Taking action
How do I define what success looks like before a small experiment?
Define success as learning enough about one important uncertainty to make the next decision, within the time, cost, permission, and safety boundaries you set before starting. Name the action, the observable signal you will record, the range that would count as useful evidence, and what each result will make you do next. A small experiment can succeed by showing that an idea should stop or change; it does not have to confirm your preferred outcome. The MARK card—Main uncertainty, Action and observation, Range and guardrails, Keep or change rule—makes that definition inspectable.
Define a learning result, not a victory condition
A small experiment is useful when it reduces a decision-relevant uncertainty. That can happen when the preferred idea works, when it fails inside a fair test, or when the result exposes a question the original plan missed. Define success around the evidence needed for the next decision, not around making one outcome look good.
The UK government's 2026 Test and Learn guidance starts with a shared, measurable outcome, asks teams to agree decision criteria before testing, and recommends decisions to continue, adapt, or stop based on evidence. Adapted to an ordinary personal experiment, this means writing the learning question and review rule while you can still see results neutrally.
Write the objective as something you can observe
Replace vague aims such as “see whether design suits me” with an action and a record: “Complete two 30-minute interface critiques this week and record whether I can explain one design choice after each.” The action shows whether the test happened; the observation helps answer the uncertainty.
CDC guidance for learning objectives is designed for training, but its structure transfers carefully: describe an observable behavior, the conditions under which it occurs, the degree or standard, and a realistic time boundary. NIST's experimental-design handbook likewise says to write down and prioritise objectives before choosing factors, responses, and the design. Neither source supplies a universal target for a personal experiment; you must choose a proportionate signal that fits the decision.
Build the definition with the MARK card
MARK is a Forever Free Compass framework, not an external research finding. Complete all four fields before starting.
- Main uncertainty: write the one question whose answer could change your next decision.
- Action and observation: state what you will do and the simple evidence you will record, keeping performance separate from interpretation.
- Range and guardrails: define what would count as a useful signal and cap time, cost, exposure, permission, and effects on other people.
- Keep or change rule: state what would make you stop, adapt, repeat, or expand at the review point.
Use evidence ranges instead of a pass-fail trap
Many small tests cannot justify a precise numerical cutoff. Use plain-language ranges when that is more honest: clear positive signal, mixed or incomplete signal, clear negative signal, and boundary crossed. Define what each range means before the result appears.
A mixed result normally calls for interpretation or one changed variable, not an automatic yes or no. A crossed guardrail is different: if the agreed cost cap, consent condition, safety boundary, or impact on another person changes, pause the test rather than averaging that problem into an otherwise encouraging result.
| Result range | Example evidence | Prewritten response |
|---|---|---|
| Clear useful signal | The action happened under the stated conditions and produced the observation you needed | Keep the direction or design the next bounded test |
| Mixed or incomplete | The action happened, but the observation was inconsistent or the setup interfered | Change one design variable and repeat only if the boundary still holds |
| Clear negative signal | The relevant assumption was contradicted inside a fair, bounded attempt | Stop or choose a different direction |
| Guardrail crossed | Cost, permission, safety, or effect on another person moved outside the stated limit | Pause now and reassess before continuing |
Keep the record proportionate
Record only what the next decision needs. A low-stakes one-week experiment may need a date, whether the action happened, one observation, and one short note about context. More fields do not automatically create better evidence, and one observation does not prove a general rule about your ability or future.
Separate execution evidence from outcome evidence. If you never ran the action, you learned about the setup or follow-through but not necessarily the direction. If you ran it under different conditions than planned, label the deviation rather than quietly treating it as the intended test.
Match the review rule to the stakes
Use MARK for low-stakes, reversible exploration. Raise the evidence standard as cost, irreversibility, safety exposure, professional duty, or impact on other people increases. A short personal test cannot establish medical effectiveness, legal compliance, financial suitability, employment permission, or another person's consent.
For those decisions, follow applicable rules and obtain qualified guidance before acting. MARK can help you write questions for that conversation, but it cannot replace expertise, formal evaluation, consent, or required oversight.
Original contribution
The Forever Free Compass take
Forever Free Compass uses MARK as an original success-definition card: Main uncertainty, Action and observation, Range and guardrails, and Keep or change rule. Main uncertainty limits the experiment to one decision-relevant question. Action and observation separates what you will do from what you will record. Range and guardrails states what evidence would be useful while bounding time, cost, permission, safety, and effects on others. Keep or change connects the result to stop, adapt, repeat, or expand. MARK is a planning heuristic, not a validated assessment, statistical threshold, or promise that a small experiment will answer the larger life decision.
Sources
- Test and Learn — HM Treasury and Evaluation Task Force. Shared measurable outcomes; decision criteria set before testing; proportionate evidence; evidence-led continuation, adaptation, or stopping.
- Design Training: Learning Objectives — Centers for Disease Control and Prevention. Observable behavior, conditions, degree or standard, and specific, measurable, achievable, relevant, and time-bound objectives.
- NIST/SEMATECH e-Handbook of Statistical Methods: What are the objectives? — National Institute of Standards and Technology. Writing and prioritising experimental objectives before choosing factors, responses, and the experimental design.
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.