Version 2026-08-27
OpenWebmaster evidence methodology
Every public fact names its source, observation or retrieval time, confidence basis, and limitation. Missing evidence remains unknown.
Four evidence planes never merge silently.
Provider operations, workspace-private evidence, public-web facts, and opted-in research aggregates have separate visibility. Search Console, first-party logs, tokens, and analytics never enter a public profile by default.
Source-specific truth.
Cloudflare Radar is an aggregate view of traffic observed by Cloudflare, not site analytics. Common Crawl proves appearance in a named collection, not Google crawl or indexing. CrUX is eligible Chrome field data aggregated over its stated 28-day period. RDAP and CT are identity and issuance evidence, not legitimacy proof.
Correlation language stays conditional.
Timeline adjacency is reported as “coincided with” or “is consistent with.” Causal wording requires a declared experiment, comparable control, observation window, sample size, and limitations.
Open Web Trust is observed, not declared.
The browser submits only a public URL. The Worker derives technical, search, transport, crawler, and eligible RDAP identity points from source-labeled observations. Weights: technical integrity 20%, search readiness 20%, identity 15%, security 15%, stability 10%, provenance 10%, crawler access 10%. Missing dimensions are excluded while coverage falls. Contribution has 0% weight. The output is not Domain Authority, a Google signal, or a ranking prediction.
Research consent and privacy.
Indexing-latency consent is purpose-specific, versioned, and revocable. Contributions contain categorical buckets and durations—not URLs, domains, queries, or provider payloads. Revocation removes the workspace's unaggregated rows. Public groups require at least five distinct workspaces.
Exports remain reproducible.
The source registry is versioned as JSON and CSV. The latency dataset discloses period, sample size, privacy threshold, method, and limitations.