# How these figures were read > Sources, the two-pass reading rule, what normalization refuses to convert, and when a product page is allowed to open. Source: https://flooringcompared.com/methodology Documents last read: 2026-11-09 Nothing here is tested. Every figure is a manufacturer's own declaration, read from a document the manufacturer published, and reproduced beside a link to that document. Where two brands measure the same property under different standards, this reference says so and refuses the conversion. ## Where the figures come from Three documents per product, and only these three: the technical data sheet, the installation instructions, and the warranty. They are the primary sources. Not the product page, not a retailer's listing, not a distributor's summary: each of those paraphrases, and drops conditions on the way. A figure that appears on a marketing page but in none of the three documents is not published here. The three carry different things and are not interchangeable. Wear layer, thickness and test results come from the data sheet. Moisture limits, flatness tolerance, expansion gaps and the maximum run without a break come from the installation instructions, usually several pages in. What the warranty excludes comes from the warranty, which is the document buyers read last and the one that decides what happens when something goes wrong. ## Read twice, published once Every cell is read twice, independently, from the same source document. If the two readings agree, the cell is published. If they disagree, nothing is published for that cell. It goes to a queue a person resolves against the document. The gate exists because these numbers sit on page three of a PDF next to four other numbers that look like them, and one reading pass has no way of noticing it took the commercial figure where the residential one was wanted. Agreement is checked after normalization, so a pass that recorded 0.508 mm and a pass that recorded 20 mil agree. It is deliberately not checked across units that measure different things: a maximum run declared in linear feet against one declared in square feet is a disagreement about what the document said, even though the two can be bridged for display once it is clear which one the maker actually wrote. Here is what this gate cannot see, stated plainly, because a check whose blind spot is unstated is worse than no check. Two readings that make the *same* error agree, and agreement is exactly what this treats as proof. Reconciliation catches careless reading; it does not catch a misleading document or a shared misconception. So one product per brand is also checked by hand against its source, chosen by a fixed rule rather than picked, and that manual check is the only thing standing underneath the blind spot. ## What normalization does, and what it refuses Three kinds of incomparability turn up in these documents and they are not the same problem. The first is one value written twice: 20 mil and 0.508 mm. It is converted and both are kept. The second is a lossy bridge: a maximum run of forty linear feet and one of 2,500 square feet both bound the same installation, but neither implies the other without an assumption about the shape of the room. Those are converted *and* the assumption is printed with the number, because a converted figure whose assumption is hidden is worse than two unconverted ones. A long narrow room reaches the linear limit first, always. The third has no bridge at all. A five pound calcium chloride limit and a ninety-five per cent relative humidity limit describe different volumes of the same slab measured over different periods; a slab can pass one and fail the other. There is no valid conversion between them, so both are published side by side, marked as not comparable. The same holds for slip resistance, where three different standards produce three different numbers from one floor, and for acoustic ratings, which describe a floor-ceiling assembly rather than a plank. ## What an absence is published with A cell reading "no declaration" is a claim about the manufacturer, not a hole in this table, so it is published with the evidence for itself. Of 222 undeclared cells in this release, 194 carry a sentence saying what was read and what was found there instead, and 46 of those also quote the document's own line, usually a figure of a different kind standing where the field would be: a dry static coefficient where a wet dynamic one was wanted, a delta against a reference slab where an impact class was wanted, a pass mark where a load in psi was wanted. The remaining 28 cells carry neither, and they are left that way on purpose. On those the one pass that held a document recorded the verdict and nothing else; writing the sentence now would mean inventing a reading nobody performed. The number is published instead, and a gate holds it: it fails if another appears, and it fails if one is resolved without the count being moved with it. ## The test behind a moisture figure Every moisture figure here carries the test it was taken under, read from the same document as the figure. Of eight makers in this release, seven declare a subfloor humidity ceiling, six of them naming an ASTM designation for it, and one naming only the apparatus the reading is taken with and no standard anywhere in its instructions. Those are not the same disclosure, and a table that prints one standard down the column reports its own inference as a maker’s declaration. Two limits that were taken by different procedures do not convert into one another. A slab can pass an in-situ probe and fail a calcium chloride test on the same day, so the comparison a reader can make down this column is between makers who named the same test. The column says which those are instead of assuming it. ## When a product gets a page Where a manufacturer publishes nothing for a field, that is recorded as a finding and shown in the table rather than left blank. A blank reads as "the maker says nothing" when it may mean "we have not looked", and those are different claims. Coverage is printed as a count on every product page, whether or not the count earns its badge. The page template was approved once. Individual pages are not: a product page is created only when its evidence clears a fixed bar, and waits in a queue until then. 10 fields must be present, each one carrying a source URL and the date the document was read. A queued cell on any of them keeps the page shut. This is what separates a reference from a page generator, and it only means anything because the build enforces it rather than the editor remembering to. ## Prices, where they appear There is no price on this site, and the reason is a measurement, not a preference. The only figure worth publishing here is one a manufacturer publishes itself: an advertised-price policy with a list behind it. That is a rule about how a product may be advertised, not a transaction price. Of eight makers in this release, one publishes both the policy and the list, one publishes the policy and keeps the list to its sellers, two publish no policy at all but sell direct and print their own retail price, and for four no manufacturer-hosted price document was found. That is a fact about the search done here, not proof that none exists. The two direct prices are not the same quantity as the advertised-price list, which is why they do not rescue the column. One is what a single seller charges today; the other is the floor below which any seller may advertise. Putting them in one column would compare unlike things while looking like a market, and a column one maker in eight can fill is not a comparison in the first place, and its blanks would be read as cheaper. If a second maker publishes a list, the decision reopens. ## What is withheld No ranking, no score, no "best". Several of these fields are not commensurable, and a composite across them would be an opinion wearing a number’s clothes. No brand pays for inclusion. No fitting advice and no labor figures: whether a floor suits a particular room is a question about that room, and the only thing this reference can honestly contribute is what the manufacturer says its own product requires. The name of the publisher is withheld by the operator's decision. What is withheld is the name only: the disclosure itself stands in the footer of every page, unchanged. Whoever publishes this reference sells the category it describes. The consequence of that decision is stated, not hidden. There is no published contact channel at present, so corrections cannot be sent to us directly. What stands in its place is structural: every figure links to the document it came from, at the page it came from, so any reader can check any cell against the primary source without asking us anything. The full disclosure, and what is counted about readers, is set out [on the about page](/about). ## This release The reuse terms are set out once, [on the about page](/about), next to the machine-readable release they apply to: what may be quoted, and what has to travel with it. They were written out here as well until the two wordings began to drift, which is the failure a reference about drifting documents should be least willing to commit. *This release, counted rather than described.* | Measure | Count | | --- | --- | | Fields in the dictionary | 56 | | Mandatory before a page opens | 10 | | Brands | 8 | | Products with a page | 16 | | Cells held back pending resolution | 20 | Field definitions: [the dictionary](https://flooringcompared.com/fields) · machine-readable release: [spec-v1.json](https://flooringcompared.com/data/spec-v1.json) --- Every figure on this site is a manufacturer’s published declaration, reproduced with a link to the document it came from. Quote it with its citation and its condition: a maximum run without its declared unit, or a slip figure without its test method, is not the fact. --- This reference is compiled and published by **the operator of this site**, a materials-only flooring retailer in the United States. It sells the category it describes here. No brand pays for inclusion and no brand is ranked; no figure attributed to a manufacturer is the operator’s own measurement, and each is reproduced with the document it came from. The counts and coverage figures are this reference’s own arithmetic over those declarations. The operator sells flooring materials. It does not sell or arrange fitting, and this site publishes no labor figures and no offers. Page views are counted with Google Analytics, along with how long each page stays in view. There is no filter, no form and no comparison box on this site. Nothing a reader could type into. [What is counted](/about#measurement). [Methodology](https://flooringcompared.com/methodology) · [Field dictionary](https://flooringcompared.com/fields) · [JSON](https://flooringcompared.com/data/spec-v1.json) · [llms.txt](https://flooringcompared.com/llms.txt) · [Privacy](https://flooringcompared.com/privacy) Documents last read 9 November 2026. This release carries 16 published products across 8 brands.