly makes the card inaccurate. That triggers the stop branch even if five people happened to find their item. A favorable main count does not erase a condition that was written to protect the usefulness of the whole arrangement.
Preserve the rule and the conclusion
Sometimes the first rule turns out to be poorly chosen. You can improve it, but identify the new rule as a revision rather than applying it backward and calling the original test successful. State whether another observation is needed under the revised definition.
Save the card with the decision note. Use honest measurement guidance when a rate or comparison needs clearer boundaries. Then put only the next justified action into the Small Steps Planner.