Does wellbeing spending actually work? The case against wellbeing-washing

A new starter joins your data team. In the first week, before they have write access to a single repository, the onboarding deck introduces the meditation app the company pays for, the monthly resilience webinar, the Employee Assistance hotline, and a Slack channel full of step-challenge leaderboards. The message underneath all of it is warm and unmistakable: we care about your well-being. Six months later the same person is answering pull-request comments at 11pm, has not taken a real lunch break in weeks, and has quietly stopped opening the meditation app. Nothing in the wellness package touches the thing that is actually grinding them down.

That gap — between what companies buy in the name of well-being and what well-being actually responds to — is the single most under-reported fact in the modern workplace. The corporate wellness business has become enormous, and it rests on an intuition that feels obviously true: give stressed people tools to relax and cope, and they will be healthier and happier. The awkward problem is that when researchers test the intuition properly, with control groups and randomisation, it mostly falls apart.

A very large industry built on a hunch

Workplace wellness is now a serious line item, not a fringe perk. Global spending on wellness interventions has climbed from around $61 billion in 2021 toward a projected $94.6 billion by 2026, and in the United States something like 80 percent of large employers offer some form of program. The catalogue is familiar to anyone onboarded in the last few years: mindfulness seminars, meditation and sleep apps, resilience and stress-management classes, on-site massage, life coaching, and an ever-expanding menu of digital “well-being solutions.”

The business case sounds airtight. Healthier, calmer employees should mean lower medical costs, fewer sick days, and sharper work. So money flows in, the perks get announced, and everyone moves on satisfied that the box is ticked. The question almost nobody in the buying chain stops to ask is the only one that matters: when you measure the outcomes carefully, does any of this actually change them?

What the independent research found

The most comprehensive answer comes from William Fleming at Oxford University’s Wellbeing Research Centre, in work published in the Industrial Relations Journal in 2024. Using Britain’s Healthiest Workplace survey data, Fleming compared employees who took part in wellness offerings against those who did not — drawing on roughly 46,000 workers across 233 organisations and around 90 distinct interventions. It is, by some distance, the largest independent look at this question we have.

The result is hard to soften. Across mindfulness, resilience training, stress management, well-being apps, sleep apps, time-management coaching, and relaxation classes, participation was associated with no measurable improvement in well-being, job satisfaction, sense of belonging, or perceived time pressure. The needle simply did not move. One category stood out as an exception in the right direction — volunteering and charity work — and one stood out in the wrong direction: resilience and stress-management training trended slightly negative on self-rated mental health. Fleming’s own summary of the implication has been widely quoted since, and it is worth taking seriously: there is nothing wrong with offering these things if people want them, but if you genuinely want to improve well-being, the lever is working practices, not individual coping tools.

A second study, by researchers at the University of Chicago and Harvard published in JAMA, attacked the same question with the gold-standard method: a randomised controlled trial. Across a large warehouse retailer — close to 33,000 employees over 160 worksites — some locations were randomly assigned a wellness program and others were not. The program did nudge a couple of self-reported behaviours: workers at participating sites reported exercising more, by roughly 8 percentage points. But on the outcomes that actually justify the spend — clinical health markers, healthcare costs and usage, absenteeism, how long people stayed, and job performance — it produced no significant effect at all.

Two very different methods, one observational at enormous scale and one randomised, landing in the same place. That convergence is the part to hold on to.

The counter-story, and why to read it warily

Type “wellness program ROI” into a search bar and a completely different reality appears. Vendor and benefits-industry write-ups routinely report glowing returns: claims that the large majority of companies measuring ROI see positive numbers, that many get two dollars back for every dollar spent, that participants spend roughly a third less in medical claims, alongside the old set-piece example of a major consumer-goods firm crediting its program with hundreds of millions in savings over a decade.

Those figures deserve real scepticism, for two reasons. The first is who is doing the counting: the upbeat numbers come overwhelmingly from the vendors and insurers selling the programs, while the null results come from independent academics with no product to move. The second reason is deeper and more interesting, because it explains how both sides can be reporting honestly. Almost every ROI claim compares people who joined a program with people who did not. But the employees who sign up for the step challenge and the meditation course are not a random slice of the workforce — they are, on average, already healthier, more motivated, and more engaged. That is textbook selection bias, and it can manufacture an apparent benefit out of thin air. The reason the JAMA trial matters so much is that randomisation breaks exactly this illusion: assign the program by chance rather than by who volunteers, and the impressive ROI evaporates.

Why coping tools cannot fix a workload problem

Step back from the numbers and the failure makes conceptual sense. Almost every popular program treats well-being as a personal skill to be trained — breathe better, sleep better, think more positively, build resilience — while leaving the actual sources of stress completely untouched. A meditation app asks an overloaded engineer to manage the symptoms of an unsustainable job. It does nothing about the workload, the unrealistic deadline, the on-call rotation, or the manager who pings at midnight.

There is a sharper version of this point that is easy to miss: the most effective way to improve mental health at work is to reduce the stress, not to add more ways of coping with it. Hand someone a resilience course in place of a fix, and you can accidentally send a second, uglier message underneath the warm one — that the strain is their personal failing to manage, not a condition the organisation created. That guilt is a plausible reason resilience training came out slightly negative in the Oxford data. You are asking people to self-optimise their way out of a structural problem, and on some level they know it.

The drivers that independent research keeps pointing to are not personal at all. They are structural: workload, pay, career progression, scheduling and flexibility, and the quality of the person you report to. For a data or software team specifically, that translates into concrete things a meditation subscription will never reach — sprint scoping that is actually achievable, protection of focused time from meeting fragmentation, realistic on-call expectations, and a manager who absorbs pressure rather than passing it straight down.

What the evidence says does work

The research is not nihilistic. It points to a rough hierarchy, and the ordering is instructive.

Strongest is changing the work itself. When organisations have genuinely reorganised time — the four-day-week trials are the clearest case — burnout fell sharply and mental and physical health improved, precisely because the intervention altered a structural driver instead of a coping skill. That is the opposite end of the spectrum from an app.

Next is real clinical care, as distinct from self-guided content. There is a meaningful difference between an app that plays you a breathing exercise and actual access to a therapist or psychiatric support. Where employees can reach genuine care, outcomes improve. The distinction worth carrying around is between care and content: one is a clinician, the other is a library you are left to navigate alone.

Then there is the quieter, more cultural variable that turns out to matter a great deal: whether people feel the organisation genuinely cares about them. Survey work in this area finds that employees who believe their employer truly cares report substantially higher well-being and are markedly more likely to stay — yet only about one in four strongly agree that their organisation does. Care, in this sense, is not a perk you can purchase and announce. It is something demonstrated through workload decisions, staffing, and how the company behaves when an employee is struggling.

And finally there is the one perk that did help in the Oxford data: volunteering. Among roughly 90 interventions, structured chances to do charitable or community work were the lone positive — most likely because they build meaning and connection outward, rather than asking an exhausted person to turn the optimisation lens back on themselves.

Reading your own employer’s wellness menu

Put all of this together and a useful word emerges for the gap between the branding and the lived reality: wellbeing-washing. It is the well-being equivalent of greenwashing — visible perks promoted while the structural sources of stress sit untouched, and sometimes a program that proclaims “we care” while the workload stays punishing, which can deepen cynicism rather than relieve it.

None of this means you should refuse the free therapy sessions or uninstall the sleep app. If a benefit helps you, use it. The point is calibration: a wellness package is a signal to read, not in itself a benefit to be grateful for. The honest questions to ask of your own employer are not “what perks do they offer?” but “do my deadlines and my workload assume I am a sustainable human being? Can I actually reach a real clinician when I need one, not just an app? Does my manager protect my time, or quietly erode it? And when someone here is genuinely struggling, what does the company do?”

If the answers are good, the meditation app is a pleasant extra on top of a healthy foundation. If the answers are bad, no quantity of mindfulness webinars will save you, and the well-being branding is doing a job — it is just not the job of improving your well-being. The data is unusually clear on this for a question that feels so soft. Well-being is built into how the work is designed, not bolted onto it afterward. The most useful thing you can do with that knowledge is stop grading your employer on the perks, and start grading it on the practices.