How we rank
A rank is an argument. Here is ours, in the open, so you can disagree with the weights instead of guessing at an advertiser.
55% · Editorial score
Daily-driver fit, the jobs the product actually does, and the failure modes we will not sand off a spec sheet.
25% · Press consensus
Where serious desks landed this cycle (Wirecutter, CNET, RTINGS-class, DPReview, Car and Driver, SoundGuys). We synthesise; we do not scrape their sentences.
10% · Value
Street price versus what you get. A 9.5 that only exists at MSRP can lose to an 8.8 on sale if the job is the same.
10% · Freshness
Daily RSS ingest. A launch, a ban, or a safety notice can move notes the same week. It will not buy first place.
Independence. Brands cannot pay to appear on a list or to move a number. They can buy display inventory on /advertise, marked as advertising, kept off the ranking templates.
Experience. Google's E-E-A-T bar wants first-hand photos. We will add them as the desk grows. Until then we label v1 as editorial composite, not lab measurement. That is the honest version of the lightning-bolt line on the homepage.
Bots. A daily cron pulls publisher RSS, counts mentions against ranked names, and writes data/signals.json. Optional Keepa / Amazon PA-API keys plug in for price history when you have them. We do not scrape Amazon HTML.