Methodology
Every figure in the assessments carries a provenance label. This page explains those labels, the scope of the evidence, and what the reports deliberately do not claim.
Confidence
The distinction is load-bearing: a confirmed figure and a derived figure warrant different decisions.
Read directly from a Cloudflare API response or analytics dataset over the stated window. Reproducible from the same query.
Computed from confirmed figures under assumptions stated at the point of use — bot-score banding, for example. The assumption travels with the number.
Named explicitly rather than estimated. Where a rule body or threshold was unreadable, the assessment says so instead of inferring it.
Scope
Stated up front so a figure is not carried further than it can support.
Presentation layer
This application reads the Markdown assessments and renders them. It does not edit them, generate summaries of them, or convert them to another format on disk. Where the dashboard shows a chart, the chart is a view of the report's own figures — and until the charts are wired to the reports, they are explicitly marked as sample data.
If a number here and a number in a report ever disagree, the report is correct.
Source