Editorial
policy
The rules we hold ourselves to, written down so you can hold us to them as well. Where we break one, the corrections log below is where it gets recorded.
Complaints about a specific review go to [CORRECTIONS EMAIL]. Cite the page, quote the passage, and say what you believe is wrong.
Principles
- 01
Every figure carries a record
No percentage appears anywhere on this site without a reference number that resolves to an entry in the evidence register, giving design, blinding, sample size, duration and endpoint. If we cannot source a figure, we do not print it.
- 02
Design outranks sample size
A randomised, vehicle-controlled trial of 60 people tells you more than an uncontrolled survey of 18,000. We grade on design first and say so in the grade key.
- 03
We separate fact from assessment
What a brand publishes is reported as fact and sourced. What we make of it is labelled as assessment. Readers should always be able to tell which is which, and to disagree with the second without disputing the first.
- 04
We publish what we could not assess
Reviews state their method on the face of the page. Criteria we could not evaluate are left blank and the coverage percentage is printed beside the score.
- 05
Brands get no advance copy
We will put factual questions to a company before publication and print the answer or the silence. We do not share drafts, scores or embargo dates, and we do not accept conditions in exchange for access.
- 06
Nothing is unpublished on request
A brand may dispute a fact and we will correct an error. A brand may not have a review taken down, edited into approval, or quietly de-indexed because it is unflattering.
Evidence grading
Randomised or split-face, controlled, blinded to at least the investigator.
Controlled but unblinded, small, or ex vivo. Directionally useful.
In vitro, open-label or uncontrolled survey. A pathway, not a result.
A manufacturer’s own customer satisfaction survey is a grade C record however many people it covers, and we say so in the review rather than in a footnote. Sample size is not a substitute for a control arm.
THE EVIDENCE REGISTER →Corrections
When we get something wrong we fix the page and log the fix here, with what was wrong and whether the score moved. We do not silently edit a published review.
Substantive corrections also carry a dated note on the review itself. Typographical fixes do not.
LOG ENTRIES BELOW ARE PLACEHOLDER STRUCTURE. REPLACE WITH REAL CORRECTIONS OR EMPTY THE TABLE BEFORE LAUNCH.| REF | DATE | WHAT WAS WRONG | ACTION |
|---|---|---|---|
| COR-003 | [DATE] | [WHAT WAS WRONG] | [WHAT WE CHANGED, AND WHETHER THE SCORE MOVED] |
| COR-002 | [DATE] | [WHAT WAS WRONG] | [WHAT WE CHANGED] |
| COR-001 | [DATE] | [WHAT WAS WRONG] | [WHAT WE CHANGED] |
Disclosures
- No contributor holds equity in, consults for, or is employed by a brand in the register.
- Where a prior working relationship exists, it is declared on the review itself, not here.
- Contributors do not accept travel, events or hospitality from brands under review.
- Panel testers are compensated for their time by us, never by a brand.
- We do not publish generated text as editorial copy under a human byline.
- Where tooling assists with research, structuring or code, a human verifies every factual claim against the primary source before publication.
- No score, grade or verdict is produced by a model.
- Product photography and test data are never generated. If we did not measure it, we do not show it.
The two lists above describe a standard, not an audit. Before launch, confirm each line is actually true of your operation and delete any that is not. A disclosure page that overstates is worse than no disclosure page.