Taking action
How do I choose what evidence to record during a small experiment?
Record only the evidence that could change your next decision: whether the planned action happened, the observable signal tied to your learning question, the conditions that may have shaped it, and any unexpected cost or effect. Decide before starting when and how you will record each item. Use the FIELD note—Focused question, Implementation check, Evidence signal, Limits and context, Decision link—to keep the record small. One log entry is evidence about this attempt, not proof of a general pattern or a high-stakes outcome.
Start with the question, not with everything you could track
Write one learning question first: “Does this kind of task give me enough useful information to justify another bounded attempt?” Then ask what observation would help answer that question. A record earns its place only if it helps show whether the action happened, describes the relevant signal, explains a meaningful deviation, protects a boundary, or informs the next decision.
CDC's evidence-gathering guidance begins the same way at a larger scale: define the evidence needed for the evaluation question, including what will be collected, how, when, and from what source. Its indicator guidance also says the indicator should align with the question. Adapted to a personal experiment, this favors a few relevant observations over an impressive-looking pile of notes.
Separate whether you ran the test from what happened
Keep an implementation check apart from the evidence signal. “I completed the planned 30-minute critique” describes execution. “I identified two design choices and could explain one trade-off” describes what you observed. If the action did not happen, you may have learned about timing, access, or follow-through, but you did not run the intended test of the direction.
The CDC Program Evaluation Framework distinguishes activities and outputs from outcomes and recommends a data-collection strategy that names expectations, methods, indicators, measures, and sources. It also notes that no single method gives a complete picture. For a small personal attempt, that does not mean collecting more by default; it means naming what each observation can and cannot tell you.
Build a minimal record with the FIELD note
FIELD is a Forever Free Compass framework, not an external research finding. Complete the five fields before the attempt, then keep the entry short.
- Focused question: write the one uncertainty this attempt is meant to inform.
- Implementation check: record whether the planned action happened under the stated conditions.
- Evidence signal: name the observable note, count, artifact, or feedback that relates directly to the question.
- Limits and context: note missing data, changed conditions, guardrail issues, and unexpected positive or negative effects.
- Decision link: state whether this evidence could justify stopping, adapting, repeating, or taking one additional bounded step.
Use one row per attempt
A compact row makes planned evidence easier to distinguish from interpretation added afterward. Use ordinary language unless a numeric measure is genuinely useful and defensible.
| FIELD item | Example before the attempt | Example after the attempt |
|---|---|---|
| Focused question | Can I learn from critiquing a real interface? | Question unchanged |
| Implementation check | One 30-minute critique on Tuesday | Completed for 27 minutes with the chosen interface |
| Evidence signal | A written explanation of one design trade-off | One explanation written; one point still unclear |
| Limits and context | Stop if access or permission changes; note interruptions | No permission issue; one interruption; no unexpected cost |
| Decision link | Repeat once if the signal is useful but incomplete | Repeat with a quieter setup; do not infer a career conclusion |
Keep collection proportionate and record surprises
The 2026 UK Test and Learn guidance says methods should be chosen proportionately for the assumption, decision, time, and resources involved; more methods are not inherently better. It also recommends systematic evidence collection and early monitoring for unintended effects. For a low-stakes personal experiment, a short contemporaneous note, a simple artifact, or one piece of structured feedback may be enough.
Add one open field for what you did not expect. That keeps the planned signal from becoming a tunnel. Record a changed cost, access problem, effect on another person, surprising ease, or new question without forcing it into the success measure. If the method changes during the attempt, label the change rather than silently merging unlike observations.
Interpret the record at the level it can support
A single FIELD note can support a decision about the next bounded attempt. It normally cannot establish a stable trait, predict a long-term outcome, or show that the action caused every change you observed. Several consistent entries may strengthen a practical pattern, but repetition alone does not remove bias, confounding conditions, or measurement limits.
Use this process only for low-stakes, reversible exploration. Do not use a personal log to establish medical effectiveness, legal compliance, financial suitability, workplace permission, safety, or another person's consent. Raise the evidence standard with the stakes and obtain qualified guidance where rules, rights, health, money, employment, or significant effects on others are involved.
Original contribution
The Forever Free Compass take
Forever Free Compass uses the FIELD note as an original evidence-planning framework: Focused question, Implementation check, Evidence signal, Limits and context, and Decision link. Focused question states the uncertainty the attempt can inform. Implementation check records whether the planned action actually happened. Evidence signal captures the observation chosen before starting. Limits and context labels deviations, missing information, guardrail changes, and unexpected effects. Decision link states what this evidence is allowed to change next. FIELD is a low-stakes reflection aid, not a validated measurement instrument, causal evaluation, or substitute for qualified guidance.
Sources
- Step 4 – Gather Credible Evidence — Centers for Disease Control and Prevention. Evidence needs tied to evaluation questions; collection method and timing; relevant indicators; data sources, burden, relevance, and quality.
- CDC Program Evaluation Framework, 2024 — Centers for Disease Control and Prevention. Data-collection strategies; expectations, methods, indicators, measures, and sources; implementation and outcome distinctions; method limitations.
- Test and Learn — HM Treasury and Evaluation Task Force. Proportionate evidence methods; systematic learning; unintended-effect monitoring; advance decision criteria; documentation and adaptation.
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.