Evidence before magnitude.
Version 2.0-pilot defines an entity–mechanism–market–time assessment. The pilot records four ordinal judgments or explicit unknowns. It does not combine them into a total.
Definition and boundary
One principal mechanism is assessed per company. Its legal entity, relevant market, counterparties, geography where material, operating stage, evidence cutoff, and invalidation condition must be explicit. Up to two additional mechanisms may be described without additional scores.
Anchored judgments
| Criterion | 0 | 1 | 2 | 3 |
|---|---|---|---|---|
| Control | Affirmative evidence of no relevant right | Rights principally support self-use | Enforceable rights affect a material counterparty | Consequential terms across multiple material counterparties |
| Substitution constraint | Routine usable alternatives | Alternatives require bounded switching work | Material cost, delay, capacity limit, or performance loss | No viable alternative at specified requirements and scale |
| Realized leverage | Counterparties bypass or neutralize leverage | Ordinary participation only | Mechanism-linked favorable terms, economics, allocation, or commitment | Such outcomes recur across counterparties or periods |
| Durability | Loss or displacement expected within 12 months | Continuation is bounded inside 12–24 months | Evidence supports the 24-month thesis, subject to named risks | Persistence or renewal is also visible across prior cycles |
Grades are ordinal judgments, not equal-interval measurements. They are never summed, averaged, normalized, or converted to percentages. If evidence cannot distinguish adjacent anchors, the result is unknown. A zero requires affirmative evidence.
Summary states
- Insufficient evidence: Control or substitution cannot be graded with at least Moderate confidence.
- Power not established in this mechanism: supported Control or Substitution is below 2.
- Structural potential: Control and Substitution are at least 2, but realized leverage is below 2 or unknown.
- Demonstrated leverage: Control, Substitution, and Realized leverage are at least 2 and the mechanism is operational.
- Demonstrated leverage; durability supported: Demonstrated leverage plus Durability of at least 2.
Announced, proposed, or contracted-but-not-operational mechanisms are prospective and cannot receive a demonstrated-leverage state.
Evidence and confidence
Every non-unknown grade must resolve to a dated claim, source, publisher, source type, observation period, locator, supporting extract or permitted snapshot hash, reasoning, material counterevidence, freshness status, and verification outcome. A graded profile requires at least two distinct documents and one primary source. Comparative claims require independent or counterparty support.
High confidence requires direct appropriate primary support, corroboration for comparative claims, a current validity check, and no unresolved material contradiction. Moderate confidence permits a visible bounded inference. Limited confidence forces the grade to unknown. Confidence is not a probability and is never multiplied into a score.
Automation, cadence, and versions
Collection, extraction, verification, contradiction checks, freshness checks, and release assembly are automated. No company row requires owner review. Unsupported claims abstain. Source text is treated as untrusted data, access restrictions are respected, and retrieval is bounded to eight documents plus two targeted follow-ups per company.
Source changes are checked monthly. Full profiles refresh quarterly. Verified material events receive dated notes. Framework, methodology, schema, pipeline, and immutable release identifiers are separate. Historical releases are corrected by a linked successor, never silently rewritten.
Legacy compatibility
The v1 weighted exposure model remains available as a labeled historical scaffold. There is no numerical conversion from its 0–10 values to v2. The legacy Strategic Consequence Engine stays pinned to v1 and is not allowed to treat v2 unknowns as zero.