# Source and coding notes

Published 21 October 2026. The frozen protocol retains its original pre-publication status and hash. Coding adjudications and supplementary observations are recorded below rather than changing that protocol.

Targeted independent review corrected CBD Dental Sydney (B09) from booking-label-only to expressly advertised direct scheduling. Retained source text describes choosing a treatment and finding a suitable time. Final booking counts are direct 3, request 6, both 1, label-only 10 and unknown 4, out of 24. No rule changed retrospectively and no booking was tested.

The source register preserves its original broad setup/collection interval. The method was written and hashed at 06:47:15.584543 UTC on 20 October 2026 before the first sampled-business search, as established by the execution order. The interval starts at 06:46:31 UTC because it includes setup. The exact first-search wall-clock time was not retained; no precise timestamp attestation is claimed.

Original collection used 72 known public search/page attempts; targeted review added 13. Two supplementary public-source reads on 21 October bring the total to 87, within the frozen 100-request ceiling. Text retrieval can use cached pages. The original HTTP refresh completed 20 of 24 reads successfully; two 403 responses, one redirect-limit result and one partial-response timeout remain limits. HTTP access does not prove a functioning contact route.

## Final pre-publication evidence review

Release review found that generic positive-code placeholders in the public CSV did not retain sufficient per-field evidence. The CSV and source-cues.json now contain specific source IDs, locators and short redacted cues. Some cues come from separately identified original readback or independent-review captures; they are not all attributed to the first extraction. MC Dental's email route is corroborated by its separately identified 20 October review capture, rather than an unretained line from its original homepage extraction.

Three positive route codes at B13 (phone/email) and B21 (phone) lacked sufficient retained original excerpts. Two new HTTPS GETs on 21 October returned 200 with zero redirects and established those advertised contact anchors. They are registered separately as S13R21/S21R21 with response and route hashes in corroboration-2026-10-21.json. They establish 21 October advertising, not reconstructed 20 October captures. The report and charts therefore identify a 20–21 October source window. The sample, definitions, contact-route and booking counts are unchanged.

B05's written-response wording changed from GENERAL_PROMISE to UNKNOWN: the retained quote/booking context did not establish that the general contact promise specifically followed a written enquiry. Its separate request/confirmation classification is unchanged. Final response counts are 2 specific / 6 general / 16 unknown, replacing the earlier 2/7/15 draft. Calculations, table, chart and alt text agree. No response speed was tested.

The evidence omits business phone/email values and personal contact names. Square-bracketed contact-value omissions are redactions, not source wording. Source URLs, dates, short excerpts, locators, hashes and unknowns remain available. Advertised wording is not tested performance. Valory AI maintains this release; corrections can identify the business/source ID via https://www.valory.com.au/contact.