Every agency quote for an immersive onboarding module assumes your building, your team and your processes will stay still long enough to justify the production cost. They will not. The most useful number in this whole argument comes from PwC, who sell VR training: their own analysis put the break-even against classroom delivery at 375 learners, and against e-learning at 1,950 learners. Below those volumes, a commissioned fixed module is the more expensive option.
Key takeaway: Commission an agency if you onboard thousands of people into a process that will not change for three years. Build it yourself if your sites, shifts and supervisors change, because the second version costs you an afternoon rather than another purchase order.
The problem worth solving
CIPD’s most recent Resourcing and Talent Planning survey found that 41% of employers whose recruits started in the last twelve months said new hires always, mostly or sometimes resigned within the first twelve weeks, and 27% said candidates sometimes failed to turn up on day one at all (CIPD, 2024). Median cost per hire sat at £2,000 for senior roles and £1,500 for everyone else.
For churn, the better figure comes from CIPD’s analysis of ONS Annual Population Survey data: around 34% of UK workers churn in a year, 27.4% moving to a new employer and 6.6% no longer working. It ranges from 25% in public administration to 52% in accommodation and food services (CIPD).
One figure to handle carefully. The widely quoted “£30,614 to replace an employee” is real but it is Oxford Economics research commissioned by the insurer Unum in February 2014, covering five sectors only and employees earning £25,000 or more (Oxford Economics). It is twelve years old and never inflation adjusted. If you use it, date it and name who paid for it.
What the onboarding research actually supports
This matters because it tells you what to build, not just whether to build.
The foundational meta-analysis covered 70 samples and 12,279 people, and found that socialisation works through three things: role clarity, self-efficacy and social acceptance. Institutionalised, structured onboarding tactics predicted role clarity at .27 and job satisfaction at .43 (Bauer et al., Journal of Applied Psychology, 2007).
The 2025 update is far larger, covering 256 studies across 33 countries, and adds perceived fit to the list. Social acceptance correlated with job satisfaction at .42, perceived fit at .46, task mastery with performance at .38. Mentoring and supporting behaviours ran .22 to .38. Undermining by existing staff ran negative .39 to negative .50 (Bauer et al., Journal of Management, 2025).
Allen’s field study tracked 222 new hires and actual turnover at twelve months, of whom 24.8% left. Two tactics significantly reduced the odds of leaving: investiture, meaning affirming the newcomer’s existing identity rather than stripping it, and serial, meaning giving them an experienced insider as a role model (Journal of Management, 2006).
Read that list and the honest conclusion is that the effect lives in structure and human contact, not in the medium. Nothing in this literature says a headset helps. What it says is that people who know their role, feel accepted, and have someone to ask, stay and perform. A walkthrough that tells someone where the break room is and who their buddy is does more of that than an expensively rendered corridor.
So the case for an immersive walkthrough is not that it teaches better. It is that it can reach people before day one, remove the wasted first morning, and be identical for every starter regardless of which supervisor is on shift.
The evidence for VR onboarding specifically
There is essentially none, and it is better to say so.
The general VR training evidence gives a small average effect: a meta-analysis of 48 studies and 3,179 students put it at g = 0.38, with mixed reality coming out negative (Coban et al., Educational Research Review, 2022). A 2025 systematic review of 201 studies on VR in employee training concluded it supports technical and behavioural skill acquisition but flagged “limited evidence of skill transfer to real-world contexts”, short-term measurement, small samples and rapid obsolescence (Applied Sciences, 2025). That review does not address onboarding at all.
The figures everyone quotes, “4x faster to train”, “275% more confident to act”, “3.75x more emotionally connected”, come from a single internal PwC pilot on new managers at twelve US locations, on one topic, not peer reviewed, with per-arm sample sizes undisclosed. PwC’s own caveat says results “will vary with other programs or populations”, and it reported retention findings as inconclusive (PwC).
The genuinely useful part of that study is the cost model, and it cuts against the people citing it. PwC’s break-even was 375 learners against classroom and 1,950 against e-learning. Most organisations are nowhere near 1,950 new starters a year into a single role.
The thing agencies cannot sell you
Here is the practical argument, and it has nothing to do with effect sizes.
ONS data puts hybrid working at 28% of working adults, but the distribution is what matters. Managers and senior officials sit at 45%. Elementary occupations are at 3%, caring and leisure services at 3%, accommodation and food services at 4% (ONS, June 2025).
Put that next to the churn figures. The workforces where a physical site induction is unavoidable are almost entirely on site, and they are also the ones with the highest turnover, up to 52% a year in hospitality. You are inducting the same roles over and over, in buildings that get reorganised, with equipment that gets replaced and supervisors who move on.
A commissioned module is a snapshot. The day the warehouse racking changes, or you open a second site, or the fire assembly point moves, you are back to the agency with a change request. Built in house, that is someone spending an afternoon with a 360° camera.
The day-one duties a walkthrough should carry
If you are building one anyway, make it carry weight rather than just looking impressive.
Health and safety. HSE’s own statement is the strongest line available: “Workers are as likely to have an accident in their first 6 months at work as during the whole of the rest of their working life.” The same page recommends induction “planned with visual aids and a physical walk-through of the workplace to identify hazards” (HSE). A 360° walkthrough is that, delivered before they arrive and repeatable for every starter. Regulation 13 of the Management of Health and Safety at Work Regulations 1999 requires adequate training on recruitment, during working hours (legislation.gov.uk).
The written statement. Since April 2020, the statement of employment particulars under section 1 of the Employment Rights Act 1996 must be given no later than the beginning of employment, and it covers workers as well as employees. Required content includes probationary period details and training entitlements, which is worth checking against what your induction actually says.
What is changing. The Employment Rights Act 2025 received Royal Assent in December 2025 and commences in phases. Already in force from April 2026: day-one paternity leave and unpaid parental leave, and statutory sick pay from day one of sickness with the lower earnings limit removed. From October 2026: the employment tribunal time limit extends from three to six months and the duty to prevent harassment arrives (gov.uk timeline).
One correction, because it is misreported constantly: unfair dismissal is not becoming a day-one right. The Act sets a six month qualifying period from January 2027, down from two years.
Right to work. Civil penalties are £45,000 per illegal worker for a first breach and £60,000 for a repeat, levels set in February 2024 (Home Office code of practice). From 1 October 2026 the scheme extends beyond employees to worker’s contract arrangements, individual sub-contractors and online matching services, with liability extending across contractual chains. This is a compliance step, not something to put in a scenario, but if your onboarding flow does not have it nailed, fix that before buying anything immersive.
What to actually build
Keep it short and specific. The NDTi finding from a different context applies here too: people are more likely to call a short training too long than a long one, because perceived length tracks engagement.
| Include | Skip |
|---|---|
| The route from the car park or bus stop to the right door | A rendered lobby |
| Where to put your things, where to eat, where the toilets are | Company history |
| The three hazards your safety officer is tired of repeating | A full regulations walkthrough |
| Fire route and assembly point from the place they will actually stand | A generic evacuation animation |
| Who their buddy is, with a name and a face | An org chart |
| One decision they will face in week one, with a wrong answer that costs something | A quiz on the values |
That fifth row is the one with research behind it. Serial tactics, giving a newcomer a named experienced person, is one of only two tactics Allen found significantly reduced actual turnover. Naming the buddy in the walkthrough is nearly free and better evidenced than anything else in the module.
Build it once and publish to browser, phone and headset from the same project, so a new starter can do it on their phone the week before they arrive. Ours runs on any device from one build, which matters here more than elsewhere, because the people you most need to reach before day one are the least likely to have a spare headset.
The short answer
Commission an agency when volumes are high, the process is stable and you will not need to change it. Build it yourself when your sites and people move, which for the high-turnover, on-site roles is always. The research says onboarding works through role clarity, social acceptance and human contact, so put those in the walkthrough and do not overpay for photorealism. And use PwC’s own break-even figures when someone tells you a fixed module is the economical option.
See how organisations are using this on the business training pages, or read the build versus buy comparison for the wider decision.