QUERY → EVIDENCE → ACTION

How we inspect a market boundary.

The method starts with a buyer decision, not a technical checklist. It then preserves the source, observation date, missingness and interpretation limit behind every claim.

THE RESEARCH SEQUENCE

Six gates, in order.

Automation directs review. It does not make the final judgement about local responsibility, proof quality or cross-language meaning.

  1. 01

    Define the buyer question

    State the decision that evidence must support and the population to which any result may apply.

  2. 02

    Lock official sources

    Use dated, inspectable population sources and publish inclusion and exclusion rules.

  3. 03

    Collect responsibly

    Use a disclosed research user agent, respect applicable robots rules, retry cautiously and retain outcome classes.

  4. 04

    Direct human review

    Use discovered routes to inspect product, purchase, operating and verification evidence in context.

  5. 05

    Test reliability

    Time the work, blind re-code selected cases and revise fields that do not produce repeatable decisions.

  6. 06

    Publish with limits

    Release denominators, missing observations, source dates, corrections and the narrowest defensible interpretation.

WHAT GETS REVIEWED

Six evidence layers.

A language route is one layer. It is never used as a proxy for a complete buying path or for AI visibility.

01

Access

Can the public page be collected under applicable robots rules, and what could not be observed?

02

Language

Can a Japanese buyer reach and understand the product and relevant use case?

03

Purchase continuity

Can the buyer move to pricing, trial, demo, contact or procurement without an inconsistent handoff?

04

Operating evidence

Who sells, implements and supports the product in Japan?

05

Verification

Can the buyer find relevant customer, partner, security, compliance and third-party proof?

06

AI-assisted discovery

Do controlled answer panels surface the company and cite the exact supporting page?

AI OBSERVATION

No universal visibility score.

AI-answer checks use a fixed prompt panel, declared locale and account assumptions, dated full answers and exact cited URLs. A mention without a citation is recorded separately. Results describe that controlled run; they do not estimate every model, user or buying situation.