Convert a noisy research finding into a decision-ready brief without laundering uncertainty into a recommendation.
A useful one-page brief does not compress claim, evidence and recommendation into one confident paragraph. It preserves the handoffs between them. The claim is what can be checked; the evidence block says how well it is supported; the decision block states what action is available under present constraints. GAO treats evidence use as a cycle of planning, assessment, decision and learning rather than a document-production exercise [1]. ONS likewise defines quality as fitness for purpose across relevance, accuracy, timeliness, clarity and comparability [2]. The brief below makes those tests operational without pretending that a source automatically dictates a decision.
Build three boxes, not a blended narrative
Box one is the claim. Write a single falsifiable sentence with subject, measure, population, period and comparison. Add a claim type: fact, estimate, signal, forecast, opinion or marketing. If any noun hides a denominator—users, deployments, revenue, failures—define it beside the sentence. Do not add implications here. A reader should be able to disagree with the interpretation while still seeing exactly what the evidence is alleged to establish.
Box two is the evidence ledger. List the originating records, not every page that repeated them. For each, state what it directly supports, the reference period, method, relevant limitation and whether another source is genuinely independent. Then assign a status such as triangulated, single-source, contradicted or unverified. ASA's rule is a useful discipline even outside advertising: hold documentary evidence for objective claims before circulating them, and do not use qualifications to contradict the headline impression [3].
Make the decision conditional
Box three begins with the decision, not a restatement of the evidence. Name the owner, deadline, reversible options and cost of waiting. Translate the evidence status into a decision posture: proceed, run a bounded test, wait for a named observation, or decline. A weak evidence base can still support a reversible discovery step; it should not quietly support an irreversible commitment. Separate business appetite from epistemic confidence so a risk-tolerant choice is not misreported as a high-confidence finding.
Add a decision rule that another person could apply. For example: proceed to a six-week workflow test if two independent customer teams complete the task under the agreed latency and review limits; otherwise retain the current process. This is deliberately narrower than 'the technology works.' Record which result would change the recommendation and when it will be checked. The brief then becomes a control surface for action, not a persuasive memo frozen at the moment it was written.
Use a hypothetical worked example
Hypothetical example: a product team claims that an assistant reduces research time by 30%. The claim box specifies median elapsed time for a defined task, among eligible analysts, during a four-week pilot. The evidence box shows 12 participants, no control group, two abandoned tasks and logs available for nine completions. The 30% is labelled a pilot estimate, not a population fact. A vendor benchmark is listed as contextual because its task and users differ.
The decision box does not approve organization-wide rollout. It authorizes a second test with a predeclared baseline, at least two real workstreams, quality review blind to treatment, and an error-cost ceiling. The trigger is operational: advance only if median cycle time improves by at least 15% while material correction hours do not rise. Missing logs count as missing, not as zero time. The owner and check date are named. One page now carries both momentum and restraint.
- Claim: one checkable sentence, definition, period and claim type.
- Evidence: origins, independence, method, limitations and status.
- Decision: owner, reversible action, threshold, deadline and reversal condition.
Take it into the meeting
- Keep the checkable claim separate from interpretation and action.
- Match commitment size to evidence strength and reversibility.
- Write the next observation and decision rule before circulation.
Sources & boundaries
Source statements are attributed; the decision process is Signal Atlas analysis. Examples marked hypothetical are teaching inputs, not observed outcomes.
- The format disciplines reasoning but does not validate the underlying sources.
- A one-page constraint may be unsuitable for high-stakes legal, clinical or safety decisions requiring specialist review.
- Decision thresholds remain organizational judgments and should not be presented as universal standards.
- Evidence-Based Policymaking: Practices to Help Manage and Assess the Results of Federal EffortsU.S. Government Accountability Office · Source publication: 2023-07-12 · Retrieved 2026-09-19
Evidence use links planning, evidence building, decisions and continuous learning. Evidence can include performance information, evaluations, statistics and other analysis.
- Quality definedOffice for National Statistics · Source publication: 2020-10-08 · Retrieved 2026-09-19
Statistical quality is fitness for purpose across relevance, accuracy and reliability, timeliness, clarity, and coherence and comparability.
- 03 Misleading advertisingAdvertising Standards Authority and Committee of Advertising Practice · Source publication: not established · Retrieved 2026-09-19
Objective claims require documentary substantiation before publication. Qualifications must not contradict the claim they qualify or omit significant limitations.