We build custom VR and AR training for industrial, energy and utility operators — tank boil-over response, lockout/tagout, hydrant and pipe inspection, truck operations, kitchen safety. The honest case for VR is not that it teaches better than a good instructor. It is that some jobs cannot be practised any other way.
You will read a lot of claims about VR training in the next month of your search. Most of them are recycled marketing numbers with no study behind them. Here is what the research supports.
A 2021 meta-analysis (Kaplan et al., Human Factors) found XR performed about the same as well-designed conventional training when transfer was measured entirely in the real world. A much larger 2026 meta-analysis (Bateman et al., Educational Research Review — 78 studies, 3,546 participants) found statistically significant positive effects for XR on psychomotor task performance.
They measure different things. The 2021 study counted only performance measured outside the simulation; the 2026 study included outcomes measured inside it. For hands-on physical procedures the evidence now leans favourable — and we'd add what the 2026 authors say themselves, that the studies vary so widely that a single pooled number isn't very meaningful.
The same authors put the real contribution of XR at "providing training in circumstances which exclude traditional methods." Danger, downtime and cost are the three things that exclude traditional methods, and they are where every project we take on lives.
Not the "10% of what we read, 75% of what we do" retention pyramid — it traces to a 1967 oil-company chart and the researcher it is credited to disowned it. Not the Walmart, Tyson or Verizon figures that circulate without a source. Not any percentage we made up about our own clients.
If a laminated job aid, a better SOP or an hour with a competent instructor solves your problem, that is cheaper than anything on this page and we will say so on the first call. We would rather lose the project than sell a $50,000 simulation of a problem a checklist fixes.
MadXR builds custom content of any kind — VR, AR, WebXR, web, mobile, websites and conventional software. This page is the training side of it. We are not a headset-only shop, and the delivery format is a decision we make with you rather than one we sell you.
High-consequence events rehearsed at zero risk: tank boil-over response, journey management, hazard recognition, emergency response. Built for Quest 3 and PC VR, with branching decisions and consequence.
Step-by-step equipment and process training with a guided teach mode and a scored assessment mode against your rubric — lockout/tagout, PPE, machine inspection, instrument operation, truck operations.
Hand-tracked mixed reality overlaid on the machine itself, plus AR work instructions and inspection guidance on tablet and phone. AR and VR training development from one team, so the content model is shared rather than rebuilt.
Photo-based site walks with pass/fail hotspots and defect identification, and full browser-delivered modules that open from a link on any device. Fast, cheap per learner, and no software to distribute.
Before development starts we need two things: a named person on your side who owns the programme after handover, and a signed-off interactive storyboard — every scene, every interaction, every line of voiceover, in a browser you can comment on. We do not start building without both.
A playable version with placeholder art so the flow, the interactions and the failure states can be tested while changes are still cheap. Most rework happens here on purpose. Your SMEs should be in a headset at this stage, not at UAT.
Final 3D assets built from your CAD or from site capture, professional voiceover, and the manipulation detail that makes a procedure feel like the procedure. This is the longest phase and the one that determines whether operators take it seriously.
One structured round of user acceptance testing, then deployment to your headsets, your LMS or the browser, and the source. We also hand over the debrief structure — the questions a facilitator asks after the sim — because that is where most of the learning actually lands. Additional scope is quoted separately rather than absorbed silently.
These are measured from our own completed projects, not estimated from a template. Storyboard review and your internal approvals sit inside these numbers — the schedule moves when sign-off does, which is the single most common cause of a late module.
| Deliverable | What it covers | Time |
|---|---|---|
| Full Unity build | Headset module, final art, voiceover, interactions | ~12 weeks |
| Teach module | Guided step-by-step instruction mode | ~11 weeks |
| Matching test / assessment mode | Scored mode built on the same scene | ~5 weeks |
| 360° module | Photo-based site walk with hotspots and questions | 3–4 weeks |
Building a teach mode and its assessment mode together is far cheaper and faster than commissioning them separately, because the scene, the assets and the interaction logic are already there. If you only have budget for one this year, build the teach mode and add the assessment next cycle.
We publish our prices. Almost no custom VR training developer does, and "request a quote" is the single most common complaint buyers have about this industry. These are the real numbers, and they are what a VR training simulator costs from us.
| Scope | Typical use | Price |
|---|---|---|
| Quick interactive experience | Single topic, micro-learning, product or concept demo | $5,000 – $10,000 |
| Interactive LMS platform | Multi-module course shell with progress tracking | $10,000 – $15,000 |
| Tier 1 — simple interaction module | Basic procedural task, up to 10 minutes, ~5 assets | $30,000 |
| Tier 2 — task training module | Multi-step procedure, 15–20 minutes, ~15 assets | $40,000 – $45,000 |
| Tier 3 — full scenario module | Comprehensive simulation with branching and consequence | $50,000 – $60,000 |
| Add-on | What you get | Price |
|---|---|---|
| Additional mode | Teach, practice or assessment mode on the same scene | +$10,000 |
| Desktop version | Non-headset build for learners without hardware | +$10,000 |
| Multi-user | Up to 8 concurrent participants in one session | +$10,000 |
| AI-powered NPC | Conversational character that responds to speech | +$8,000 |
| Reactive NPC | Character that responds to learner actions | +$4,000 |
| AI coach / chatbot | In-module assistant with feedback at the end | +$4,000 |
| Simple NPC | Scripted background or instruction character | +$2,000 |
| Progress tracking dashboard | Admin view of completion and scores | +$3,000 |
| LMS / SCORM / SSO integration | Reporting into your existing system | +$2,500 |
| Additional module or level | Extra scenario reusing the built environment | +$2,000 |
| Additional language | Full voiceover and on-screen text | +$2,000 |
| Custom 3D art | Bespoke modelling beyond the tier asset count | +$3,000 |
| Gamification | Scoring, badges, leaderboards | +$1,500 |
| Custom branding | Your identity through the interface | +$1,500 |
Terms are 30% at kickoff, 35% at UAT and 35% on final delivery, with one round of UAT included. Additional scope is quoted separately. What drives price is not headset count or minutes of runtime — it is the number of distinct assets we have to model and the number of ways a learner is allowed to get it wrong.
In our experience the software is almost never the reason a programme dies. These four things are, and all of them are decided on your side of the table. We require a named client-side owner before the build starts, because a module without one has nobody to schedule learners into it.
The module ships, the champion moves teams, and six months later nobody can say who is responsible for scheduling learners or fixing a login. This is the number one killer and it costs nothing to prevent.
An immersive experience is memorable on its own. It only becomes training when a facilitator asks what happened, why, and what you would do differently. Without a structured debrief you have bought an experience, not a training programme.
Who charges them, who patches them, who owns the enterprise MDM, what happens when a battery swells. Headsets without an owner end up in a cupboard within a quarter. Plan this before you buy hardware — or skip hardware and go browser-based.
A completion record is not a competency record. Decide up front what the simulation is evidence of, what still requires a signed field observation, and how the two are recorded — before an auditor asks you.
Energy, industrial, utility, construction and food-service operators — work where the subject is equipment, procedure and consequence. Below is what we built, not what it achieved; we do not publish outcome numbers we cannot substantiate.
A tank boil-over emergency scenario on Quest 3, plus journey management and hazard recognition training.
Lockout/tagout, 3S, PPE and machine inspection, plus inverter deep-dives using mixed reality and hand tracking.
Truck operations training with a matching assessment mode built on the same environment.
Water and sewer operations, hydrant work, and CCTV pipe defect identification for inspection crews.
A browser-based site walk with an AI companion, delivered bilingually in English and Spanish.
Emergency response training built in Unreal Engine for a utility environment.
MadXR publishes its prices. A Tier 1 simple interaction module — a basic procedural task, up to 10 minutes, around 5 assets — is $30,000. A Tier 2 task training module covering a multi-step procedure of 15 to 20 minutes with around 15 assets is $40,000 to $45,000. A Tier 3 full scenario module, meaning a comprehensive simulation, is $50,000 to $60,000. A quick single-topic interactive experience is $5,000 to $10,000, and an interactive LMS platform is $10,000 to $15,000. Add-ons are priced separately and listed on this page. Terms are 30% at kickoff, 35% at UAT and 35% on final delivery, with one round of UAT included.
These are measured from our own completed projects rather than estimated. A full Unity headset build takes about 12 weeks. A teach-mode module takes about 11 weeks. A matching test or assessment mode built on the same scene adds about 5 weeks. A 360 photo-based module takes 3 to 4 weeks. Storyboard review and your internal approvals sit inside those numbers, so the schedule moves when sign-off does.
It depends on the skill, and the honest answer moved this year. A 2021 meta-analysis (Kaplan et al., Human Factors) found XR performed about the same as well-designed conventional training when transfer was measured entirely in the real world. A much larger 2026 meta-analysis (Bateman et al., Educational Research Review, 78 studies and 3,546 participants) found statistically significant positive effects for XR on psychomotor task performance. The two are not a clean contradiction — the 2021 study counted only performance measured outside the simulation, the 2026 study included outcomes measured inside it — but for hands-on physical procedures the evidence now leans favourable. We would add what the 2026 authors say themselves: the studies vary so widely that a single pooled number is not very meaningful. The strongest case is still the one neither study disputes — a tank boil-over, an arc flash, a confined-space entry, or equipment you cannot take offline. If a job aid or a good instructor would work, buy that instead.
No. We build the same training for the browser, for desktop, for tablet and phone, and for headset, and we frequently ship a desktop or browser version alongside the VR one so people without a headset can still be trained. Headsets are worth buying when the physical actions matter — reaching, positioning, moving through a space under pressure. If the learning objective is recognition or decision-making, a browser build usually reaches more people for less money.
Four things, and almost never the software. No named internal owner after handover, so nobody schedules learners. No debrief built around the simulation, so the experience is memorable but nothing is drawn out of it. Headsets with no device-management plan, so they end up uncharged and unpatched in a cupboard. And treating completed the simulation as competent on the job, which is a records problem that surfaces during an audit. We require a named client-side owner before the build starts.
Yes. Client-specific content, the 3D assets built for you and the deployed application are yours, with no per-seat licence and no platform subscription. Host it yourself, run it through your own LMS via SCORM or xAPI, or leave it with us. We retain only our general-purpose tooling and frameworks, which you get a licence to use as embedded in the deliverable.
Send us the procedure, the incident you're trying to stop repeating, or the task nobody can practise without shutting a line down. We'll tell you what it would cost, how long it would take, and whether VR is the right answer — including when it isn't.
Start a conversation