Methodology

Every result combines two dimensions, always shown together: Risk (intensity of negative signals) and Confidence (amount, diversity, and recency of evidence). We never show a bare score without this context.

Factors considered

• Number of independent technical sources and their configured reliability.
• Moderated community reports, weighted by verification level.
• Temporal and geographic diversity of reports.
• Association with campaigns reviewed by a moderator.
• Signal age (signals lose weight over time).
• Open disputes, which cap the displayed risk and confidence.

What we don't do

We don't use an opaque machine-learning model as the scoring core: every point is traceable to a named factor, shown under "Why this result" on every entity page. We don't use categorical labels like "confirmed scammer" or "100% safe".

The scoring engine is versioned: every weight change creates a new version, history is never overwritten. See also our data sources.