It is the most frequently asked question and the least frequently answered. We are going to answer it with what is available, including the part that doesn't work in our favor.
The number everyone cites, with its full context
PwC's 2020 study compared in-person classroom, e-learning, and virtual reality for the same course. Its most useful finding is not the one everyone quotes: the economic break-even point for virtual reality is 375 learners versus the classroom, and 1,950 versus e-learning. At 3,000 learners, virtual reality is 52% more cost-effective. The upfront investment in content is up to 48% higher.
This study has to be cited with its full context, or a technical client will take it apart in the first meeting: the population was new office supervisors, the content was inclusive leadership —soft skills, not plant operations— and PwC sells virtual reality services. It is not an independent trial.
That said, it is the most honest published threshold in this market.
The other number, which points the opposite way
A study published in Applied Sciences monitored six operators assembling a transmission wiring harness using guided augmented reality, compared with a traditional five-day program. The initial cost of the system was €3,472; the cost per trainee of the traditional program, €374.
Break-even is reached with the fourth employee trained.
It is descriptive monitoring with six people and no control group of its own. That limitation has to be stated, but the order of magnitude of the threshold is correct.
Four and 375 don't contradict each other. They are different things.
Here is the only useful answer that can be given.
A high-production virtual reality experience, with generic content and aimed at soft skills, needs massive volume to pay for itself. An augmented reality guide or a simulator built on a specific procedure at your plant pays for itself almost immediately.
Choosing the wrong one is the most expensive mistake in this market.
And there is a data point that explains why. The largest available meta-analysis of virtual reality training pooled 184 studies from 164 sources with 9,007 participants. The average effect is d = 0.541: real, solid, and far from the miracle that the industry's marketing promises.
But the truly useful finding is the moderator. When the fit between the task and the technology is high, the effect rises to d = 0.820. When it is low, it falls to d = 0.207, practically nothing. And the type of hardware —immersive headset or monitor— did not turn out to be a significant moderator.
In other words: the headset is not what produces the result. What produces it is instructional design and the fit with real operations.
The third angle: time, not equipment
A transfer study with a control group at Embry-Riddle Aeronautical University measured how many hours of real flight time are saved for every hour in a simulator before the first solo flight. The transfer ratio was 0.34: every ten hours in the simulator saved 3.4 hours in the aircraft.
But the number that matters to operations is a different one: the training calendar dropped from 80.6 to 59.5 days, 26% less.
The 0.34 is interesting because it is honest. One hour in a simulator does not replace one hour of real operation; it replaces a third. The gain is not in substitution: it is that the scarce, expensive time on real equipment is used when the person already knows the basics, instead of being spent teaching the basics.
How to calculate yours
Three questions, and that is enough for a calculation you can defend to finance:
- How many people go through this procedure each year? Include contractor turnover, which is almost always forgotten.
- How much does each run cost today? Labor hours, instructor, equipment downtime, consumables, permits.
- How closely does the exercise match the real task? If the scenario is generic, you are at d = 0.207 and the numbers don't work. If it replicates your procedure, you are at d = 0.820.
If a vendor doesn't explain which of these paths applies to your case, they haven't done the math.
And that is exactly the conversation worth having before looking at a catalog of experiences.
Sources: PwC US — The effectiveness of virtual reality soft skills training, 2020 · Sabol et al. — Applied Sciences 16(5), 2026 · Embry-Riddle Aeronautical University — JAAER, 2025 · Howard, Gutworth & Jacobs — Computers in Human Behavior, 2021