What the use case is
XR training puts a learner through a procedure in a simulated environment, so the rehearsal happens somewhere that practising for real would be dangerous, expensive or disruptive.
That framing sets the boundary usefully. The technology earns its cost where the real thing carries consequence: a live rail corridor, a working plant, an operating theatre, a warehouse mid-shift. Where the task can simply be practised at a desk, a headset adds cost and removes nothing.
XR training is not a replacement for instruction. In the published record it consistently sits alongside classroom teaching and supervised practice rather than displacing either, and it is used for the middle step between being told how to do something and doing it under supervision. Treating it as a substitute for the whole programme is the most common way a pilot disappoints.
It is also not one thing. A guided walkthrough with no assessment, a branching procedure with a score, and a physics-accurate simulation of a machine are three different products with an order of magnitude between them in cost. Buyers frequently use one word for all three.
What the published record shows
Training is the most common use case in this register, and safety training is the most common variety within it.
That is a statement the register can support, because every case study on a record traces to a public source. Across the studios currently on record, training and simulation attaches to more companies than any other use case, and the pattern inside it is consistent enough to be worth reading directly.
Hazard and safety rehearsal dominates. 4Experience publishes safety training for STRABAG and driver training for DPD. Bit Space Development documents a near-miss simulation for Canadian National Railway and power tool training for the Manitoba Construction Sector Council. Each rehearses a situation that is genuinely hazardous to stage.
Equipment and procedural operation is the second cluster. Nsflow publishes forklift operator training for an automotive manufacturer and machine operator training under restrictive conditions. These are repeatable tasks with a correct sequence, which is the shape the technology handles best.
A third and smaller group covers judgment rather than procedure, such as soft skills and threat response. It appears in the record but far less often, and a buyer scoping it should expect a thinner evidence base than for procedural work.
What the record does not settle is effectiveness. Case studies describe what was delivered, not what changed, and very few publish an outcome measured against a control. That absence is worth stating plainly. Anyone citing a retention or speed figure for XR training should be asked which study it comes from, because the number circulating in vendor material is usually traceable to a single vendor-funded piece of research rather than to a body of evidence.
How to scope a programme
Scope on the number of distinct procedures, not on hours of content or headcount.
A procedure is the unit that drives cost. Each one is a separate flow to design, build, test and validate with someone who performs it, and adding a second procedure to an existing environment is far cheaper than adding a second environment. A programme covering six tasks in one plant room is a fundamentally different job from six tasks across six sites.
Decide early whether the training must assess. Assessment changes the product: it requires a defined correct path, tolerance for acceptable variation, a scoring model and somewhere for results to go. Retrofitting assessment into a walkthrough is close to a rebuild, and it is the single most common mid-project scope change.
Settle the reporting path before the build. Training that produces no record is hard to justify a second year, and the integration into an existing learning system usually sits with the client's IT function rather than with the studio, which makes it the longest lead item on many projects.
Plan for content maintenance. Procedures change, and a simulation that teaches last year's process is worse than no simulation. Ask at proposal stage what a change costs and who can make it.
What it costs
No credible public source publishes what an XR training programme costs, and this register does not publish one either.
Price data on studio directories is self-reported through a submission form, which makes it a marketing figure rather than an observation. What can be stated is what moves the number, and for training the drivers differ from other use cases.
- Procedure count and branching. Every decision point multiplies the states that must be built and tested. A linear walkthrough and a branching procedure with failure paths are not comparable products.
- Fidelity of the simulated equipment. Representing a machine visually is inexpensive. Making it behave correctly when operated incorrectly is a different discipline and the largest single cost jump in this category.
- Assessment and reporting. The scoring model, the data it produces and the integration that carries it into a learning system.
- Fleet operation. Devices, charging, hygiene, provisioning and the person who runs the sessions. These are operational lines, they recur annually, and they are routinely missing from a first budget.
The recurring cost is the one buyers underestimate most. A programme is not a project that ends, and the second year carries device replacement, content updates and support.
Which technology fits, and which does not
Virtual Reality fits when the environment is the point, and Augmented Reality fits when the real equipment is in front of the learner.
That division is more reliable than any hardware comparison. If the trainee must be somewhere they cannot safely go, the environment has to be simulated, which means Virtual Reality. If the trainee is standing at the actual machine and needs guidance registered to it, replacing their view is exactly wrong, and Augmented Reality is the correct and usually cheaper answer.
Mixed Reality sits between them and earns its premium when the learner must use real tools while digital guidance responds to the real room. That is a narrower case than it is sold as, and the evidence standard this register applies is worth pointing at whenever a proposal claims it without a delivered example.
Two cases do not fit XR at all. Knowledge that can be read and tested does not need a headset, and putting a slide deck in one is a well-documented way to spend a budget on novelty. Tasks requiring fine force feedback, where the skill lives in the hands rather than the sequence, remain poorly served without specialist hardware that most programmes cannot justify.
What the published record does not tell you
Training case studies describe what was built and almost never describe what it achieved.
Across the training work this register documents, roughly one in nine names a target device and about one in twenty-five carries a date. Client names appear far more often. The result is a corpus that establishes a studio has delivered training software and leaves the two questions a training buyer actually has, which device and how recently, to be asked directly.
A third omission matters more than either. Very few published training case studies state an assessment method, a completion rate or a comparison against the instruction the programme replaced. A studio is not obliged to publish a client's internal results and often cannot, so the absence is not evidence of a weak outcome. It does mean a portfolio cannot settle whether a studio has built training that was measured, which is a different capability from building training that was delivered.
Ask for one programme where results were tracked, who defined the measure and whether the studio saw the data. The answer separates a studio that shipped a course from one that helped a client prove it worked.
Further published work in this category
Published training work clusters in industrial and institutional settings, and the commissioning body shapes what gets documented.
Draw and Code publishes a training application built for Saint-Gobain in a manufacturing setting, and HQSoftware publishes occupational training simulations without naming a client. Those two documents sit at opposite ends of the disclosure range, and both are common. A named manufacturer usually signals a procurement process with a stated objective, while an unnamed programme is often work delivered under an agreement the studio cannot describe.
Platform-level training work is a separate commissioning pattern worth recognizing. Nsflow publishes an industrial training management platform delivered for Grupa Azoty, which is a system for administering training across a workforce rather than a single simulated procedure. The distinction matters at scoping time, because a platform carries user management, reporting and integration obligations that a single module does not, and those are usually what makes the second year cost more than the first.
Where these programmes fail
Training programmes fail most often because nobody agreed in advance what would count as success.
A simulation that is enjoyable, well built and unmeasured cannot be defended at renewal. The failure is not the software and it is rarely visible during delivery, which is what makes it common. Deciding before the build whether the programme must demonstrate competence, reduce incidents, shorten instruction time or simply replace an existing course determines the assessment design, and assessment designed afterwards rarely fits.
The second failure is treating the content as finished. Procedures change, equipment is replaced and regulations are updated, so a simulation encodes a process at a point in time. A programme with no update path degrades into training that contradicts current practice, which is worse than no simulation at all.
Third, the cohort is misjudged. Training designed for confident users fails for a workforce with limited headset experience, and the first session is then spent on the controls rather than on the procedure. Onboarding, supervision and the realistic tolerance of new users belong in the scope, particularly where attendance is mandatory and the audience did not choose to be there.
A fourth failure belongs to whoever owns the content afterwards. A studio that delivers a finished simulation and no means of editing it has handed over an asset that only it can change, which is a reasonable arrangement when both sides choose it and an unwelcome discovery when they did not. Ask whether procedures can be updated by the client, what that costs and what skills it requires, because a training programme that cannot be corrected within a week of a procedure changing will be contradicted by the work floor before it is retired.
One final point applies whatever the subject. Simulation earns its cost where the real thing is dangerous, expensive, rare or impossible to rehearse, and it competes poorly with a well-run classroom where none of those apply. A studio willing to say that a proposed programme does not need Virtual Reality is giving better advice than one that accepts every brief, and that answer is worth more at the scoping stage than a portfolio.
Where to go next
Every studio named here is on the public record with its sources and its verification date, and the full list sits on the register.
Rankings are scoped rather than global, because a studio strong in warehouse safety rehearsal is not thereby strong in clinical simulation. The most useful preparation before requesting proposals is to write down the procedures, decide whether each one must be assessed, and confirm where results need to land. Training is delivered on all three technologies the register tracks, and the choice between them is a scoping decision rather than a supplier one. Virtual Reality development carries the deepest published training record, Augmented Reality development covers the widest supplier pool, and Mixed Reality development is the smallest of the three but the most credible where a trainee must see the real workspace.
Two other pillars share most of these suppliers. Healthcare is the sector with the most documented training work and the heaviest governance around it, and product visualization draws on the same studios for a different job, which is worth checking before assuming a portfolio is training-specific.
