Methodology

#How topics are selected

Topics are chosen where Weavidence can add genuine explanatory value: a public- health or methodological question that benefits from clearer synthesis, a worked example that makes a difficult decision inspectable, or a product design that illustrates a broader principle without becoming a disguised feature list.

Potential traffic is not sufficient reason to publish. A topic should have a bounded question, an identifiable audience, sources that can be checked, and a credible explanation of what a reader will understand or do differently.

#How questions are scoped

Each article is built around one answerable question, stated in its metadata and answered directly near the beginning. Scope is narrowed until:

• the intended reader and decision are clear;
• material terms can be defined;
• the evidence can be assessed within one article;
• exclusions can be stated honestly;
• the conclusion does not require several different questions to be collapsed into one.

A broad topic may become a cluster of linked articles rather than one long page that treats every issue superficially.

#How claims are classified

During drafting, claims are distinguished as:

• product fact — a current, inspectable property or status of Weavidence;
• product design intent — what a product is designed or planned to do;
• external factual claim — a statement about the world requiring an appropriate external source;
• methodological principle — a claim about research, analysis, assessment, or evidence practice;
• interpretation — an editorial reading of evidence or design;
• inference — a conclusion drawn from stated evidence rather than directly stated by a source;
• recommendation — advice whose applicability and limitations must be made clear;
• illustrative example — a synthetic demonstration rather than empirical evidence.

This classification determines what kind of source and qualification are needed.

#How sources are identified

For methodological and public-health claims, the preferred hierarchy is:

1. current official institutional guidance;
2. current reporting or methodological standards;
3. primary research;
4. high-quality systematic reviews and consensus reports;
5. authoritative methodological references.

Foundational older sources are retained where they introduced a concept still in use, but are supplemented when current practice, limitations, or applications have developed. Broad Method Guides normally use more than one source so that a single document is not made to support definitions, application, and limitations beyond its scope.

For Product Notes, sources may include Weavidence's typed public content, architecture documentation, public pages, and direct implementation evidence. These sources establish what the project says, intends, or implements; they do not independently validate outcomes.

#How a source is verified

Before a source supports publication, the editor checks as far as the available environment permits:

• title, author or institution, year, and publication venue;
• stable URL, DOI, PMID, or official document location;
• whether a newer edition or superseding guideline exists;
• the population, setting, and question addressed;
• whether the cited passage actually supports the claim;
• whether the article is extending the source beyond its original domain;
• whether disagreement or important uncertainty should be represented.

Search-result snippets and secondary references can identify candidate sources; they do not by themselves establish the full claim. Where the primary content cannot be inspected with adequate confidence, the claim is narrowed, labelled as uncertain, or held from publication.

#How citations are used

Inline citation markers link each material claim to a specific reference. Every listed reference must be used, and every citation identifier must resolve. Automated validation enforces that structural relationship.

Citation density follows the claim rather than the paragraph count. One source may support several adjacent definitions, while a sentence combining distinct claims may require several sources. Sources are not added merely to make a bibliography look larger.

#How evidence is synthesized

Insights is not automatically a systematic-review publication. Unless an article explicitly reports a systematic search and selection method, it should be read as a source-grounded editorial synthesis.

The synthesis should:

• identify the role of each source;
• distinguish consensus, framework, empirical result, and editorial inference;
• represent important differences between sources;
• avoid turning absence of evidence into evidence of absence;
• keep applicability and implementation context visible;
• state what the article did not assess.

Evidence Briefs carry a higher expectation of explicit search scope and source selection than Method Guides or Concept Explainers.

#How review levels are assigned

#Editorial

Used for Product Notes and other content whose main claims concern Weavidence's own documented design or for low-risk explanatory material. The review checks status accuracy, source linkage, limitations, clarity, and policy compliance.

#Methodological

Used when the article makes material claims about study design, epidemiology, statistics, reproducibility, assessment, or public-health method. The editor checks the argument against appropriate external sources, tests definitions and examples for internal consistency, and verifies that recommendations do not exceed the evidence.

A methodological label does not imply independent peer review. When no reviewer is named, the review should be understood as internal methodological checking by the accountable publisher. A future independent review must identify the reviewer, role, scope, and conflicts rather than being implied by the label.

#Clinical

Used when content could materially guide an individual clinical decision. It requires an authorized named reviewer in the article metadata and is not published without that record. Most Insights content deliberately remains at the level of method, public-health systems, or research practice and does not enter individual clinical guidance.

#How examples and worked cases are governed

Every example is identified as one of:

• synthetic;
• illustrative adaptation;
• de-identified real case;
• public documented case.

A synthetic case should expose realistic ambiguity, missing information, alternative interpretations, and limitations. It must not imply that a named public authority or customer experienced the scenario. A real or de-identified case requires lawful use, permission where applicable, and enough provenance to prevent a fictionalized success story.

Worked Cases should normally make the reasoning chain visible:

problem
→ question
→ design or evaluation plan
→ data and assumptions
→ analysis or assessment
→ interpretation
→ limitations
→ decision

#How uncertainty is represented

Articles include a visible limitations section proportionate to their claims. Uncertainty may concern:

• source quality or currency;
• transfer between settings;
• measurement validity;
• missing populations or evidence;
• disagreement between frameworks;
• product maturity;
• untested implementation assumptions;
• the distinction between reproducibility and correctness.

Qualifications are placed where they affect interpretation, not saved only for a final disclaimer.

#How Product Notes differ from external evidence articles

A Product Note is an accountable inside explanation of what Weavidence built, plans, or learned. It may use internal documentation as a primary source and should link to public evidence where possible. Its commercial relationship is inherent and disclosed.

A Method Guide, Concept Explainer, Worked Case, Commentary, or Evidence Brief making general external claims must use external sources. It cannot inherit credibility from a Product Note or treat a product implementation as proof of a methodological principle.

#When current sourcing cannot be verified

An article remains draft or review when:

• a material source cannot be authenticated;
• the primary text cannot be inspected adequately;
• current guidance may have superseded the cited version;
• a required reviewer is unavailable;
• the article cannot distinguish evidence from inference;
• an example risks implying unsupported effectiveness or a real event.

Held articles receive no public route, sitemap entry, feed item, search record, or social image. This is a publication control, not a cosmetic label.

#What automated checks do

Repository validation checks structural properties such as:

• schema and status;
• unique slugs, titles, and descriptions;
• citation identifiers and reference use;
• source-count and source-type minimums for evidence-sensitive content;
• dates and revision history;
• topic membership and orphan prevention;
• exclusion of held articles from public output;
• canonical metadata, sitemap, feed, and social assets.

These checks can detect omissions and contradictions. They cannot determine that a causal claim is valid, a source is interpreted correctly, or a worked example is professionally appropriate. Human review remains necessary.

#Publication decision

An article is published only when:

• its question and scope are stable;
• material claims are supported or explicitly identified as interpretation;
• sources and links are verified to the required confidence;
• examples are labelled and non-deceptive;
• limitations are visible;
• review level is appropriate;
• no placeholder or unresolved citation remains;
• the accountable publisher accepts responsibility for the page.

Passing the build is necessary, not sufficient.