Taking action
What is the difference between testing an assumption and evaluating an outcome?
Testing an assumption asks whether one uncertain link, condition, or mechanism is plausible enough to guide the next step. Evaluating an outcome asks whether the intended change occurred, for whom, under what conditions, and—when the claim requires it—whether the action contributed to that change. An assumption test can refine a plan without demonstrating the final outcome. Name the link under test, the observation it could produce, the decision it can inform, and the conclusion it cannot support. Use stronger outcome evaluation when the decision or claim depends on the result itself.
Ask whether you are checking a link or judging the result
Testing an assumption examines one uncertain part of the story: whether people will engage, whether a step can be delivered, whether a mechanism behaves as expected, or whether a necessary condition is present. Evaluating an outcome examines the intended change itself and the evidence needed to describe what happened, for whom, in what context, and with what confidence.
The 2026 Test and Learn guidance says focused tests can examine critical assumptions and early mechanisms before more robust evaluation. The 2026 Magenta Book defines evaluation more broadly as systematic assessment of design, implementation, and outcomes. A focused test may strengthen a plan; it does not automatically answer whether the whole approach achieved its intended outcome.
Use the BRIDGE–OUTCOME split
BRIDGE–OUTCOME is a Forever Free Compass framework, not an external research finding. Use the left side to bound a focused test and the right side to recognize when the question requires broader evaluation.
- Belief: state the one assumption or causal link you are uncertain about.
- Result at the link: describe what would be different locally if that assumption held.
- Indicator: name an observable signal that is close to that link, not merely completion of the activity.
- Decision: write the narrow next choice the observation could inform.
- Generalization limit: state the people, setting, duration, and conditions the observation does not cover.
- Evidence fit: match the method to the consequence of being wrong, then switch to OUTCOME evaluation when the intended change itself carries the decision.
Keep the causal pathway visible
The Government Analysis Function's Theory of Change toolkit recommends making the links from activities to outputs and outcomes explicit, along with their assumptions, evidence, risks, and context. That makes it possible to test one link without pretending the entire pathway has been confirmed.
Suppose a person is considering a weekly workshop to improve a practical skill. A focused test might ask whether the workshop format produces usable practice and feedback. An outcome evaluation would ask whether the skill improved under relevant conditions over a suitable period, while considering baseline, context, other influences, and the strength of conclusion needed. The first question can inform whether to continue or redesign the workshop; it cannot by itself demonstrate the final outcome.
Match the question to the evidence
The Test and Learn annex advises selecting methods proportionately to the assumption, decision, risk, time, and resources. It also says early mechanism evidence should be followed by more robust evaluation when the claim or decision concerns broader effects. The useful transfer to a low-stakes personal choice is to let the literal question determine the evidence rather than choosing a convenient activity first.
If the question is whether a reminder is noticeable, a short usability check may fit. If the question is whether the reminder caused a durable change, the evidence burden is different. Broader claims may require a defined outcome, a baseline or comparison, repeated measurement, attention to alternative explanations, and a method capable of supporting causal language.
| Question | Appropriate label | What it may inform |
|---|---|---|
| Can the step be delivered as planned? | Feasibility or implementation assumption test | Revise, continue, or stop the delivery approach |
| Does the expected short-term mechanism appear? | Mechanism or pathway test | Investigate the link further or prepare stronger evaluation |
| Did the intended change occur? | Outcome evaluation question | Describe outcome evidence within the method's limits |
| Did this action cause the change? | Impact or causal evaluation question | Support attribution only when the design can address alternatives and the counterfactual |
Do not promote a local observation into a whole-outcome claim
A positive assumption test can show that one barrier was absent or one mechanism is plausible. Other links may still fail. The intended outcome may take longer, vary across conditions, or be shaped by outside events. Record the narrow observation and the next question instead of treating a passed link as proof that the whole approach works.
The Magenta Book separates process questions about delivery and design from outcome and impact questions about what changed and whether change can be attributed to the intervention. It also distinguishes emerging evidence used during implementation from more conclusive assessment after implementation. The label should follow the question and method, not the reader's enthusiasm for the result.
Know when to move from BRIDGE to OUTCOME
Move to an OUTCOME question when the next decision depends on the intended change rather than on one local link; when you want to make a public, durable, or causal claim; when effects may differ across people or settings; or when unintended consequences and cost matter to the decision. Define the outcome, baseline or comparison, horizon, context, relevant groups, and conclusion strength before collecting evidence.
For ordinary low-stakes reflection, BRIDGE may be enough to choose one bounded next action. It does not certify safety, effectiveness, causation, or generalisability. Do not use an informal test to bypass consent, confidentiality, professional duties, formal research requirements, or qualified guidance for health, legal, financial, employment, housing, safeguarding, or other consequential decisions.
Original contribution
The Forever Free Compass take
Forever Free Compass uses the BRIDGE–OUTCOME split as an original evidence-scope card. BRIDGE names one Belief or causal link, the Result expected at that link, the Indicator to observe, the Decision the observation can inform, the Generalization limit, and the Evidence strength required. OUTCOME names the intended change, relevant people and conditions, comparison or baseline, timing, unintended effects, and the strength of conclusion sought. The split prevents a focused link test from silently becoming a whole-outcome claim. It is not a validated evaluation, research, causal-inference, statistical, or decision method.
Sources
- Test and Learn — UK Government Evaluation Task Force. Focused testing of assumptions and mechanisms; early signals; proportionate method choice; the relationship between iterative testing and robust evaluation.
- Magenta Book: Central Government guidance on evaluation — HM Treasury and UK Government Evaluation Task Force. Evaluation of design, implementation, and outcomes; process, outcome, and impact questions; attribution; context; emerging versus conclusive evidence.
- The Theory of Change Process – Guidance for Outcome Delivery Plans — UK Government Analysis Function. Causal pathways from activities to outcomes; explicit assumptions; evidence strength; risks; context; indicators; monitoring and evaluation planning.
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.