Publishing principles

Methodology and evidence standards

Measurement Press treats claims as decisions waiting to happen. A claim deserves more trust when the evidence trail, comparison group, uncertainty, incentive structure, and practical consequence are all visible.

Reader standard: strong pages should let a careful reader see what is known, what is inferred, what would weaken the claim, and which next guide can help them apply the same test.

Evidence trust route

Use this route when a reader wants to check how evidence standards, source quality, funding, privacy, advertising separation, and corrections fit together before relying on a page.

Reader questions

Methodology questions

Use these checks when a reader needs the short version of how evidence, source limits, causal language, case studies, ads, and corrections are handled.

Evidence grade

How are evidence grades assigned?

Evidence grades describe how close the source is to the claim, whether the denominator and comparison are visible, and whether the method can support the sentence being published.

Scope

What happens when a source is useful but narrow?

A useful but narrow source can support a bounded observation, example, or directional signal. The published claim should name the scope, limitation, and next check instead of treating the source as universal proof.

Causal language

When can a page use causal language?

Causal language should be reserved for evidence with a design that can support the counterfactual question, such as a valid experiment, holdout, matched comparison, or model with clearly stated assumptions and limits.

Case studies

How should case studies be used?

Case studies are worked examples for recognizing a measurement or framing pattern. They should not be treated as allegations about named companies or as proof that the same effect applies in every setting.

Separation

Can advertising affect the method?

No. Advertising support should stay separate from evidence grades, source treatment, claim language, guide placement, corrections, and editorial conclusions.

Correction

What happens when better evidence changes a claim?

The page should narrow, revise, or correct the claim so readers can see what changed, which evidence changed confidence, and whether the conclusion still holds.

Start with the decision

Every article, guide, case study, and source page starts by asking what decision the reader is being asked to make. A media frame may affect whether a reader believes a public claim. A measurement result may affect budget, renewal, creative, or inventory decisions. A case study may affect whether a pattern is recognized in a new setting.

The publication standard is to make the decision visible before judging the evidence. That keeps a useful but narrow source from being treated as proof of a broader claim, and it keeps a measurement method from answering a question it was not designed to answer.

Method pathways

Source trail

Evaluate evidence before accepting the frame

Use the evaluation desk when a story, report, vendor result, chart, or quote needs its source, denominator, comparison, and claim language inspected.

Reader job

Choose the right evidence workflow

Use the taxonomy when the issue is not yet clear: media framing, measurement design, campaign reporting, buyer readiness, or inventory quality.

Reference quality

Prefer primary material and method-bearing sources

Use the source library when a page needs official documentation, standards, research methods, measurement references, or inventory checks.

Accountability

Correct the claim when better evidence changes confidence

Use the corrections policy when an error, stale source, missing denominator, or overstated causal phrase affects what a reader should believe.

Evidence grades

Evidence grades describe how close the source is to the claim and how much method detail a reader can inspect. A grade is not a reputation score for a person or institution. It is a page-level judgment about whether the cited support can carry the sentence being published.

GradeMeaningRequired supportClaim language
PrimaryDirect record or original method-bearing source.Original document, dataset, transcript, filing, paper, standard, or official methodology note.Can support a bounded factual description when the denominator and date are clear.
Strong secondaryCredible analysis grounded in primary material.Named author or institution, visible citations, method transparency, and clear limits.Can support synthesis when the page identifies what is analyzed and what remains uncertain.
DirectionalUseful but incomplete evidence.Partial data, narrow sample, survey, vendor readout, modeled estimate, or expert interpretation with stated limits.Should use careful verbs such as suggests, is consistent with, may indicate, or warrants review.
UnsupportedThe current support cannot carry the claim.Missing citation, vague sourcing, circular summaries, anecdote presented as prevalence, or a measurement result used beyond its design.Should not be repeated as fact until adequate evidence is found.
Evidence action

Turn the evidence grade into the next check.

After a source is graded, send the reader to the guide that narrows the claim, requests support, or chooses a stronger test before the sentence travels.

Claim review sequence

1. State the claim plainly.

Rewrite the headline, vendor result, chart takeaway, or case-study lesson as one testable sentence. Decide whether it is descriptive, comparative, predictive, or causal.

2. Identify the source role.

Separate original records, direct participants, data owners, vendors, analysts, commentators, sponsors, and secondhand summaries. Source proximity and source incentives both matter.

3. Find the denominator.

Name the population, opportunity base, reporting base, time window, and exclusions. Counts, rates, lift figures, and percentages are weak until the base is visible.

4. Name the comparison.

Ask what the reader is comparing against: prior period, peer group, baseline, holdout, matched geography, modeled counterfactual, or no comparison at all.

5. Preserve the strongest weakener.

List the source, counterexample, sensitivity check, pre-period trend, alternate denominator, or uncertainty interval that would materially lower confidence.

6. Match the verb to the evidence.

Before publication, adjust causal verbs, certainty words, and summary language so the final claim does not outrun the support.

Media analysis method

Media analysis inspects the editorial moves that shape reader confidence: framing, comparison class, base rate, source weighting, headline scope, causal verbs, quote balance, anecdote selection, timeline order, and omitted disconfirming context. A critique is not publishable unless it identifies a specific editorial move and a concrete correction or narrower alternative.

Readers should be able to distinguish a reporting error from a framing problem. A factual error may require a correction. A framing problem may require a clearer denominator, a fairer comparison, a weaker verb, a visible counter-source, or an explanation that the evidence is illustrative rather than representative.

Measurement analysis method

Measurement analysis defines the estimand, counterfactual, method, assignment or matching process, eligible population, time window, outcome quality, uncertainty, and decision relevance. Attribution, marketing mix modeling, lift testing, attention measurement, survey research, and brand studies should not answer questions they were not designed to answer.

A measurement page should separate observation from causal interpretation. Credited outcomes are not automatically incremental outcomes. Survey lift is not automatically market impact. A model fit statistic is not automatically a valid budget recommendation. Useful results can still be published when their scope is clearly bounded.

Case-study method

Case studies are worked examples based on realistic business mechanics. Each one defines the business incentive, the advertised proof claim, the likely bias mechanism, the naive read, and a better comparison. The goal is to help readers recognize a measurement pattern, not to accuse a specific company.

A case-study conclusion should travel only as far as the evidence allows. The page should make clear whether the lesson is about selection bias, baseline choice, outcome quality, audience eligibility, survey recruitment, conversion lag, or generalizability.

Advertising and sponsorship separation

Advertising may support the publication, but it should not determine editorial conclusions, evidence grades, source treatment, corrections, or guide placement. Sponsored, affiliate, or commercial relationships should be disclosed when they are relevant to reader trust. The public reader-facing standard is kept on the advertising standards page.

Editorial pages should remain readable and complete on their own. Internal links, source references, and article structures should serve the reader's next question rather than push them into unrelated commercial paths.

Revision and correction standard

Corrections identify the original problem, the improved claim, the evidence that changed the assessment, and whether the conclusion changed. Routine updates for source freshness, accessibility, search metadata, or internal links do not require a public correction unless they alter a claim, evidence grade, disclosure, or reader decision.

When a page changes because better evidence narrows confidence, the revision note should be close to the affected article or guide. The correction should make the page easier to trust, not harder to inspect.

Reference standards

For grouped workflows that map official references to media-framing, citation, measurement, attention, marketing mix modeling, and inventory checks, use the source library.