Seven more sitemap URLs were a publishing observation, not market validation

Key finding

A larger sitemap supports a narrow statement: the publisher advertised more URLs at the later observation. It does not, by itself, support a statement about people finding those pages, using them or paying for anything. In our sample the distinction mattered because there was measurable publishing activity while search and commercial outcomes remained UNKNOWN.

Observation period and sample

We compared point-in-time observations made on 12 October and 19 October 2026. The sample was the same ten publicly listed experiment properties, including two properties on external domains. This was an inventory sample selected by our own portfolio, not a representative sample of internet businesses. We did not sample every day between the observations.

The four unchanged sites were industries.datumharbor.com (19), case.shengyibiaoben.cn (63), free.datumharbor.com (21) and signals.shengyibiaoben.cn (11). Across the ten sitemaps the totals were 205 and 212 URLs. These are advertised URL counts, not counts of original articles or indexed pages. Increases can include pages of different types.

Method and public evidence

For each property we fetched its public homepage, robots file and sitemap with certificate verification enabled. We parsed the sitemap XML and recorded its URL count. We then compared the same field for the same hostname in the two dated observations. The subtraction is reproducible from the sanitized first-party observation dataset. The dataset exposes public domains, dates and counts only; it excludes private traffic records, credentials and internal operational reports.

The comparison describes the snapshots. It cannot identify when each extra URL first appeared or establish that every URL existed throughout the interval. We did not derive a publish date for an individual article from a sitemap total. We also did not count these changes as publications made by DatumHarbor: the properties operate independently.

What this evidence does and does not answer

The operational question was whether the experiment surfaces changed. For six properties the answer was yes on the advertised URL-count measure. The business question was whether those changes helped validate demand. The available evidence did not answer that question. There was no matched query, click, inquiry or transaction series in this comparison.

That gap should remain visible. UNKNOWN means the relevant measurement is unavailable to this comparison. It does not mean there were zero visitors or zero customers, and it is not evidence that an experiment failed. Equally, seven extra URLs are not seven new customer needs. Converting a publishing count into a demand count would change the unit without adding evidence.

A reusable decision rule

1. State the observed unit precisely: here it was a URL listed in a sitemap.
2. Compare like with like: the same property, source field and observation method.
3. Record a technical change without upgrading the experiment lifecycle.
4. Request the next relevant measurement: index coverage for an indexing claim, query and click evidence for search exposure, or attributable inquiries and transactions for a commercial claim.
5. Keep the original observation even when the next measurement is unavailable.

This rule lets us acknowledge real work without overstating its effect. It also prevents the opposite error: treating missing outcome data as proof that no useful work happened. The record can say that publication surfaces expanded and that the business consequence is not yet established.

Limitations

The sample is small and selected. Only two snapshots were compared. A sitemap can change for reasons other than a new research article. We did not audit each new page's originality, claims or conversion path. Search indexation, impressions, clicks, ranked queries, AI citations, leads and revenue are UNKNOWN for this comparison. No causal effect or general success rate can be estimated from these observations.

Conclusion and next measurement

The supported conclusion is seven additional advertised URLs across six observed properties, not market validation. A future result would need a clearly defined outcome, an observation window, an attributable source and its limitations. Until then, we preserve the technical change and leave the outcome open.

See the experiment inventory, our lifecycle and evidence principles, and the companion method on incomplete review coverage.