Distinct research signals
The Expert Score measures product quality across eight categories. The Trust Score measures company, disclosure, permission and evidence factors. Risk Level summarizes material downside concerns. Research Status, Evidence Confidence, Research Coverage and Research Freshness communicate different dimensions of uncertainty. None predicts profit or replaces a user's security assessment.
Controls, capabilities and practical usability.
What can be traced to current, relevant sources.
Permissions, incidents and unsupported claims.
Expert Score weighting
Each category receives a 0–100 score with a concise product-specific rationale. The published weight is applied once; values are summed and rounded.
Example: an 80 Security score contributes 16 points because Security has a 20% weight. A polished feature set cannot erase missing performance evidence because each category remains visible.
Calculated Trust Score
Thirteen factors cover identity, registration, team, history, pricing, security, API permissions, withdrawals, documentation, performance claims, support, incidents and regulatory information. Factor weights total 100%, and production code calculates the final score. Editors do not enter it manually.
Unknown is a valid result. Not Disclosed, Not Verified and Not Available are not silently converted into positive assumptions.
Risk Level and Research Status
Risk Level
Low, Moderate, High or Critical reflects material product, permission, incident and trust concerns. It is not market-risk tolerance or a promise of safety.
Research Status
Verified, Partially Verified, Unverified or Under Review describes the state of the research record. Under Review flags profiles such as Quadency where current operator or product evidence remains materially uncertain.
Evidence Confidence, coverage and freshness
Evidence Confidence does not mirror the score. High means broad, current coverage with a verified research status; Moderate means useful evidence with limitations; Limited means sparse sources, important gaps or an Under Review profile. A product can have a mid-range score and Limited confidence.
Research coverage reports nine areas: operator, terms, pricing, API documentation, security documentation, regulatory search, incident search, claims and current status. Coverage means the area was reviewed—not that the product is safe or the outcome favorable.
Research Freshness is derived from dated evidence-area checks and configured review intervals. Current, Review Due, Partially Stale and Under Reverification are maintenance states—not scores and not product-risk labels. Stale information remains explicitly scheduled for reverification rather than being declared incorrect automatically.
Source hierarchy and contextual attribution
Company terms, documentation, pricing, security pages and product disclosures. Best for what a provider says or offers, not independent validation.
Regulatory databases, official notices and company registries. Used within their jurisdiction and entity scope.
Audits or research with a clear method, accountable publisher and relevant scope.
Credible reporting that adds context but may not establish technical implementation.
Source authority is contextual, not a numerical quality score. Ten URLs controlled by one company can document ten details, but they remain primary evidence from one provider and do not become independent corroboration through repetition.
- Important legal, pricing, security and regulatory sections link to relevant sources in context.
- Source labels distinguish company documentation from independent evidence.
- An audit of a token or isolated component is not treated as assurance for an entire platform.
Claim verification
Claims are stored separately from scores with their category, original source, evidence, status and analyst conclusion.
Evidence supports the material meaning.
Only part is supported or important limits apply.
Sufficient supporting evidence was not found.
Available evidence does not permit a reliable conclusion.
Existing evidence supports that the wording creates a materially inaccurate impression.
Incidents, score history and corrections
Material updates record the date, reason and relevant evidence. When a genuine update changes Expert Score, Trust Score, Risk Level, Evidence Confidence or Research Status, the previous and new values can be preserved. We do not move scores or dates to manufacture activity.
When a dataset expansion materially changes a study finding, the current study is updated with its new dataset version and research date and the change is recorded. Current research is not silently presented as though the revised statistic had always applied.
A “no qualifying incident identified” entry describes a bounded search, not proof that no incident occurred. Companies may submit evidence and a labeled response; submissions never change a score automatically. Read the corrections process.
Browse the consolidated research changelog or inspect the public dataset summary.
Automation or AI may assist organization and quality checks, but the accountable research desk remains responsible for sources, conclusions, corrections and publication.