How this site verifies what it publishes
Every figure here is read off a manufacturer's own document, then read again by someone who sees only that document and nothing we recorded. Where the two readings disagree, nothing publishes.
Checked 63 manufacturer manuals on file Spot an error?
The reason to explain this in public is that it is checkable. Every count on this page is read from the repository when the site is built, so it cannot claim coverage the data does not have.
Three stamps
- Captured63manufacturer documents
- Read twice, and agreed96of 209 fields
- Documented absences10answers no maker publishes
Pass one: capture
A figure is pulled from the manufacturer's official document, never from a reseller's rehost when an official copy exists and never from memory. The file is archived here with its revision code, and the field records the source, the section, and the date it was captured.
Pass two: the second reading
A separate session is given only the archived documents and asked what they say. It does not see what we recorded. Exact match is the only pass condition: no averaging, no judgement calls, no splitting the difference. Where the two passes disagree the field stays unpublished until a person resolves it against the document.
This catches things that careful reading does not. It has already caught a warranty term attributed to the wrong clause, a part number that was the housing rather than the cell, and a figure we had marked verified that a second document in our own archive disproved.
Recording what a manufacturer does not say
Some of the most useful facts in this category are absences. Haywardpublishes no rated cell life anywhere we have looked, and that is worth knowing before you try to work out cost per hour. An absence here is a claim like any other: it names every document searched, and it is verified by a second check returning the same nothing. It is never an excuse for a blank.
An absence is also scoped mechanically. A claim about a brand has to be checked against every document we hold for that brand, because a narrow check under a broad sentence is how a false statement passes verification honestly.
Corrections
If something here is wrong, it gets checked against the primary document and logged either way, including when the check finds we were right. Corrections are recorded with the date and what changed, in the changelog, rather than quietly edited.