Evidence Governance
How evidence becomes publishable.
Published evidence must pass governed thresholds for confidence, verification, and review. These standards exist before any outcome is published. They define what qualifies as proof.
Classification
How proof is classified
Data derived from real assessments, decisions, and verified outcomes. Sample size and verification method are attached.
Statistics computed across multiple cases. Published only when minimum sample thresholds are met.
Illustrative examples showing how the system classifies and governs. Not verified outcomes. Visibly labelled.
Anonymised case dossiers built from real operating patterns. Not live data. Labelled accordingly.
Verification
How outcomes are verified
Captured from user feedback. Never represented as independently verified. Not eligible for publication.
Derived from observed user actions within the system. Tracked through execution records and commitment checkpoints.
Supported by documentary evidence: contracts, financial records, organisational data. Requires source tracing.
Independently reviewed and confirmed by a human operator. Highest individual verification method.
Hierarchy
Evidence origin hierarchy
Not all evidence carries equal weight. The system distinguishes between:
- Operator-confirmed — highest trust. Human review of documentary evidence.
- Documentary — source-traced to financial, contractual, or organisational records.
- Behavioural — observed actions within the system over time.
- Self-reported — user-submitted feedback. Useful for calibration. Not publishable as proof.
Integrity seals
What the seal levels mean
Every evidence asset receives an integrity seal based on confidence, verification method, and data completeness. See the full seal registry for details.
Verified internally. Not publishable. Minimum 85% confidence.
Outcome-supported. Publication-eligible after human review. Requires behavioural or documentary evidence and financial impact.
Operator-confirmed with documentary trace. Requires 90%+ confidence, financial impact, and contract trace.
Reserved. Not currently issued. Will require repeated verified patterns across multiple independent cases.
Publication
When evidence may be published
Publication requires at minimum a Silver integrity seal.
Aggregate metrics require a minimum of 15 verified cases before any public claim is made.
Case studies require human review and anonymisation verification before publication.
Self-reported outcomes are never publishable as proof. They are used for internal calibration.
Review
The human review gate
No evidence auto-publishes. Every case draft, aggregate metric, and proof block passes through human review before reaching any external surface. The review confirms anonymisation, accuracy, and that no identifying information can be derived by cross-referencing public data.
Privacy
Anonymisation and privacy controls
All published evidence is anonymised. No client name, individual name, or organisation identifier appears in any public proof unless expressly authorised. Financial figures are presented as ranges where identification risk exists. Timeframes are generalised when specificity could enable de-anonymisation.
Boundaries
What we do not publish
- Internal classification methods or computational structures
- Routing logic or decision-engine internals
- Exact admission or refusal thresholds
- Internal state architecture
- Proprietary operating mechanics
- Individual respondent data
- Unapproved or suppressed evidence
- Client-identifying information without permission
- Outcomes below the publication seal threshold
Labels
What proof labels mean
This illustrates how the system classifies proof. It is not a published verified outcome.
Anonymised case built from real operating patterns. Not live data.
Outcome verified through the integrity seal system and approved for publication.
Not enough verified cases to publish aggregate metrics. This is honest, not evasive.
Current status
Current publication posture
Some outputs displayed on this platform are demonstration patterns. Published evidence requires human review. Self-reported outcomes are not treated as publishable proof. Aggregate metrics require minimum cohort thresholds. Client-identifying data is not published without explicit permission.
We publish our standards before our outcomes. The standards are the proof that the institution exists.