Evidence and methodology

What we use as evidence

We go to the primary document: the code section, the standard, the evaluation report, the test, or the agency guide, rather than an article about it. Every published article ends with a numbered source list, linked where the document is online. We group the sources by who published them:

• Government and public agencies. Federal and state agencies, statutes, regulators, and national laboratories.
• Codes, standards, and evaluation reports. Model codes, consensus standards, ICC-ES evaluation reports, and the IBHS FORTIFIED standards.
• Research and field studies. Peer-reviewed papers, university theses, laboratory reports, post-storm field studies, and policy research.
• Industry and trade groups. Trade associations, industry manuals, and consultancies. Useful, and treated as industry sources, not independent ones.
• Manufacturers. A maker's own literature. Treated as the maker's claim, not as independent confirmation.
• Insurers. Insurer and insurance-pool guides and program pages.
• News and trade press. News reports and construction or insurance trade press.
• Other. Everything else, listed by web address: law firms, brokers, nonprofits, forums, and reference sites.

Manufacturer and trade-association material can be useful, especially for how a system is installed. Our rule is to label it as such and not to treat it as independent confirmation. No company pays to be listed in the library.

How a claim is checked

• Independent drafts, compared claim by claim. Major articles start as several independent AI research drafts. Where they disagree, we go back to the primary source.
• Read, not summarized. Our rule is that a page says a document supports a claim only if the document was opened and read. Where we could not confirm something, the page says so, or the claim is left out.
• No unsourced numbers. No performance, cost, or savings figure is published without a source attached. Figures we cannot confirm are cut.
• No winner by default. Comparisons name trade-offs. Where no source ranks two systems, the page says that rather than picking one. The comparison sheet stays unscored until a source is attached to each cell.

What a source shows, and what it does not

A product test describes the tested product under the test’s conditions. An assembly rating belongs to the assembly, not to one board in it. Neither tells you how a whole house will do in a storm, which also turns on the roof, the openings, the connections, the installation, and the site. We try to say which of those a source actually covers. Even IBHS, which writes the FORTIFIED standards, says that no home is 100% “hurricane-proof” (FORTIFIED FAQ).

Standards are not wall systems

Some names on this site are ways to build a wall: ICF, SCIP, CMU, SIP, or a wood frame. Others are standards a house is built or re-roofed to. FORTIFIED is the second kind. IBHS describes it as “a voluntary construction and re-roofing Program”, and a FORTIFIED designation is a written certification, issued by IBHS after an evaluator’s documentation is reviewed, that a home was built, renovated, or re-roofed to the standard (FORTIFIED FAQ). IBHS says projects permitted on or after November 1, 2025 must meet the 2025 FORTIFIED Home Standard (FORTIFIED technical documents).

When a page here says “FORTIFIED frame”, it means a wood or light frame built to that standard. FORTIFIED is not a wall material, and it is not something we sell.

How to read a page

• Sources. The numbered list at the end of each article. Short references in the text, such as (FEMA FS 26, p. 1), point to an entry in that list.
• Evidence trail. Each compare page opens with a summary of its sources grouped by the type of publisher, its review status, and its tagged statements. The grouping is read from who published each source. It is not a quality score.
• Claim tags. Some statements carry a tag in brackets:
  • [code]: the statement rests on a building code, a standard, or the code path for that system.
  • [tested]: the statement rests on test results, field data, or a published study.
  • [manufacturer]: the statement comes from manufacturer or trade literature.
  • [judgment]: the statement is our reasoning from the sources, not a finding stated in them.
• Review status. A page shows “Expert review in progress” until a named reviewer signs it. Today 0 of 32 published articles are signed. Reviewers will be named, with their credentials, on the pages they sign.
• Last updated. Each article shows the date it was last changed.

Publishers we cite

Built from the source lists of the 32 published articles: 596 citations in all, of which 4 have no web link and are not counted below. Each line links to one document we cite from that publisher. Publishers we have not sorted into a type are listed under Other by web address.

Corrections and limits

If something is wrong, tell us on the contact page. Confirmed errors are fixed, and methodology changes are logged on the changelog.

Everything here is education, not engineering advice or a stamped design. Build with licensed professionals. How we are paid is on the disclosures page.