FlatAEC research · Method pilot

What gets lost when air-barrier product evidence is republished?

A point-in-time public-web comparison of four manufacturer product records with third-party technical-library pages.

Observed: August 30, 2026
Scope: U.S. · English · public pages
Sample: Four exact product pairs

Third-party pages reliably preserved the product name, broad category, and a product-data link. What they did not preserve in visible landing-page text was the decision context a technical researcher may need to select the correct evidence.

Across the four scored pairs, application conditions, performance or standards, limitations, revision dates, and exact links back to the manufacturer's product record were not visible in any third-party overview.

We call this evidence compression, not misinformation. The supporting facts may still exist inside attached PDFs. The research problem is the additional work required to open, identify, and reconcile those documents—sometimes across multiple variants, jurisdictions, or general manufacturer resources.

Pilot results

Visible landing-page evidence was scored separately from evidence that may exist inside an attached document.

Evidence field	Result	Status
Exact product identity	4 of 4	Preserved
Broad category or function	4 of 4	Preserved
Product-data document link	4 of 4	Available
Clear installation or specification resource	3 of 4	Mostly available
Detailed application conditions in landing-page text	0 of 4	Not visible
Performance values or standards in landing-page text	0 of 4	Not visible
Product limitations in landing-page text	0 of 4	Not visible
Visible revision date for the core product record	0 of 4	Not visible
Exact link to the manufacturer product page	0 of 4	Not available
General manufacturer documents mixed with product documents	4 of 4	Observed

Four product comparisons

Manufacturer pages were treated as the reference record. Third-party pages were evaluated only for visible consistency and completeness.

Pair 01

Observed

Blueskin VP100

Identity survives; the use boundary moves out of view.

Henry identifies the product for wood-framed residential and multifamily construction and exposes data, installation, evaluation, specification, and warranty resources. The third-party page preserves the name and category and lists 16 product-specific documents, but its visible overview does not carry that use boundary, performance values, or limitations. Two entries are simply labelled Product Data without visible revision dates.

Manufacturer source ↗Third-party page ↗

Pair 02

Observed

CCW-705

The overview is accurate but materially thinner.

Henry exposes composition, above-grade use, transition and flashing applications, covering conditions, compatibility, codes, a guide specification, and architectural details. The third-party page accurately summarizes the broad function, while its product-specific set contains product data, SDS, technical notes, and a brochure rather than the fuller implementation record.

Manufacturer source ↗Third-party page ↗

Pair 03

Observed

ExoAir 230

Document volume does not automatically create variant clarity.

Tremco visibly distinguishes 230 and 230LT and states application temperature, exposure limit, service temperature, substrates, application methods, and NFPA 285 context. The third-party page preserves the broad vapor-permeable, UV-resistant description and lists 18 product-specific documents, including U.S., Canadian, 230, and 230LT records, without explaining those conditions in the overview.

Manufacturer source ↗Third-party page ↗

Pair 04

Observed

Sto Gold Coat 81636

One product resolves to three specification paths.

Sto explains Substrate-Driven, Medium-Build, and High-Build paths with different wet-film thickness conditions. The third-party page preserves the exact product number and high-level description and offers seven product documents, but its guide-specification link is specifically High-Build and the overview does not explain that three paths exist.

Manufacturer source ↗Third-party page ↗

Why the variant check matters

A current manufacturer page exists for PERM-A-BARRIER VPS 30, while the third-party record found during discovery was titled PERM-A-BARRIER VPS. GCP's portfolio contains both names.

We excluded the case instead of reporting an obsolete product or naming conflict. Exact entity and variant resolution must happen before comparison.

GCP's VPS 30 guide specification directs users to the current manufacturer record and warns that information on other websites may not be current or location-applicable. That product-specific warning is not evidence that every third-party listing is stale; it demonstrates why source lineage and jurisdiction matter.

What the pilot establishes—and what it does not

Supported

Exact identity and high-level function survived in the four exact pairs.
Important selection conditions were compressed into attachments or left outside visible overview text.
Document volume did not replace product, variant, jurisdiction, revision, and source clarity.

Not established

That an attached document was technically incorrect or obsolete.
That one third-party platform performs worse than distributors or other libraries.
That evidence compression causes rankings, citations, specifications, or sales.
Any engineering, compliance, safety, or project-suitability conclusion.

The next six biweekly observations

The pilot becomes useful when one frozen panel is observed repeatedly instead of replaced with unrelated topics.

01
Baseline
What gets lost when air-barrier evidence is republished?
02
Freshness
Can a researcher identify the current technical data sheet?
03
Identity
How often do product variants become ambiguous outside the manufacturer site?
04
Conditions
Do third-party pages preserve application limits and exclusions?
05
Durability
Which technical-document links change or break over eight weeks?
06
Quarterly update
What changed, what stayed stable, and which corrections were made?

Method and publication standard

Selection

Current public U.S. manufacturer product page.
Exact product name or number on the third-party page.
Air, vapor, weather, or water-resistive barrier use.
Ambiguous variants recorded but excluded from scoring.

Capture

Record date, URL, title, product identifier, and source type.
Score visible HTML separately from attached documents.
Re-open consequential sources on publication day.

Quality control

Resolve entity and variant before scoring.
Re-score at least 20% of a full panel independently.
Publish the field dictionary, raw rows, exclusions, and corrections.

Interpretation

Use observed, not visible, and not evaluated precisely.
Do not infer technical truth from a marketing summary.
Do not convert association into ranking or commercial causality.

Research disclosure

This independent pilot used public sources only. The named companies were not contacted and did not participate. Findings describe what was visible at the cited URLs on the observation date. No engineering, architectural, legal, code, certification, safety, compliance, or project-suitability determination was made.

The method is designed to produce original, inspectable research rather than a universal score. Google's people-first guidance emphasizes original research or analysis, clear sourcing, authorship, and an explanation of how content was produced. See Google Search Central's people-first content guidance and its guidance on AI-assisted content.

Request a correction or review the evidence policy

Where does your product evidence become difficult to verify?

Name one product family, one third-party path, and the buyer decision that should follow.

This work cannot guarantee search rankings, AI citations, specifications, quotes, shortlists, contracts, leads, pipeline, or revenue.

Request a Spec Audit