Regulated Commerce Research Methodology

Scope before score

This is a published research method, not an industry ranking or a certification program. The initial sector focus is cannabis. This release contains technical education and synthetic examples; no pilot records or external-cohort findings are published.

The existing scanner covers a subset of public signals. Manual interpretation and connected evidence remain separate work. Missing optional MCP, A2A, OpenAPI or llms.txt surfaces are not automatically failures.

Evidence states

Observed

A specified observer directly recorded the result under stated conditions.

Declared

The operator or a document makes the claim; operational verification is separate.

Unknown

The evidence needed to answer the question is unavailable or insufficient.

Not applicable

The check does not apply to the declared task or scope.

Contradictory

Two relevant representations or observations conflict.

Manual review required

An automated check cannot safely settle the interpretation.

The 12-check review

1. Public response integrity

   Inspect the final response and content type, not the status code alone.

   Evidence lane: public snapshot

2. Declared crawl policy

   Record the applicable robots rules without calling them proof of access.

   Evidence lane: public snapshot

3. Discovery links

   Inspect sitemap and internal discovery links; verify sampled destinations.

   Evidence lane: public snapshot

4. Canonical identity

   Compare the intended domain, canonical URL, title, and entity identifiers.

   Evidence lane: public snapshot

5. Semantic content

   Check whether the response contains useful text and ordinary links.

   Evidence lane: public snapshot

6. Structured-data consistency

   Compare declared structured facts with visible content; a parse success is not truth verification.

   Evidence lane: manual review

7. Age-gate boundary

   Record what is visible before interaction; do not circumvent eligibility controls.

   Evidence lane: manual review

8. Representation parity

   Compare sampled HTML and machine representations for conflicting facts.

   Evidence lane: manual review

9. Optional interfaces

   Inspect only interfaces actually claimed by the operator. Missing MCP/A2A is not automatically a defect.

   Evidence lane: conditional

10. Verified request evidence

    Require authorized first-party observations and a stated identity verification method.

    Evidence lane: connected evidence

11. Attributed referrals

    Require authorized referral/session evidence; do not infer visitors from crawler counts.

    Evidence lane: connected evidence

12. Business outcomes

    Require a documented attribution window and paid/account outcome evidence.

    Evidence lane: connected evidence

How to run a bounded review

1. Define the task, subject and authorization before collecting data.
2. Keep public requests, authorized browser tests and connected provider evidence separate.
3. Record observer, methodology version, observation time, outcome and limits in an access-controlled working record.
4. Review contradictions and coverage gaps before interpreting results.
5. Give each proposed change a repeatable acceptance test; preserve retest history.
6. Publish only synthetic examples or records explicitly approved for release.

Confidentiality and independence

Implementation access does not grant publication permission. Private business identities, URLs, screenshots, customer data and distinctive operational details do not belong in the public library. Operator-managed work is not independent validation and must not enter an external benchmark denominator.

A benchmark requires a defined external cohort, actual observations, coverage reporting, independent interpretation where claimed, and publication review. This release has no benchmark results and no composite readiness ranking.

Corrections and current interfaces

Submit a private methodology question or correction. Do not include credentials or private evidence links. Accepted corrections should be versioned rather than silently changing historical findings.

Machine-readable library contract · Current product capabilities · Research library