Editorial standards and testing method

Every factual claim on this site has to come from one of five evidence classes. If it fits none of them, it does not get published.

1. The five evidence classes

Class A

Direct test

We ran it. A generated file, a clean installation, a captured result with the version and date recorded. This is the only class that supports a compatibility claim.

Class B

Vendor documentation

The vendor states it in public documentation. Adequate for pricing, system requirements and feature lists — not for whether a document renders correctly.

Class C

Published standard

Defined by a specification such as ISO/IEC 29500 for Office formats or ISO 32000 for PDF. Used to explain what the format requires, not what a product does.

Class D

Independent corroboration

Reproduced across independent sources — bug trackers, developer forums, other testers. Treated as a lead, not a result, until we test it ourselves.

Class E

Labeled as unverified

Anything we have not reproduced is written as uncertain, in plain words. We do not dress a guess as a finding to make a paragraph read better.

2. What a test means here

A test is: a file we generated, with known formatting; opened in a clean installation with no add-ins and no custom templates; observed directly; recorded with the build number and the date. Anything less is a demonstration, and we label it as one.

We capture screenshots and keep the files. If a claim on this site is challenged, the evidence either exists or the claim comes down.

3. Limits we accept

4. Commercial independence

Affiliate commissions fund this site. They do not set scores, soften caveats, or remove a limitation from a page. The rules we hold ourselves to are on the affiliate disclosure page.

5. Corrections

Reproducible corrections are investigated. If a finding changes, the article is updated and the change is noted rather than quietly edited — a review that silently rewrites its results is worse than no review.