Evidence-support infrastructure for harm-minimisation visibility.
Standards Protocol · Intelligence Observatory · NSW Gaming Observatory. A longitudinal, public-data interpretation environment built to survive auditor, regulator, academic, and media scrutiny simultaneously.
The observatory is currently in Stage 1: building the governance spine (evidence taxonomy, methodology registry, provenance chain, audit chain, confidence framework, language policy, institutional-use classification, known-limitations registry, and immutable version snapshots). Public observatory views (venue / LGA / state) appear from Stage 4.
Seven classes every output is tagged with: observed, derived, modelled, proxy, interpretive, scenario, demonstration.
Every metric the platform produces must have a registry entry: formula, source dependencies, assumptions, limitations, version history.
OpenHash-chained audit log, SHA-256 source lineage, four institutional confidence levels driven by six weighted inputs (incl. historical completeness across 4 continuity factors).
Institutional use classification
Every output is tagged with its intended audience. Stage 7 role-gating (5 audiences) is built on these tags.
- PublicCleared for public release. Observed and derived outputs only, with full provenance and limitations disclosure.
- AnalystFor authorised analysts. May include modelled outputs with confidence framing. Not for public surfaces.
- AdminFor administrators and governance custodians. Includes parser logs, audit chain inspection, methodology revisions.
- ResearchFor approved researchers under data-access agreement. Permits export of aggregated or derived series with provenance.
- DemonstrationFor walkthroughs, regulator demos, presentation environments. Clearly labelled DEMONSTRATION. Never operational.
Known limitations registry
8 platform-wide limitations that every output must respect or disclose.
- temporalReporting lag
- granularityAggregate-only observations
- data-qualitySchema inconsistencies across releases
- identityVenue alias instability
- inferentialInability to infer causation
- inferentialVenue metric gating
- temporalHistorical archive separation
- transparencyPublication coverage disclosure
Staged build plan
- Stage 01Evidence Architecture Hardeningin progress
- Stage 02Forensic Ingestion Layerplanned
- Stage 03Canonical Warehouse · Identityplanned
- Stage 04Public Observatoryplanned
- Stage 05Deterministic Metrics Engineplanned
- Stage 06Dashboard Mode (secondary)planned
- Stage 07Private Analytical Layerplanned
- Stage 08Reform Impact Monitorplanned
- Stage 09Comparative Observatory Engineplanned
- Stage 10Research · Reporting · Exportplanned
- Stage 11Operational Admin · QAplanned
- Stage 12Guided Demo · Presentationplanned
- Stage 13Final Hardening · Validationplanned
One stage at a time. No blending. Each stage is locked before the next begins.
Stage 3 · Canonical Entity Resolution Layer
LOCKED · v3.0-LOCKEDLongitudinal venue identity architecture. Canonical entities are continuity-mapping overlays — not legal determinations. Original source identity values are preserved write-once and never overwritten. Every resolution decision is reversible and immutably audited.
Stage 4 · Deterministic Rank-Based Venue Observability Engine
observation only · no dollars · no inferenceDeduped venue-quarter observations + per-venue movement timelines (rank Δ, EGM Δ) + observation coverage classifications + publication coverage classifications. Rank-only publication basis preserved — NSW does not publish dollar values in the B/C streams. Internal validation surface, not a public dashboard.
Public-output boundary policy
Public surfaces show only observed and derived outputs. The platform never publishes risk scores, intervention recommendations, predictive claims, or enforcement determinations.
Language policy lists 36 prohibited terms across 4 severity levels and 27 permitted alternatives.