Taking action
When is a small test the wrong tool for a decision?
A small test is the wrong tool when the decision cannot safely be made by trial: a meaningful downside is not reversible; consent, permission, law, policy, or professional safeguards set a boundary; requirements are fixed and leave no room to iterate; the observation cannot answer the actual decision; or a small evidence set would be too weak for the stakes. Use the BRAKE screen before testing. If any boundary fails, switch to existing evidence, qualified guidance, formal evaluation, or direct planning instead of experimenting.
Do not test a boundary that already decides the method
A small test is useful only when there is genuine room to learn safely and change course. It is the wrong tool when a trial would cross a non-negotiable boundary, expose someone to material harm, produce evidence too weak for the decision, or test something that fixed requirements already settle. In those cases, choose the required process, stronger evidence, qualified guidance, or a direct plan.
The 2026 Test and Learn annex says iteration is most useful under uncertainty when adaptation is possible. It is less useful when the problem and solution are already understood, requirements are fixed, or delivery must follow a fully planned sequence. It also treats testing as preparation for robust evaluation, not a replacement for it.
Run the BRAKE screen
BRAKE is a Forever Free Compass framework, not an external research finding. Apply it before you write a hypothesis, duration, or success measure.
- Boundaries and authority: identify consent, permission, law, policy, contract, safety, confidentiality, or professional requirements. If the test would bypass one, stop and use the authorized route.
- Reversibility and exposure: name the plausible downside, who carries it, whether it can be contained, and what cannot be undone. If the exposure is material or not yours to accept, do not experiment informally.
- Answerability: write the exact decision and the observation the test could produce. If that observation cannot distinguish the relevant options, use a different method.
- Known or fixed route: check whether credible existing evidence, a settled requirement, or a tightly sequenced process already determines what must happen. If so, plan execution rather than retesting the route.
- Evidence matched to the decision: ask whether the proposed evidence is strong enough for the consequence of being wrong. If not, narrow the decision or escalate the evidence method.
Separate uncertainty from room to experiment
Uncertainty alone does not make a test appropriate. You also need a changeable feature, an observable result, permission to act, a safe stopping point, and a decision that the result could actually inform. A tightly sequenced requirement may still contain uncertainty, but there may be no responsible room to improvise during delivery.
The Test and Learn annex explicitly distinguishes settings that benefit from iterative feedback from settings with fixed requirements or little scope for iteration. The practical transfer is simple: when the route is fixed, spend effort on planning, assurance, or implementation rather than inventing variation for its own sake.
Match the evidence to the size of the decision
A reversible, low-cost next step can sometimes use a modest observation. A decision involving a broad commitment, public claim, another person's welfare, or a material loss needs evidence and assurance proportionate to those stakes. Do not shrink the evidence method while leaving the consequence unchanged.
The Cabinet Office playbook says a test plan should name risks, ethical considerations, and the decisions that results will support, and that the planned evidence should match the size of the decision. GovS 010 similarly requires analysis to be sufficient for the question, fit for purpose, and proportionate to uncertainty, risk, value, and impact.
Treat consent and safeguards as design limits, not test variables
Do not test whether another person will tolerate pressure, disclosure, risk, or an unapproved change. Consent must be informed and voluntary; permissions and safeguards apply before the experiment. If a person cannot be protected, the method is wrong even if the possible learning seems useful.
The Cabinet Office research-safety guidance requires risks and potential harms to be assessed, consent and data handling to be planned, and a better method to be chosen when participants cannot be protected. For personal decisions, the relevant lesson is not to imitate formal research. It is to recognize that curiosity never overrides another person's authority, privacy, or safety.
Choose the replacement method deliberately
A failed BRAKE screen does not mean doing nothing. It tells you why a test is unsuitable and therefore what kind of next step fits better.
| Why the test fails | Use instead | Immediate next action |
|---|---|---|
| A rule, permission, or safety boundary governs the choice | Required process or qualified guidance | Identify the authority and obtain the applicable requirement before acting |
| The downside is material or not meaningfully reversible | Risk reduction, formal review, or a smaller decision | Reduce exposure or move the decision to the person with authority and expertise |
| The observation would not answer the real decision | Targeted research or clearer decision framing | Write the decision, then identify the information that could actually change it |
| The route is fixed or already well understood | Direct planning and implementation assurance | Translate the known requirement into sequence, ownership, and checks |
| The evidence would be too weak for the stakes | Stronger analysis or formal evaluation | Match the method and assurance level to the consequence of error |
Keep informal tests within low-stakes authority
BRAKE does not certify that a test is safe, ethical, legal, valid, or sufficient. It is a reflection aid for recognizing when an informal experiment should not carry the decision. Do not use it to bypass clinical, legal, financial, employment, housing, safeguarding, privacy, or other professional requirements.
When the screen passes, the next step is still modest: define one focused learning question, pre-commit the evidence and decision rule, keep the action bounded, and record limits. When it fails, respect the failure. Choosing a better method is progress, not reluctance.
Original contribution
The Forever Free Compass take
Forever Free Compass uses BRAKE as an original test-fit screen: Boundaries and authority; Reversibility and exposure; Answerability; Known or fixed route; Evidence matched to the decision. BRAKE asks whether trial is an appropriate way to learn before helping design the trial itself. A failed screen redirects the reader to a safer or more decisive method. It is not a validated risk assessment, research protocol, legal or clinical screen, evaluation standard, or substitute for consent, required processes, or qualified guidance.
Sources
- Test and Learn — UK Government Evaluation Task Force. When iterative testing is and is not valuable; fixed requirements; scope for adaptation; proportionality to risk and potential harm; separation from robust evaluation.
- Testing, learning and iterating (phase 4) — UK Cabinet Office. Test objectives; risks and ethical considerations; advance decision criteria; evidence matched to the size of the decision; pause and stop branches.
- Conducting research safely — UK Cabinet Office. Risk and harm assessment; safeguarding; informed consent; ethics; data protection; choosing a better method when people cannot be protected.
- Government functional standard GovS 010: Analysis — UK Government Analysis Function. Fit-for-purpose analysis; proportionality to uncertainty, risk, value, and impact; sufficient design for the question; validation and assurance.
This is educational reflection, not medical or mental-health care. If distress is persistent, severe, or involves immediate safety, contact a qualified professional or local emergency service.