Research question: Do the 12 campaign article files named by the publication-quality test contain operational, claim-level source records, and what evidence controls should govern their replacement?
Audit scope and denominator
This is a bounded Real Estate Luxury repository audit, not a survey of the luxury real estate industry. The denominator is the 12 article slugs explicitly listed in lib/__tests__/campaign-content-quality.test.ts. The observation unit is one MDX article file. The audit used the file contents present at review time; it did not inspect a content-management system, unpublished notes, analytics, approval messages, source documents stored elsewhere, or the conduct of individual editors.
The question is deliberately narrower than whether the prose sounds responsible. A statement that editors should record a source is not itself a source record. For this audit, an operational register had to connect a material claim to retrievable evidence and a review decision. At minimum, that meant a claim identifier or quoted claim, a source owner or publisher, a URL or internal record locator, a date checked, an approval state, and an accountable reviewer. A structured row could be embedded in the article or maintained through an unambiguous locator, but general workflow prose did not qualify.
“External source URL” means an HTTP or HTTPS destination in a Markdown link in the article body. Frontmatter canonical URLs do not evidence article claims and were excluded. “Substantive words” follows the campaign test method: remove Markdown headings, replace linked text with its label, lowercase the remainder, and count alphanumeric tokens, including internal apostrophes or hyphens. The threshold is 600 words per article, not 600 words averaged across the set.
Rubric and reproducible method
The file list came from the campaign test rather than publication-date inference. Frontmatter was separated before body measures. Each file was checked for existence, substantive length, ordered H2 sequence, external Markdown destinations, and at least one use of source, evidence, or record language. A separate scan counted explicit approval or review terminology.
Those lexical measures do not prove an operational register. A public procedural guide may appropriately contain no claim-level row because it makes no property-specific assertion. When evidence is private, a controlled locator can be more appropriate than a public URL. The audit therefore reports visible guidance and distinguishes it from a future transaction record that should connect a specific proposed claim to its evidence, status, and reviewer.
Measured audit results
| Measure | Result | Denominator and rule |
|---|---|---|
| Expected article files present | 12/12 | All named campaign paths exist. |
| Files reaching 600 substantive words | 12/12 | Per-file campaign tokenizer. |
| Files with at least five H2 headings | 12/12 | Structural minimum. |
| Distinct level-two heading sequences | 12 | Unique ordered sequences across 12 files. |
| Files containing source, evidence, or record guidance | 12/12 | At least one bounded lexical match in body copy. |
| Files containing explicit approval or review terminology | 12/12 | Every body contains at least one bounded approval or review term. |
| Files providing an external source URL | 0/12 | HTTP or HTTPS Markdown destination in body copy. |
The results show substantive, independently structured guides with evidence-oriented language. The zero-link count is not a defect for this set: the articles avoid externally verifiable numerical, legal, market, or performance claims and use generalized examples. Real property documents, owner instructions, inspection records, vendor invoices, or showing protocols may require restricted access rather than publication as hyperlinks.
What an operational register should capture
A practical register should use one row per material factual claim, not one row per article. Suggested fields are: claim ID; exact claim text; claim type; source title; source owner or issuing body; public URL or controlled record locator; source publication or effective date; date checked; relevant excerpt, page, or dataset field; geography and time period; known limitation; editor; subject-matter approver where needed; approval state; publication destination; next review date; superseded-by or correction ID; and retention status.
That structure separates several decisions that boilerplate can blur. Evidence answers “what supports this wording?” Approval answers “who accepted it for this use?” Freshness answers “when was it last checked?” Authority answers “is this person permitted to release a client or property fact?” None substitutes for another. A source can be authentic but stale, current but outside its geography, or accurate yet not authorized for public disclosure.
The register should also preserve negative decisions. If an editor removes “minutes from the airport” because no route, traffic convention, and check time are defined, the decision row prevents the unsupported wording from silently returning. If a property document is replaced, the previous locator should remain associated with the superseded claim rather than being overwritten. Corrections should identify the affected claim, prior wording, corrected wording, discovery time, correction time, responsible reviewer, and every channel requiring an update.
Data sources and references
The following ten references were selected for editorial sourcing, correction, records, publishing, provenance, and preservation, not for housing-market statistics. Each destination was retrieved from the named publisher during source verification. Availability confirms that a page was reachable; it does not make every recommendation legally binding or guarantee future availability.
- The Society of Professional Journalists Code of Ethics says journalists should verify information before release, use original sources when possible, identify sources clearly, update and correct information, and acknowledge mistakes promptly. Those propositions support source identity, checked date, and correction fields.
- Reuters Journalistic Standards place accuracy ahead of speed, discuss source credibility, prefer named sources in most circumstances, and call for prompt, clear, transparent corrections. They support a documented review gate and correction trail, not a promise that a workflow will prevent errors.
- The BBC Accuracy Guidelines call for first-hand sourcing where possible, checks on facts and statistics with important caveats, authentication of documentary or digital material, and corroboration of claims. A register can record those checks without claiming that one source is always sufficient.
- The NPR Ethics Handbook describes accuracy, transparency, and accountability, values information gathered and verified by the newsroom, and says errors should be corrected and learned from. It supports recording both verification and what happened after an error.
- The Online News Association’s Corrections ethics module examines whether a correction should be acknowledged where the error occurred and how corrected information should reach platforms where the original spread. It supports channel and correction-status fields rather than silent replacement.
- The U.S. National Archives’ Universal Electronic Records Management Requirements organize federal electronic-record needs around capture, maintenance and use, disposal, transfer, metadata, and reporting. This study borrows those lifecycle categories as design prompts; it does not claim that federal records rules govern a private editorial register.
- The National Archives Guide to the Inventory, Scheduling, and Disposition of Federal Records establishes inventory and disposition as distinct records-management activities in its federal context. For an editorial team, the cautious analogy is to inventory evidence and assign retention status, subject to the team’s actual legal and contractual duties.
- The Library of Congress Recommended Formats Statement publishes format preferences, an evaluation matrix, and a change log for preservation planning. It can inform durable evidence-format choices; it cannot certify the truth of the content preserved.
- The Government Digital Service guidance to plan new GOV.UK content separates decisions about whether and where to publish and places them within user-need and content-management guidance. It supports recording audience, purpose, and destination before approval, while applying directly to GOV.UK rather than this site.
- The W3C PROV overview defines provenance in terms of the entities, activities, and people involved in producing data or another thing so that quality, reliability, or trustworthiness can be assessed. That model supports linking a claim, evidence object, editorial action, and responsible person without asserting that metadata proves the underlying claim.
Together, these sources support a control design: identify evidence, verify within its scope, record who acted, preserve usable records, and make corrections visible. They do not establish that the 12 articles are factually wrong, that a particular record-retention period is legally required, or that adoption will improve rankings, inquiries, sales, valuation, or client trust.
Release and maintenance criteria
The current set passes the campaign's file, length, structure, filler, and similarity controls. Source-register acceptance remains claim-dependent. A material externally verifiable claim should resolve to a row with a source owner, exact URL or controlled locator, checked date, scope note, editor, and approval state. Private facts should use access-controlled locators and record publication authority. Public links should be retrieval-checked, with redirects or failures logged.
Every published correction should connect the original claim, new wording, reason, approver, timestamp, and affected channels. Expired or superseded evidence should trigger review rather than automatic republication. If a guide makes no external or property-specific factual claim, an empty public URL register can be truthful; it must not be backfilled with generic citations merely to raise a count.
Limitations and inference boundary
This was a structural content audit of 12 local files. It did not fact-check a live property proposition, inspect private evidence, test access permissions, interview editors, or measure whether a source register changes error rates. Markdown URL counting misses evidence held outside the files, while lexical scans can miss equivalent wording. Reachability is time-sensitive, and an HTTP response does not prove a page is correct or suitable for a particular claim.
The supported inference is narrow: the current guides contain evidence-oriented instructions without pretending that private records are public citations. A claim-level register would make later property-specific review more reproducible. This audit cannot establish that implementation will eliminate errors, satisfy every jurisdiction, protect confidential information, improve rankings, or produce commercial outcomes.