Provenance
Can the source, retrieval and lineage of the evidence be traced?
Evidence · provenance · integrity · verification · validity
Trust should be inspectable.
SCC Nexus does not use “verified” as a blanket badge. Each project is assessed against four separate questions: where the evidence came from, what protects it, what was actually checked, and what the result can legitimately support.
Traceable sourcesFail-closed gatesDated checksBounded claims
Evidenceinspect · challenge · revisitIntegrity
What controls reduce accidental, sensitive or unsupported publication?
Verification
What tests, reconciliations or human reviews have actually been performed?
Validity
What can the output reasonably support — and what does it explicitly not prove?
Flagship evidence systems
Proof that is repeatedly regenerated.
These are not static case studies. They are recurring evidence systems with public outputs and protected review layers.
Companies House Intelligence
National-scale public-record interrogation.
5,698,274rows scanned150,000priority-exported recordsRecurringreview cycle
Public redacted intelligence is separated from a protected/unredacted analytical review layer. Source-led ingestion, deterministic review indicators, validation, hashes/manifests and archived run artefacts support successive comparison.
Open sccnexus.com ↗Dated public proof: June 2026 run 27907375771.England Air-Quality Evidence Observatory
A weekly monitoring evidence cycle.
238/240valid station-days99.2%completeness9,696evidence rows
Scheduled harvesting is followed by source/integrity checks, validation gates, public-surface reconciliation, review-pack generation and archive controls. Protected review material is kept separate where unrestricted publication is not appropriate.
Open sccairquality.com ↗Dated accepted proof: 22 Jun–21 Jul 2026 panel.Audit trail
What a serious reviewer should be able to ask for.
The answer differs by project, so SCC Nexus exposes the control rather than claiming a universal certification.
Origin
What authoritative or primary source was retrieved, when, and under what identity?
Transformation
What code, calculation, filtering or enrichment changed the source into the analytical record?
Integrity
What completeness, schema, source, duplicate, limitation or publication gates were applied?
Verification
What tests, reconciliations and human-review decisions were actually completed?
Version
Can the evidence state be tied to a dated release, archived artefact, manifest or hash where one is produced?
Validity
What can the result support — and which causal, clinical, legal, regulatory or financial conclusions remain outside scope?
Designed to withstand scrutiny, not to borrow authority.
These controls make the work inspectable and challengeable. They do not constitute independent certification. SCC Nexus states external validation only when documentary evidence of that validation exists.
Project evidence ledger
Open the detail when you need it.
The full audit evidence remains available, but it no longer has to dominate the page.
01SCC VapingActive research · identity updated September 2026Evidence triage with deliberate abstention
Study records, protocols and post-publication notices remain distinguishable before automated triage.
Deterministic fixes precede ML assistance and ambiguous evidence is allowed to abstain.
10 high-consensus, 35 moderate and 661 abstentions in the reported triage snapshot.
Valid for evidence discovery and triage; not scientific-quality appraisal, causal inference or clinical guidance.
02UK Fusion Evidence ObservatoryResearch observatory · reviewed August 2026Human decision remains mandatory
Diagnostic review records can retain FAIR-MAST UUID, shot, signal, interval, version/commit, reviewer and decision.
The engine fails closed if an export attempts to assert scientific verification or publication approval.
Human diagnostic review is separated from automated assistance and from the public Observatory.
Valid for provenance-led evidence linkage and review workflow; automated annotation is not scientific verification.
03SCC StocksOperational research service · reviewed August 2026Research prompts, not trading execution
Research prompts use public market/company information and are separated from trading execution.
Risk gates, manual review and paper-trading separation prevent automatic execution or advice.
Repeatable research workflow with human review of highlighted opportunities and risks.
Research and screening aid only; not financial advice or a prediction of outperformance.
04LewesLiveOperational local platform · reviewed August 2026Public presentation separated from discovery
Event and directory records are intended to retain source links/context through upstream engines.
Quality gates can withhold incomplete, blocked, duplicated or unsafe material.
LewesLive receives validated outputs rather than directly exposing raw discovery records.
Local discovery service; not an authoritative statutory register.
05Local Events EngineReusable infrastructure · reviewed August 2026544 confirmed test functions
Internal-audit, public-build, LewesLive-import and test-review outputs remain separated.
Public-build validators, blocklists, provenance requirements and source-health reporting act as gates.
544 test functions confirmed; builds can be withheld when quality gates fail.
Valid for discovery, normalisation and publication preparation, subject to validation.
06Local Directory EngineDeveloping reusable infrastructure · reviewed August 2026Provenance-aware public/private boundary
Listing discovery is designed to preserve where a business/service record came from.
Private implementation, credentials and source-specific tactics are separated from public output.
Current verification is principally architectural/public-surface based.
Developing infrastructure; not an independently validated national or statutory directory.
External validation remains separate. SCC Nexus does not imply independent academic, regulatory or third-party validation unless it is specifically documented.