Units and joins
candidate_id joins panel, route, body, header and late-check rows. It identifies a publisher candidate, not a parent company. Null means the value was not established. It is not zero, no, or a failed website.
publication is the name retained in the collection review and may be null; input_domain always identifies the candidate. category is a supplied editorial stratum: national, regional or specialist. These labels are not weights.

Panel
documented_us_status: supported means the supplied review retained explicit US-audience/edition or US-local coverage evidence; unknown means the documentary threshold was not met. This is not a nationality classifier.
primary_outcome: first-pass collection state, not a live website status or quality rating.
attempt_record_present: a workflow attempt record exists. It does not necessarily mean an HTTP request reached the publisher.
initial_url/final_url: recorded requested start and terminal workflow URLs. A terminal workflow URL need not be a successfully fetched destination.
completed_at_utc: workflow completion timestamp from its source record. It is not a separately measured timestamp for each redirect hop.
body_sha256: checksum of the retained body, including zero-byte captures where present.
measurement_wave: baseline, expansion, or local_completion. Do not treat these as identical network environments.
visible_cross_host_redirect: belongs to the 15 observed cross-host routes in this release.

Primary outcomes
text_with_records: a captured body contains lines matching domain, seller identifier and DIRECT/RESELLER-like fields. Not complete syntax or seller validation.
homepage_unknown: a current homepage host was not established; no ads.txt attempt was completed.
robots_unknown: access policy could not be resolved under that collection wave's rules; a later stage may have stopped after a recorded redirect.
robots_disallowed: policy evaluation stopped the workflow.
blocked: collection recorded a blocked workflow. This does not assert a specific origin HTTP status across all rows.
not_measured_environment_blocked: the execution environment prevented measurement.
http_error: recorded terminal HTTP error.
non_text_response: captured response failed the text screen.
network_error: network-level failure in that observation.
unsafe_redirect: a redirect failed a collector safety prerequisite; not a finding that a website is malicious.

Redirects and headers
origin_host: publisher-side hostname issuing the reported HTTP redirect.
target_host/location: destination hostname and literal Location value from retained headers.
redirect_status: origin response's 3xx code, excluding upstream proxy CONNECT responses.
target_requested: whether the workflow made a request to the reported destination.
terminal_text_observed: the route ultimately returned text passing the record-pattern screen.
observed_at_utc: source attempt completion timestamp; not hop-specific timing.
header_sha256/original_evidence_path: original retained header checksum and archive-relative path. Headers may contain more than the single extracted response. Complete originals are not in this release.

Host summaries
publisher_count: number of observed publisher candidates pointing to that exact hostname. Not all customers, domains globally, or independent companies. Hostname grouping is not ownership grouping.

Late checks
initial_outcome and late_outcome remain distinct. Different execution contexts mean a changed result does not isolate time, geography, proxy or client as the cause.
