How I test
Short version: a product only gets a testing label after it has been at my desk through normal working days. Everything else says so on its face.
Tested at my desk
Bought or borrowed, used daily for at least two weeks, measured and photographed.
Researched, not yet tested
Specs, published measurements, warranty terms and owner reports — no hands-on time yet. We say so rather than pretending.
Tested and rejected
It came to the desk and did not earn a spot. We keep these visible.
The protocol
- Wear or use the product for a minimum of two weeks in a normal working day, not a staged photo session.
- Record the desk geometry it was used with: desk height, monitor height, chair, and hours per day.
- Take the same three photos for every product: front, side at the desk, and the wear point after two weeks.
- Log what actually changed — pain level out of 10 at the same time each day, and how long the effect lasted.
- Publish the failures. A corrector that dug into the armpit after four days is more useful than a spec sheet.
What gets recorded
Every trial logs the same fields: the desk geometry it ran under, hours worn or used per day, the discomfort score at the same time each day, and what went wrong. Those logs appear inside the guide, not in a separate marketing page, so you can check the claim against the record while you read it.
Where the non-tested claims come from
When a product has not been to the desk, the recommendation is built from these sources and nothing else:
- Cleveland Clinic ergonomics guidance
- Mayo Clinic office ergonomics
- OSHA computer workstation eTool
- Cochrane / PubMed reviews on posture correctors and back pain
- BIFMA chair testing standards
- Manufacturer spec sheets and warranty terms, read in full rather than summarised from a listing.
- Owner reports at volume, including the one-star reviews, which is where failure modes live.
Scores
This site does not publish an invented rating out of ten. Where a rating appears it is the retailer's aggregate owner rating, labelled with its source and review count. A number made up by a reviewer who never used the product is decoration, not information.
Corrections
If a pick turns out to be wrong, the guide gets edited and the change is stated in the article rather than quietly swapped. Products that fail testing stay visible with the failure written down.