Open the visual model ↗
Name the kind of claim first
A regulator’s published text, a platform’s eligibility page, a self-reported publisher result, a survey estimate, and an editorial hypothesis answer different questions. Keeping each label attached prevents a document collection from turning into a false consensus. Repetition across three briefs may mean they share the same source.
The supplied research uses fact, vendor claim, estimate, inference, and judgment labels. Its privacy report also separates technical observation, contractual promise, legal text, and interpretation. Both schemes can be useful if the reader sees which evidence supports the particular sentence.
- Fact from a primary source: say exactly what the source establishes.
- Vendor claim: identify whose product, policy or performance claim it is.
- Estimate: state the sample, method, period and limits.
- Inference: show the reasoning connecting evidence to a conclusion.
- Editorial judgment: make the recommendation and its assumptions explicit.
Compare the same unit and period
A pageview rate cannot be compared directly with a session rate, an impression rate, or revenue per subscriber. A revenue share depends on the pool to which it applies. A payment statement can represent an earlier earning period. Cost treatment can make a nominally higher rate the weaker business result.
| Field | Question |
|---|---|
| Population | Independent publisher, large group, nonprofit outlet, newsletter, or another business? |
| Period and geography | When was the result earned, observed, and paid? In which markets? |
| Unit and currency | Pageview, session, impression, subscriber, order, license, or package? |
| Gross and deductions | Before or after buyer fees, network share, refunds, commissions and payment fees? |
| Costs and time | Which cash costs, owner labor, fixed overhead and obligations are included? |
| Evidence method | Audited records, source disclosure, survey response, vendor estimate, or hypothetical arithmetic? |
| Transferability | What conditions could make this result fail elsewhere? |
A proposed study has no results
The imported strategy proposes a revenue survey, search-loss panel, partner comparisons, a consent observatory, an ad-network test, newsletter-rate research, a crawler census, and case interviews. Its proposed sample sizes, cell suppression rules, and deadlines are planning assumptions. None is evidence that the work was conducted.
Before presenting a result, record recruitment, participation, selection bias, missing data, exclusions, instrument design, and changes over time. A panel of volunteers can be useful without representing every publisher. A before/after move can coincide with seasonality or a product change; it does not establish causation by itself.
A real case needs a real subject, agreed disclosure scope, original documents or attributed self-report, baseline, costs, outcome, and limits. A hypothetical model needs visible assumptions and checked arithmetic. Do not move one into the other category because it makes a stronger headline.
Use uncertainty to design the next test
If a revenue rate is unknown, leave it unknown and explore a sensitivity range chosen for the exercise. If a source page conflicts with another page from the same company, record the conflict and seek a dated answer before a consequential commitment. If a policy change is only proposed, keep it out of instructions written as current obligations.
The practical result is a smaller set of defensible comparisons and a clearer verification queue. A source register lets readers inspect where a statement came from. It does not, by itself, prove that every quotation, date, claim, or inference is correct.
Sources & limits
Methodological guidance adapted from the supplied research. It contains no measured revenue benchmark, completed survey, original case study, or forecast.
Adapted from supplied research. See the evidence and review method. Section source notes identify supporting references; operational suggestions remain editorial judgment.
Source claims and editorial judgments remain separate. Send a correction with the passage and supporting evidence.
