Every product DeskAudit recommends goes through the same pipeline: research, shortlist, purchase at retail, hands-on audit, then publication — with flaws published alongside praise. This page documents the exact protocol: what we measure, how we measure it, and what counts as a pass or a fail. If a claim on this site can't be traced to one of these methods, it shouldn't be here — tell us.

The pipeline

  1. Research. We map the category: every plausible contender, its manufacturer-published specs, its review consensus across retailers, and its recurring complaint patterns. Anything with a disqualifying pattern (systematic breakage reports, fake-review signals, unpublished critical specs) is cut before it costs us a dollar.
  2. Shortlist. From research, we pick the 5–8 products worth buying — covering the best overall, budget, and key use cases. The shortlist and the reasons are published in each roundup's verification note.
  3. Purchase at retail. We buy every product at full retail price, with our own money, from the same stores our readers use. No review units, no discounts, no exceptions — see what we don't do.
  4. Hands-on audit. Each product goes through the per-criterion protocol below, on a real desk, in a real home office. Every measurement is recorded; every failure is photographed.
  5. Publication. Picks, scores, and full results go live — including the products that failed. Failed products stay on the record in a retired-picks section, not deleted.

The per-criterion protocol

Stability: the wobble test

Method: Product loaded to 70% of its manufacturer-rated capacity on a standard 30-inch desk. We apply a controlled lateral push at the platform edge and measure deflection with a ruler against a fixed reference, then type for 60 seconds on a keyboard placed on the same desk surface while observing the platform.

Pass: Deflection under 5 mm with no visible rocking during typing. Fail: Visible rocking, walking feet, or deflection over 10 mm. Anything between is "acceptable with caveats" and noted as such.

Load capacity: the honest weight test

Method: We load the product incrementally to 100% of its manufacturer rating using calibrated barbell plates, hold for 24 hours, and check for permanent deflection, cracking, joint loosening, or leg creep with calipers before and after.

Pass: No measurable permanent deflection and no structural distress at 100% of rating. Fail: Any cracking, joint failure, or permanent deflection over 2 mm. We also report the load at which distress begins, so readers know the real ceiling — not just the marketing one.

Height accuracy and adjustability

Method: Every claimed height setting is measured with a steel tape, three times, and compared against the published spec. Adjustment mechanisms are cycled 25 times to check for loosening, sticking, or wear.

Pass: All settings within 3 mm of published specs; mechanism still firm after 25 cycles. Fail: Any setting off by more than 6 mm, or a mechanism that loosens, sticks, or requires tools the manufacturer didn't mention.

Assembly: the Allen-wrench ritual, timed

Method: One tester, the included instructions, a phone stopwatch, no prior assembly of that product. We record total time, tools required vs. tools included, and every point where the instructions were unclear or wrong.

Pass: Assembled in under 20 minutes with included tools and comprehensible instructions. Fail: Missing hardware, instructions that don't match the parts, or a process requiring unlisted tools. We publish the timed result either way.

Fit and finish

Method: Visual and tactile inspection under desk lighting: sharp edges (checked with a fingertip sweep and, where suspect, a cloth snag test), coating consistency, weld and joint quality, foot/pad completeness against the parts list.

Pass: No sharp edges, complete hardware, consistent finish. Fail: Sharp edges, missing parts out of the box, or finish defects affecting function. Cosmetic-only defects are noted but don't fail a product alone.

Long-term sag check

Method: After the load test, products are kept loaded at 70% of rating for 30 days in normal office conditions. Height is re-measured at day 30 with calipers.

Pass: Height change under 2 mm. Fail: Sag over 3 mm — the classic plastic-riser failure mode, and exactly why this test exists.

Documentation honesty

Method: We compare every manufacturer claim — dimensions, capacity, materials, warranty terms — against what arrives in the box and what our tests show.

Pass: All material claims accurate; warranty terms honored in a test support contact. Fail: Any false material claim (e.g., "metal" parts that are plastic) or warranty/support that doesn't respond within 5 business days. Findings are published verbatim in each review's "honest part."

Scoring

Each product gets a DeskAudit score out of 10, weighted: stability 25%, load honesty 20%, adjustability/fit 15%, assembly 10%, fit and finish 10%, long-term sag 10%, documentation honesty 10%. Scores are research-based until the hands-on audit completes, and every score is labeled as such. A product that fails any safety-relevant criterion (load, stability) cannot score above 5.0, no matter how nice it looks.

Launch status

DeskAudit launched on October 17, 2026. The hands-on protocol above activates with our first purchased batch — the products are shortlisted, the test bench is specified, and the first audit results will be published as soon as testing completes (target: January 2027).

How launch picks were selected meanwhile: every pick in our launch articles was chosen from verifiable public evidence only — manufacturer-published specifications, aggregated retailer review consensus (tens of thousands of ratings), and recurring complaint patterns in owner reviews. Each launch article carries a clearly-labeled verification note stating exactly this. No launch pick is described as hands-on tested, no measurement is presented as ours, and any spec we couldn't verify is marked "not published by manufacturer" rather than guessed. When the first audit batch completes, every launch article will be updated with measured results — and any pick that fails will be demoted or retired on the record.

What we don't do

  • No sponsored placements. Nobody can pay for a pick, a rank, or a score. Ever.
  • No free review units. We buy at retail, with our own money. A manufacturer sending us a free product would disqualify that product from review, not fast-track it.
  • No pre-publication manufacturer review. Manufacturers don't see our reviews before you do. We contact them for comment on failures when the claim is serious — and we publish their response or their silence.
  • No affiliate-driven rankings. Our affiliate relationships (see affiliate disclosure) pay the same regardless of which product you pick, and picks are finalized before any affiliate link is attached.
  • No deleted failures. Products that fail stay published with their failures documented. A review site that only shows winners is a catalog, not an audit.
  • No fake precision. If we didn't measure it, we don't print a number. Estimates are labeled as estimates; gaps are labeled as gaps.

Corrections

We get things wrong sometimes. When we do, we correct the article, note the correction with a date at the bottom of the page, and — if the error affected a pick or a score — we say so explicitly. See our editorial policy for the full corrections process.

Questions about our methods? Contact us. Read why DeskAudit exists and how we're funded.