# Frozen method: public website enquiry options

Status: local research draft only; no publication authorised.
Version: 1.0. Frozen UTC: 2026-10-20T06:47:15.584543+00:00

Question: What contact routes, appointment/request distinctions and written response expectations are advertised on a small, search-discovered set of Australian service-business websites?

Evidence class: observed public website content on the access date. This is neither observed call behaviour, a survey, a model, a representative national estimate nor a product test.

Design: search-based convenience sample, target 24 businesses: 3 industries (plumbing, dental, accounting), 4 metros (Sydney, Melbourne, Brisbane, Perth), 2 distinct businesses per cell. Industries span urgent service, appointment-based service and professional enquiries; metros provide geographic variety without implying coverage. Equal cell sizes are design choices, not population weights. No customer/tenant data used.

Discovery order: industry order above, then metro order above. Exact queries: `{industry} {metro} contact`. Use the public web search tool once per cell. Preserve returned source order and query output. Take the first two eligible distinct businesses encountered in the returned order (not based on contact features). If fewer than two eligible businesses, make one replacement query `{industry} {metro} contact local business` and use its returned order. Record all returned candidates up to the second included business and all attempted candidates. Search rank means order surfaced by this tool, not a Google ranking.

Eligibility: a primary business website offering the named service and explicitly serving/located in the metro; a readable primary page supports identity/location/service. Exclude directories, aggregators, articles, job/recruitment pages, unrelated services and duplicate business/brand domains already included anywhere. A multi-location brand can enter only once. No exclusion based on contact-route result. An inaccessible primary site is a failed attempt and replaced by the next candidate, unless another readable primary page from the same domain is available within budget. Do not bypass restrictions. If the cell remains short, disclose achieved denominator and the shortfall; do not hand-pick.

Public page budget: at most 100 page/search requests in total, including 12 initial query requests, up to 12 replacement searches and a maximum of 3 primary pages per included business. Stop at 100 even if incomplete. HTTP redirects count in the local HTTP log; any HTTP request and web open of the same URL count separately. Search-engine internal fetches are unknowable and not described as measured. Maximum concurrency: 2. Prefer one contact/home landing page, then an explicitly linked booking/contact page where needed; no exhaustive crawl. External booking destinations may be observed as outbound links but not visited. No forms submitted, calls, account/session access, paid APIs, private records or browser interaction.

Coding unit: business. Every positive code must cite source ID and a short exact excerpt or advertised link label. Collect date/time, source URL, tool/access method and evidence. Store only relevant public text excerpts/labels; remove phone/email values and personal contact names from cached excerpts where not needed.

Predefined fields:
- phone_advertised: YES if explicit phone/contact-by-call route is displayed; UNKNOWN if not established.
- email_advertised: YES if email route/address is displayed; UNKNOWN otherwise.
- enquiry_form_advertised: YES if a general enquiry/contact/quote form or explicit link is displayed; UNKNOWN otherwise. A booking-request form is also a written contact route but separately coded.
- appointment_or_job_path: DIRECT_SCHEDULING_ADVERTISED only if a primary page explicitly advertises an online booking/appointment/slot-selection flow; REQUEST_ONLY_ADVERTISED if it explicitly says request/book via a form or callback that staff will arrange/confirm, without a direct flow established; BOTH if both are explicit; UNKNOWN otherwise. These are advertising observations, never proof of completion or live availability. A bare Book Now button alone is BOOKING_LABEL_ONLY, not proof of direct scheduling.
- written_response_expectation: SPECIFIC_TIME if a reply/callback timeframe is explicitly attached to a written enquiry (quote words, do not assume compliance); GENERAL_PROMISE if reply is promised without timing; UNKNOWN otherwise. Opening hours, emergency availability and appointment time are not response promises.
- chat_advertised: YES only for an explicit primary-page live-chat/chat route; UNKNOWN otherwise, including script widgets not visible to text access. Never code absent because text lacks a widget.
- observation_notes: ambiguity, accessibility/JavaScript limits, observed route relationship.

No negative absence claim: UNKNOWN means not established within these pages/access method. YES counts use the full achieved business denominator and UNKNOWN shown separately. No imputation. Summarise raw n/N; percentages descriptive only. No significance tests, confidence intervals, causal claims, missed-call rates, response-speed tests, revenue estimates, industry league table or national prevalence. Report cell counts and unknowns. No retrospective method changes; deviations separately dated.

Reproducibility: CSV business rows with per-field source/evidence pointers; source register with excerpts/time; discovery/attempt log; standard-library Python counts and editable SVG charts; report with limitations and Valory commercial involvement. Hash frozen method before collection and include it in final package.