How we research

Where the numbers come from

Every dimension on this site is computed from a single file of constants rather than typed into a page by hand. That is not tidiness, it is accuracy: when the same figure is written into forty pages independently, some of those copies are wrong, and nothing catches it. Because the calculators and the fit tables read the same constants, a page and the tool it links to give the same answer by construction.

Each figure is labelled a standard or a convention. Standards are published: ANSI/HFES 100 for workstation fit, ANSI/BIFMA X5.1 and X5.11 for chair testing, X5.5 for desks, G1 for the ergonomics guideline, plus manufacturer specifications and plain trigonometry. Conventions are rules of thumb, principally the stature multipliers. Where no documented figure exists, the calculators return nothing rather than inventing one.

How products are sourced

The catalog holds 316 records across 57 categories, ranging from $8.79 to $1585.00. Each one came from a real API response, and only factual fields are stored: the identifier, title, brand, price, image and any load rating stated in the title. No affiliate URL is stored, because the link is assembled when the page renders.

A record is rejected, not corrected, when:

• it has no price, or no image, since a card with an invented figure is worse than no card;
• its price falls outside a sanity band for its category, which almost always means the search returned something other than what was asked for;
• it is an accessory returned by a query for the product itself, such as a seat cushion surfacing under a chair search;
• it is a renewed or used listing, whose price and availability are too volatile to anchor a recommendation.

The mistake this process exists to prevent

Catalogs like this are often deduplicated by title similarity, keeping one record per apparent product. On this catalog that would be actively destructive, because Amazon truncates long titles on its own. Two genuinely different chairs, at $788 and $862, arrive with byte identical titles. A similarity check would treat them as one listing, keep the cheaper, and silently delete the other.

So records here are deduplicated by their Amazon identifier and by nothing else. A model designator, a size, a load rating or a motor count is a real difference between products, and it is usually the entire reason the dearer one costs more. Where two records do share a title, the page says so and shows the price that distinguishes them.

How products are chosen for a page

Grids sample across the whole price range of a category rather than showing the cheapest few. Sorting by price and taking six is the easiest way to make an entire category look cheaper than it is, and it hides exactly the products a reader researching a serious purchase came to see. Roundups present three tiers, and the top tier is the dearest genuine option rather than a capped one.

What we do not do

We do not test products in person and we never write as though we had. Phrases like "we tested" and "hands on" do not appear on this site, because they would not be true. The single exception is the long term review of the Autonomous Desk 5 Pro, which is first person because the author has owned and used that desk, and has owned that brand's desks since around 2015. It is labelled, and it discloses that the desk was supplied for review.

Prices shown are the ones returned when the page was last built and they change constantly. Every product links to its live listing, which is the price that counts.