SSStrategicSignal

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.

Effective September 15, 2026Immutable scoring versionExplicit unknown states

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.

The core distinction is rights: access for inference is not training; access is not ownership; a partnership is not a data transaction unless the corpus is strategically material and evidenced.

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.

1Data Asset Acquisition
2Proprietary Corpus License
3Training Rights License
4Governed Live Data Access
5Federated / Controlled Learning Rights
6Sovereign / On-Prem Learning
7Strategic Model Co-Development
8Proprietary Data + Equity Partnership
AStranded / Separable
BEmbedded / Continuing
CFinite Corpus
DRenewable Corpus

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.

Uniqueness / scarcity12
Historical depth8
Decision → outcome richness13
Decision Value Density13
Real-world grounding10
Domain economic value9
Refreshability8
Proprietary advantage9
Model-learning usefulness10
Non-replicability4
Rights usability4

Transaction Signal Score

Measures how strongly the transaction evidences standalone economic value for proprietary corporate data.

Disclosed economics16
Clean price discovery16
Separability of data consideration15
Explicit training / model-improvement use14
Clarity of rights purchased12
Competing bids10
Strategic buyer quality6
Repeatability / precedent value6
Independent model-uplift evidence5
Decision Value Density asks how economically consequential the decisions represented in the corpus are. Drilling, yield, clinical, underwriting, fraud, or safety decisions can carry far more information value than routine low-stakes records.

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.

01Public signal
02Candidate + de-duplication
03Qualification test
04Evidence collection
05Classification + scoring
06Human review
07Publication
08Ongoing re-review

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.