A reporting checklist for message-test results

Begin with an audit block

At the top of a result page, identify the study ID and version, decision question, sponsor and funder, field dates, geography and language, target population, sample or recruitment source, mode, achieved usable sample, primary outcome, and whether the work was pre-specified. Link the full methods, stimuli, questionnaire, aggregate data or codebook, and correction history.

If no sponsor paid for or influenced the work, say so. If a sponsor supplied claims, stimuli, or review comments, describe that role. A global sponsorship policy does not replace a proximate study-level disclosure.

Make design and provenance reproducible

Archive the exact stimulus for every condition, including images, qualifiers, disclosures, call to action, destination, device layout, and version hash. Publish the questionnaire in administered order with instructions, response options, randomization, display logic, and preceding context that could affect an answer.

Describe eligibility, recruitment, incentives, assignment, exposure rules, duplicate controls, quality screens, exclusions, missing data, weighting, coding, statistical model, multiplicity handling, and software version. Label every material deviation from the decision brief with its timing and rationale.

Show flow, denominators, and data quality

Report invitations when known, starts, screened out, assigned, exposed, completed, excluded, and analyzed. Give the reason and count for each exclusion. Denominators should travel with every percentage so readers can tell whether a subgroup or item has missing data.

Publish balance or implementation checks that were planned, assignment anomalies, timing distributions, duplicate rates, coding agreement where applicable, and sensitivity analyses. Do not hide an operational defect because the final estimate looks plausible.

Report estimates before labels

For each primary and guardrail outcome, report the estimate by condition, the contrast, uncertainty interval, analysis base, and unit. Distinguish exploratory subgroup or secondary findings from confirmatory outcomes and show all decision-relevant directions, including unfavorable and inconclusive results.

Avoid “winner,” “no difference,” or “representative” unless the design and decision rule justify the term. A non-significant result does not prove equivalence, and a small p-value does not measure practical importance, study quality, or the probability that a hypothesis is true. Explain the practical threshold and the action actually taken.

Worked result shell

A compact opening could read: “From September 10–12, 2026, 612 eligible opt-in panel members were randomly assigned to two fictional message versions. In this recruited sample, version B increased accurate unaided restatement by an estimated 8 percentage points, with the stated interval, while the false-automation guardrail moved by an estimated 1 point. The sample was not probability-selected, so the levels should not be generalized to all U.S. or Canadian adults.” Use actual values only after a real study; this sentence is a reporting illustration.

Follow with the participant flow, exact result table, sensitivity analysis, false or partial interpretations, decision, limitations, and downloadable evidence package. Preserve the original release when a material correction is issued and document what changed.

Final failure-mode check

Do not publish until each of these risks has an explicit answer:

• A press-style headline outruns the population, measure, or uncertainty.
• The report omits question wording, preceding context, or a stimulus detail that could change interpretation.
• Exclusions, subgroup choices, or outcome definitions were made after viewing results and are unlabeled.
• A sponsor’s role is hidden or the sponsor can suppress valid unfavorable findings.
• Only the preferred outcome appears while guardrails or wrong-answer categories are absent.
• A correction overwrites the prior result without a dated record.

Sources and scope

1. Disclosure StandardsAmerican Association for Public Opinion Research

   Detailed professional disclosure reference for sponsor, sample, recruitment, mode, dates, weighting, precision, questions, and presentation.

2. Statement on Statistical Significance and P-ValuesAmerican Statistical Association

   Peer-reviewed principles limiting common overinterpretations of statistical significance and p-values.

3. Standards and Guidelines for Statistical SurveysU.S. Office of Management and Budget

   Federal reference for documentation, data-quality evaluation, analysis, release, and preservation of survey information.

4. Questions to Ask About SurveysStatistics Canada

   Practical official guidance for evaluating population, sampling, questions, timing, sponsor, quality, and appropriate conclusions.

Sources support the specific statements described above; they do not validate this publication’s rubric, guarantee a compliant execution, or replace context-specific professional advice.