Methodology v1.0.0
A market record built for evidence, not volume.
Include only an observable transaction or governed arrangement in which a proprietary corporate corpus, operating history, institutional memory, or controlled rights to that corpus are strategically material to AI training, post-training, customization, validation, evaluation, or improvement. Ordinary enterprise AI deployments and retrieval-only connectors are excluded unless the AI counterparty receives separately material data rights.
Qualification test
A record qualifies only when a proprietary corporate corpus is identifiable, strategically material to the AI system, and subject to an observable rights or access arrangement. A model merely connecting to customer data is not enough.
Locked V1 taxonomy
One primary transaction structure is assigned to each observation. Asset state, corpus renewal, data type, and individual rights remain separate facets, preventing category overlap.
Two scores, two questions
Factor ratings run from 0 to 5 in half-point increments. For observed factors, sum weight × rating ÷ 5, divide by observed weight ÷ 100, and round once. A score is null below 70% coverage. Unknown factors are null; zero represents evidenced weakness, not missing evidence. Scores are qualitative analysis, not measured performance or transaction valuations.
AI Data Asset Score
Measures the intrinsic strategic usefulness of the underlying data asset, not the quality of price discovery.
Transaction Signal Score
Measures how strongly the transaction evidences standalone economic value for proprietary corporate data.
Evidence and uncertainty
Primary evidence is preferred. Secondary reporting can corroborate or fill a public-record gap. Every record separates disclosed facts from Strategic Signal analysis, distinguishes dataset-specific economics from bundled program value, and preserves unknown or not-disclosed terms.
Candidate-to-public workflow
Approval records the authenticated reviewer, timestamp and exact record fingerprint. Changing a record or its evidence requires another approval. Legacy observations without this evidence are explicitly provisional and excluded from approved aggregates.
Monitoring can create candidates, but it cannot publish records. Weak observations are rejected and retained in a review ledger so repeated false positives improve future qualification.
Versioning
Methodology versions are immutable. Corrections to observations are recorded in release metadata; scoring-rule changes require a new methodology version and a versioned restatement. V1 is intentionally not an index. Longitudinal measures will be calculated only after the observation base is sufficiently broad and stable.