The XR Spec Gap: What Vendors Do Not Publish
Across 123 live BestInXR posts, 190 cells read Not disclosed or Not published. The gap is not random. Vendors withhold the four things buyers ask for first, publish the specification that no longer separates one product from another and state results with no measurement method attached. This records where each of those happens.
In short:
- 61 of 123 posts carry at least one undisclosed figure. Those 61 posts contain 190 occurrences of Not disclosed or Not published, each one a field BestInXR tried to fill from a primary source and could not.
- Field of view stopped separating display glasses. XREAL publishes 57 degrees and RayNeo publishes 59, which is the number every one of them advertises and the one that no longer discriminates. What actually differs is brightness, from 700 to 1250 nits, and price, from $279 to $599.
- Haptics publishes no comparable measure at all. No supplier publishes a comparable measure of force or fidelity, which leaves price and feedback point counts as the only figures to compare.
- 19 statements across 18 posts record a figure as the company's own rather than audited. That is the honest label for most published percentages in this industry.
How this was counted
Two editorial tokens make the count reproducible. BestInXR writes Not disclosed when a figure cannot be verified from a primary source, and Not published when a supplier publishes nothing for that field at all. Neither is ever an estimate, and neither is ever left blank, because a blank cell reads as incomplete research rather than as a finding.
Counting those two strings across every live post gives 190 occurrences in 61 posts, from a corpus of 123. The figure is a floor rather than a total: prose statements like "none publishes a headquarters address" are a gap too and are not in the count.
Nothing here is a criticism of any individual supplier. A private company is under no obligation to publish its headcount. The point is narrower and more useful: when an entire category withholds the same field, a comparison built on that field is not possible, and a buyer should stop looking for it and ask directly instead.
What vendors do not publish
Four fields go missing more than any others, and they are the four a buyer asks for first.
| Field | How often it is missing | A category where none of the ranked companies publish it |
|---|---|---|
| Founding year | The single most common gap | 3D car and automotive configurators: ZeroLight, Emersya, Hapticmedia and Aircards publish none |
| Team size or headcount | Nearly as common | AR development for ecommerce, where none publishes a headquarters address either |
| Rates or project minimums | Common enough to exclude it as a ranking criterion | AR and VR development for retail: pricing is not a ranking factor because none publishes rates |
| Named clients | Less common, and the strongest evidence when present | Automotive configurators: none publishes named automotive customers |
The pattern is consistent enough to be predictive. In the service categories, where the product is people, the missing fields are the ones that describe the people: when the company started, how many of them there are and what they charge. In the hardware categories, where the product is a device, those fields are usually published and the missing ones are measurements.
This is also why third-party directory profiles are so widely cited. Project minimums, hourly rates, employee bands and award lists appear on aggregator sites and almost nowhere on the vendors' own sites. BestInXR does not carry those figures, because a profile field entered by a vendor into someone else's database is not a primary source, and around 20 such claims across the corpus remain unresolved for exactly that reason.
Specs that have converged
A specification stops being useful when every product in a category publishes the same value. Three categories have reached that point.
Display glasses and field of view. Field of view is the number every manufacturer in the category advertises. XREAL publishes 57 degrees, RayNeo publishes 59 and Viture sits between them. A three-degree spread across the category means the headline spec cannot decide a purchase. Brightness and price still can: brightness runs 700 to 1250 nits and price runs $279 to $599 for hardware that is otherwise very similar.
AI glasses and capability. Meta sells six no-display models across three brands between $224 and $499, and what the extra money buys is frames, fit and lenses rather than capability. The one measurable difference inside that range is generation: the $224 pair records 1080p video against up to 3K on the $379 pair. Every other difference in that tier is a styling difference described in specification language.
Headsets and the specs that are not published at all. Meta does not list weight on its own comparison page. Apple does not publish per-eye resolution for Vision Pro and gives a combined figure of 23 million pixels instead. Varjo publishes a full specification sheet and no price. Each of those is a deliberate choice about which number a buyer gets to compare.
Two more categories publish no comparable measure at all. In haptics, neither bHaptics nor SenseGlove nor anyone else publishes a comparable measure of force or fidelity, so feedback point counts and price carry the whole comparison. In full body tracking, no supplier publishes a latency figure measured the same way as any other, which means the specification that matters most to the experience is the one that cannot be compared.
Figures published without a method
A percentage with no stated measurement is not a measurement. This is the most widespread disclosure problem in the industry and it is concentrated in the commerce categories.
In virtual try-on, conversion and engagement percentages appear on every vendor site in the category with no stated measurement. Almost every published uplift figure lacks a comparison group, a period or an attribution rule. Where a comparison group is absent, a shopper who chose to open a try-on tool is being compared against the average shopper rather than a matched one, so purchase intent is doing part of the work attributed to the software.
The reliable route, where a buyer needs a number, is a holdout test on their own traffic across a full merchandising cycle. That is more work than reading a case study and it is the only version of the claim that transfers.
One published figure in the whole corpus states its own error rate. Fittingbox publishes pupillary distance measurement within 1 millimeter for 7 out of 10 measurements. That is how a measured claim should read: a quantity, a tolerance and a success rate. It is worth naming because it is rare rather than because it is remarkable.
Figures that are the company's own
Nineteen statements across eighteen posts record a figure as company-published rather than audited. The label matters because these are usually the largest numbers on a vendor's site.
Examples, each recorded in the post that found it: a SlimeVR Discord community the company says has 70,000 members, KAT VR claims of support for more than 3,000 games and more than 50,000 users worldwide, a stated 200 percent boost to customer engagement and satisfaction, more than 200 XR-enabled projects and a homepage claim to be the world leader in a category.
None of these is necessarily wrong. All of them are unverifiable from the outside, and the distinction between a company-reported figure and an audited one is the distinction a procurement process is built to make. A named customer stating a figure over a multi-year period is materially stronger evidence than an unattributed vendor statistic, though it remains company-reported.
What to ask for instead
When a field is missing across a whole category, stop treating its absence as a differentiator and ask for it directly.
For a service supplier, ask for the founding year, the size of the team that would work on your project rather than the company, a rate card or a project minimum and two named clients you may contact. All four are routinely withheld in public and routinely supplied on request.
For a hardware supplier, ask which figures are measured and which are claimed, and under what conditions. A claimed field of view and a measured one are different kinds of fact, and the corpus labels them separately for that reason.
For any published percentage, ask three questions: measured against what comparison group, over what period, and under what attribution rule. A figure that cannot answer all three is a marketing claim, and it should be weighed as one.
For anything you plan to write into a document, check what the supplier publishes today rather than what an article said last year. BestInXR's XR Graveyard records what has been withdrawn, the XR hardware price index records what each product costs with the date it was read, and the XR ownership map records who actually owns the company you are about to contract with.
Frequently Asked Questions (FAQ)
1. Why do XR companies not publish their prices?
Most XR development work is scoped per project, so a single price would be misleading, and most suppliers treat rates as commercially sensitive. The practical consequence is that price cannot be used as a comparison criterion in the service categories at all, which is why BestInXR excludes it as a ranking factor where none of the ranked companies publish rates.
2. What does "Not disclosed" mean on BestInXR?
That the figure could not be verified from a primary source. "Not published" means the supplier publishes nothing for that field. Neither is ever an estimate, and neither is left blank, because a blank cell reads as incomplete research rather than as a finding. There are 190 of the two across 61 live posts.
3. Is field of view still a useful spec for AR glasses?
Not on its own. XREAL publishes 57 degrees and RayNeo publishes 59, with Viture between them, so the headline number no longer separates the products. Brightness, which runs from 700 to 1250 nits, and price, which runs from $279 to $599, still do.
4. How should I read a vendor's published uplift percentage?
As a hypothesis to test rather than a forecast. Ask what comparison group it was measured against, over what period and under what attribution rule. In virtual try-on specifically, those percentages appear on every vendor site in the category with no stated measurement method.
5. Why does BestInXR not use figures from directory profiles?
Because a profile field a vendor entered into another company's database is not a primary source. Project minimums, hourly rates, employee bands and award lists appear on aggregator sites and almost nowhere on the vendors' own sites, and around 20 such claims across the corpus are recorded as unresolved rather than repeated.
6. Which XR spec is measured rather than claimed?
Very few, and the distinction is stated wherever it applies. Manufacturer figures for resolution, weight and field of view are the manufacturer's claims unless a measurement method is published alongside them. The clearest published example of a properly measured claim in the corpus is Fittingbox's pupillary distance measurement within 1 millimeter for 7 out of 10 measurements.
