On 10 August 2026 we opened two web pages about the same machine. Husqvarna's US site says its DXR 145 demolition robot has an operating weight of 2,315 lb. The product page at an authorized US retailer for the same SKU says 2,171.52 lb. The gap is 143 lb, or 6.6%, and neither page explains it.
That is a small discrepancy on a boring specification, and it is exactly why it is worth starting here. Operating weight is the most testable number a machine has. You can put a robot on a scale. If two parties in the same distribution chain cannot agree on the weight, it is worth asking who, if anyone, checks the numbers that cannot be put on a scale — depth rating, ingress protection, payload, coverage rate, uptime.
The answer, across every category we have priced this year, is almost always: nobody. The manufacturer declares, the dealer copies, the aggregator copies the dealer, and the buyer treats the fourth-hand copy as corroboration.
The self-declaration stack
Here is the pattern, from our own reporting, with dates.
Depth ratings. In our three-brand underwater ROV comparison on 6 August we noted that every depth rating in the piece — QYSEA's 100 m, Chasing's 100 m and 200 m, Blue Robotics' 300 m aluminium enclosure — came from the manufacturer. There is no third-party pressure-test certificate published for any of them. A depth rating is a structural safety claim on a vehicle that will be sent somewhere a human cannot follow, and it rests entirely on the vendor's own testing.
IP ratings. Our pipe-inspection crawler survey on 3 August found the cheapest machine in the catalog, a $9,695 crawler, rated IP67 to 1 m, while $38,995 and $48,995 machines were rated IP68 to 10 m. IP is a real IEC standard with a real test method — and it is also, in practice, declared rather than certified for this class of equipment. The number on the datasheet is what the vendor says the enclosure did.
Coverage rate. On 30 July, pricing LionsBot commercial cleaning robots, we found the US distributor advertising ~13,563 ft²/h for the R3 Vac while LionsBot's own specification is 500–800 m²/h. Worse, the manufacturer's own English-unit conversion on that line is internally wrong: it prints 5,382–10,764 ft²/h, but 10,764 ft² corresponds to 1,000 m², not 800. A fleet sized on the distributor's number would be short by a factor of about 2.5.
Payload. On 31 July we opened KEENON's page for the S100 and found the headline reading "100kg+ Load Capacity" while the parameter table on the same page printed a maximum payload of 300 kg — with dimensions, weight and battery identical to a different model's table. One page, one machine, two payload figures, three times apart.
Price, of all things. Even numbers published by a government are not immune to laundering. The microprocessor ankle orthosis code L2221 took effect on 1 April 2026, and the figure the orthotics trade has circulated for it is $1,889.92. That number is real — but it is the arithmetic mean of 53 jurisdiction allowances, and no jurisdiction is paid it. The largest single group, twelve states plus the District of Columbia, is allowed $1,663.48, which is 12.0% lower. A statistic became a price in one hop, and it happened inside the most transparent pricing regime in this industry.
The one place a third party actually stands in the way
There is an exception, and it is instructive precisely because of what it does and does not check.
In the US durable medical equipment channel, CMS contracts a body called the PDAC — the Pricing, Data Analysis and Coding contractor, operated by Palmetto GBA — to run Coding Verification Reviews. A manufacturer submits a product; if the PDAC agrees it meets a code's descriptor, the product enters a public classification list; if it is not on the list, claims are denied for incorrect coding. We wrote that up on 7 August as a market-access barrier hiding inside a billing rule. It has teeth: Fior & Gentz's Neuro HiTronic microprocessor KAFO knee joint was PDAC-verified for L2006, reported by The O&P EDGE on 21 May 2025, and that verification is what makes a roughly $39,500 allowance billable.
Now notice what the PDAC is verifying. It is verifying that a device matches a *descriptive* code — is it custom fabricated, does it have swing and stance phase microprocessor control, does it include the charger. It is not verifying that the knee flexes to a stated angle, that the battery lasts a stated number of hours, or that the enclosure survives a stated ingress test. The one third-party gate in the industry that actually stops money checks paperwork, not performance.
Elsewhere there is no gate at all. CE marking and FCC authorisation are conformity declarations, not performance audits. UL and TÜV certifications, where they exist, cover electrical safety. ISO 10218 and ISO/TS 15066 — which we have covered in our robot safety standards guide — govern how a robot must behave around people, not whether its brochure is accurate. None of them certify that a marketing number is true.
Our view: the industry has built a fairly serious apparatus for safety and a fairly serious apparatus for billing, and essentially nothing for performance claims. That is a defensible allocation of regulatory attention right up to the point where the performance claim *is* the safety claim — which is what a depth rating, an IP rating and a load rating all are.
Where the liability actually lands
Ask the awkward question. A 100 m-rated ROV floods at 80 m. Who is answerable?
In practice the buyer is, and the buyer usually discovers this in the warranty terms rather than in the specification. Our cobot after-sales comparison on 29 July found that warranty clocks start variously at shipment, delivery or installation depending on the brand, and that FANUC does not publish a CRX warranty period at all while marketing "8 years of ZERO maintenance" — a figure derived from accelerated life testing, which is a research result presented in the position where a commitment would go. A rating that appears in marketing and not in the warranty is a claim, not a promise, and the difference only becomes visible after something breaks.
There is a smaller version of the same confusion in accessory catalogs. QYSEA sells a 300-metre tether spool at $1,850 for a product line whose machines are, by the manufacturer's own specification, rated to 100 metres. The tether number is horizontal range, not depth, and QYSEA is not claiming otherwise — but a 300 m figure sitting on the accessory page of a 100 m machine is precisely the sort of number that gets copied into a spec table by someone in a hurry.
Four checks that cost nothing
For buyers, the practical response is not to distrust every datasheet. It is to run the cheap tests that catch the failures we actually keep finding:
- Open the manufacturer page and the dealer page side by side. Disagreement on a testable number — weight, dimensions, battery capacity — is a reliable signal that the harder numbers were also copied rather than checked.
- Check a spec sheet against itself. Two of the five failures above were internal contradictions on a single page: a headline that disagreed with its own table, and a metric-to-imperial conversion that was wrong by 25%.
- Ask which number is in the warranty. If a rating appears only in marketing, price the risk of it being wrong, because you are carrying it.
- Ask who tested it, and ask for the report. Not to receive it — you often will not — but because the answer tells you whether the number came from a test rig or from a product manager. In the categories where we have asked, that distinction is nearly the whole story.
Sources
- Husqvarna Construction US, demolition robot model specifications, read 10 August 2026; Contractors Direct product page for the same SKU (970509705), read 10 August 2026.
- CMS DMEPOS fee schedule public use file DME26-B (April 2026), parsed 10 August 2026; Össur R&R, CMS Announces New Orthotic and Prosthetic Codes and Payment Amounts, 19 March 2026.
- The O&P EDGE, HCPCS Code Approved for Microprocessor-Controlled KAFO, 21 May 2025.
- Össur R&R on Coding Verification Review requirements, 19 June 2024, as cited in our 7 August analysis.
- GrabaRobot original reporting: ROV brands compared (6 Aug 2026), pipe inspection crawlers (3 Aug 2026), LionsBot price guide (30 Jul 2026), service robots by venue (31 Jul 2026), cobot warranty and service compared (29 Jul 2026).
What is fact and what is ours. The specification discrepancies, the fee-schedule arithmetic and the PDAC mechanism are documented facts with the sources above. The judgement that the industry regulates safety and billing but not performance claims, and that buyers carry the resulting risk, is ours.
