Methodology v1.3.0 — readiness definitions

Canonical readings

The readiness triple

The canonical register projection carries integer point values from 0 to 100. The overall readiness projection carries ratios from 0 to 1. Eligibility validates each projection in its native form and requires exact scalar agreement without imposing a monotonic order; web-pv does not derive one projection from the other.

Claimed

#claimed

The canonical register projection’s current claimed-readiness point value.

Published register values are integer points from 0 to 100. The web layer reads the server-derived value and does not calculate it from page input.

Verdict eligibility separately requires a complete claimed ratio in the canonical overall readiness row. web-pv does not derive either projection from the other.

Reported

#reported

The canonical register projection’s current reported-readiness point value.

Published register values are integer points from 0 to 100. The value is part of the public readiness triple even when a compact register row does not display it.

The reported axis is assessed independently from claimed and verified; no monotonic order is required for eligibility.

Verified

#verified

The canonical register projection’s current verified-readiness point value.

Published register values are integer points from 0 to 100. The web layer displays the server-derived value without smoothing or client-side recomputation.

The verified axis is assessed independently from claimed and reported. It summarizes current evidence; it is not a universal certification of fitness for a particular use.

Published grading contract

Grading, index points, and ordering review

Grading contract

#grading-contract

Claimed, Reported, and Verified index points are editorial assessments under the current methodology contract.

Automated collection currently updates derived context, confidence, freshness, and append-only change events. It does not by itself move an axis value.

A future evidence-to-score movement engine requires a separately versioned methodology change before it may change this contract.

Ordering anomaly

#ordering-anomaly

A derived state indicating that Reported exceeds Claimed, Verified exceeds Reported, or both, under the current projection.

The accepted states are consistent, reported_gt_claimed, verified_gt_reported, and both_adjacent_violations.

An anomalous row is provisional on register and map surfaces and excluded from citable verdicts while review is pending or rejected. Approval is bound to the exact C/R/V values and methodology version.

Index points

#index-points

Ordinal 0–100 display values for each readiness axis.

Index points are ordinal; differences within the stated resolution are not decision-significant.

The published display resolution is one index point; bands are the primary summary, while the integer triple remains prominent.

Exact public formula

Gap

#gap

gap = claimed − verified

Claimed points minus verified points in the canonical register projection.

The eligibility check requires an integer from -100 to 100 and requires gap = claimed − verified exactly.

A positive value means claimed is above verified in the current projection; a negative value means verified is above claimed. It does not infer intent or wrongdoing.

A row with any adjacent ordering anomaly is excluded from citable verdicts until the exact C/R/V fingerprint and methodology pin are approved; rejected rows remain excluded.

Classifier output

Strength classes

The canonical register projection’s directional watch-pressure class.

Eligible register rows use exactly Weak, Growing, Strong, or Critical. The web layer reads the class; it does not reproduce an upstream classifier.

Strength is not a readiness score or truth verdict. Noise belongs to watch classification and is not an eligible register-readiness strength value.

Class	Current web treatment
Weak	Eligible register class; rendered with the lightest bracket treatment.
Growing	Eligible register class; rendered with a stronger bracket treatment than Weak.
Strong	Eligible register class; rendered with a stronger bracket treatment than Growing.
Critical	Eligible register class; rendered with the heaviest bracket treatment.

Noise is used by watch classification. It is not accepted as a complete register-row strength for verdict eligibility.

Read-model context

Band, freshness, and public evidence treatment

Readiness band

#readiness-band

The canonical overall readiness row’s server-derived band.

The complete-row gate accepts unknown, emerging, watch, ready, or mature. web-pv validates membership and does not calculate the band.

Bands use the verified ratio: unknown below 0.15, emerging from 0.15, watch from 0.35, ready from 0.55, and mature from 0.75.

The band is the primary summary; the integer Claimed/Reported/Verified triple remains visible as the first-class reading.

Confidence state

#confidence-state

Freshness classification for qualifying evidence behind the current readiness row.

The live read model accepts fresh, stale, or deadman together with a parseable computed-at timestamp and a 0-to-1 confidence scalar.

Freshness uses validated or Tier 1/2 signals only. Automated unvalidated Tier 3 context signals do not refresh the state.

A stale or deadman state describes the reading’s monitoring freshness; it does not silently replace the canonical values with synthetic ones.

Refresh cadence

#refresh-cadence

Qualifying-evidence collection runs through an automated lane every six hours; derived context, confidence, freshness, and change events update from the evidence the lane admits.

The six-hour cadence describes collection opportunities, not a promise that every source or readiness axis changes on that schedule.

Only validated or Tier 1/2 signals refresh freshness; automated unvalidated Tier 3 context signals may add context but do not make a reading fresh.

Evidence-tier treatment

#evidence-tier-treatment

Tier information is consumed only as an already-derived aggregate from the canonical readiness projection.

Eligibility requires source_tier_counts to be a structured aggregate. The current web-pv gate does not assign tier names, publish weights, or recalculate readiness from those counts.

A public unit may describe the aggregate it actually receives; missing tier detail stays missing rather than being estimated.

Evidence-stance treatment

#evidence-stance-treatment

Stance may be shown only as an approved, derived aggregate when a public verdict projection supplies it.

The current readiness/register eligibility projection does not read per-signal stances. web-pv does not infer a stance mix or recompute readiness from raw stance rows.

When a later public verdict projection supplies support, contradiction, or neutral counts, the unit presents those counts as context and leaves absent groups unestimated.

Fail-closed publication gate

Verdict eligibility

#verdict-eligibility

The intersection of complete canonical overall-readiness and register rows for one canonical slug.

Load the complete canonical overall-readiness projection to its exact count.
Load the complete canonical register projection to its exact count.
Validate each projection half without passing raw rows beyond the helper.
Exclude pending or rejected anomalies and require a known row-level methodology pin.
Publish only canonical slugs present in both complete sets.

web-pv loads both projections to their exact counts, rejects malformed or duplicate readiness membership, validates each row’s public scalar contract, and sorts the resulting slug intersection.

A missing or incomplete projection half is excluded. Pending and rejected ordering anomalies are excluded, and an unknown methodology pin fails closed before a citable verdict is served.

Fetch, pagination, or count mismatch fails closed instead of publishing a partial eligible population.

Decision Brief mapping

Persona verdicts

#persona-verdicts

Rules v1.0.0 map the same canonical readiness and evidence measures into one closed classification for each operating frame. The mapping is deterministic and total; it does not use private records or the paid Decision Card audience field.

CTO

cto

Portfolio and operating-risk frame for deciding whether a technology belongs in production, a bounded pilot, active tracking, or outside the current plan.

Architect

architect

Integration and reversibility frame for deciding whether current evidence supports production architecture, an isolated pilot, monitoring, or no active design work.

Security lead

security_lead

Assurance and control-evidence frame with a higher bar for production adoption and stronger penalties for claim gaps and contradiction.

Closed verdict vocabulary

ignore → track → pilot → adopt

The order runs from no active work through monitoring and bounded evaluation to a production-adoption classification. It is a class, not fitness-for-purpose advice.

Deterministic rule table

Persona	Track from	Pilot from	Adopt from	Gap penalty	Stance factor	Tier factor
CTO	25	55	78	0.25	8	8
Architect	22	50	75	0.3	9	10
Security lead	30	62	85	0.4	12	12

Start with verified readiness. A positive claimed-minus-verified gap is subtracted at the listed persona rate. Strength, support-minus-contradiction share, Tier 1 plus half-weighted Tier 2 share, and deterministic evidence recency then adjust the score. Scores below Track map to Ignore.

verified Start with the canonical verified-readiness point value.

gap Subtract the positive claimed-minus-verified gap using the persona weight; a negative gap adds at most five points.

strength Apply the published Weak, Growing, Strong, or Critical strength adjustment.

stance-mix Apply the support-minus-contradiction share from the public stance aggregate.

tier-mix Apply the Tier 1 plus half-weighted Tier 2 share from the public tier aggregate.

recency Apply a deterministic adjustment from the verdict dateline to latest-observed date; absent dates receive the disclosed lowest adjustment.

Checkpoint derivation

Re-evaluate one published factor at a time and report only the first threshold crossing that changes the persona classification. Statements describe conditions and make no timing or outcome promise.

derived from current public signals; not fitness-for-purpose advice

Interpret with the dateline

Limitations

Evidence lag

#evidence-lag

Current evidence may lag behind a recent technology or deployment change. Read the computed-at date and fresh, stale, or deadman state along with the scalar values.

Coverage variance

#coverage-variance

Current evidence coverage may differ by technology and discipline. Incomplete projection pairs are excluded from verdict eligibility, and absent aggregates are not estimated.

No universal fitness claim

#no-universal-fitness-claim

A reading summarizes current evidence in the canonical projections. It may support a decision, but it does not establish fitness, safety, compliance, or suitability for every context.

Confidence heuristic

#confidence-heuristic

The current confidence scalar is a bounded count-and-tier heuristic over current evidence. It may support context, but it does not measure source independence or resolve contradiction direction; those inputs await a separately versioned redesign.

Authorship and public-data boundaries

Disclosures

EU AI Act Article 50 split

#art-50-split

Citable units identify AI-assisted assembly separately from dated human editorial judgment. Automation does not fill or relabel a human editorial section.

Methodology definitions and version decisions are human-reviewed methodology judgment. AI-assisted assembly in a published unit is labeled on that unit rather than being presented as human editorial judgment.

Rule #38 derived-results boundary

#rule-38-boundary

Public methodology and units expose derived results and approved public aggregates only. They exclude hidden weights, raw signals, trust records, Wisdom principles, source excerpts, and internal grader prompts.

The JSON twin carries this same public contract. It is not a corpus export or a route to restricted records.

Prospective changes

Revision policy

A material definition or interpretation change advances the methodology version and adds a dated revision. The current page, JSON twin, dateModified, and discovery metadata move forward together. Historical units keep their original dateline and methodology-version pin.

Revision register

v1.0.0 2026-07-12 Initial public methodology

v1.1.0 2026-07-17 Added the refresh-cadence definition making the automated collection lane’s schedule part of the public methodology.

v1.2.0 2026-07-17 Added the versioned persona-verdict rules, closed decision vocabulary, threshold table, checkpoint derivation, and fixed Decision Brief limitation.

v1.3.0 2026-07-19 Published the grading contract, signed-gap and ordering-anomaly gate, verified-band thresholds, ordinal index-point resolution, qualifying-evidence freshness, row-level pins, and confidence-heuristic limitation.

Canonical human URL: https://pragma.vision/methodology · Machine twin: /api/methodology.json