Workplace wellbeing: Why engagement does not equal better health

Posted in partnership with Spital Clinic

Research covering 46,336 UK workers found no benefit from eleven common wellbeing interventions. The procurement habits that built that market have not changed, and engagement dashboards are the reason why.

The scale of the problem is not in dispute.

Health and Safety Executive statistics show that 1.9 million workers in Great Britain suffered from work-related ill health in 2024/25, that 964,000 of them had stress, depression or anxiety, and that 40.1 million working days were lost.

The estimated annual cost of injuries and ill health arising from current working conditions was £22.9 billion in 2023/24.

Employers have responded to that with procurement.

Over the past decade the corporate wellbeing market has filled up with digital point solutions: meditation apps, resilience platforms, mood trackers, step challenges, webinar series. Most are bought on a per-employee licence, reported on through engagement dashboards, and renewed annually.

A declaration of interest. Spital Clinic sells corporate healthcare to London employers, and is therefore part of this market rather than a neutral observer of it. That is also why the evidence below is worth taking seriously, because it does not flatter anyone selling into it.

The dominant model has been tested, and it did not hold up

In 2024, Dr William Fleming at the University of Oxford’s Wellbeing Research Centre published an analysis in the Industrial Relations Journal covering 46,336 UK workers across 233 organisations.

It compared people who took part in workplace wellbeing interventions with people who did not, across eleven common individual-level practices: mindfulness, resilience and stress management training, relaxation classes, wellbeing apps and similar.

The finding was that none of them showed a benefit. Not a small benefit, not a benefit in some subgroups.

Some showed a small negative association, which the author attributes to selection effects, with people already struggling being more likely to sign up and not receiving what the product promised. The pattern held across job levels, organisations and industries.

That is an unusually clean result for social science, and it indicts a product category worth a great deal of money. The sector’s response has largely been to carry on selling.

Why engagement metrics survive bad evidence

The reason this persisted so long is that the industry measures the wrong thing, and does so sincerely.

Engagement metrics are easy to collect, improve reliably with better design, and look like progress on a slide. Downloads, logins, session length, challenge participation, satisfaction scores.

None of them is an outcome. They measure whether someone opened a thing, not whether their health changed, and the gap between those two is where the entire problem sits.

Worse, engagement metrics are gameable in ways outcomes are not. An app with a daily streak mechanic will show rising engagement whether or not anyone is better.

A provider reporting 4.6 out of 5 satisfaction has told you that people found the experience pleasant, which is not the same claim as having helped them.

HR teams are not being foolish here. They are buying under real pressure, with limited clinical expertise in the room, from vendors presenting the only metrics anyone has asked for.

The failure is a market design failure, not a competence one.

The honest version of the alternative

The obvious move for a clinical provider at this point is to say the answer is clinical services, meaning its own. The evidence does not support stating that confidently.

Two things can be said with reasonable confidence. The first is that organisational factors have a better evidence base than individual ones.

Workload, autonomy, job security and the quality of line management do more to shape health at work than anything you can install on a phone, and no amount of procurement substitutes for fixing them.

The second is narrower and more relevant to anyone selling health services into employers. Removing access barriers is a different intervention from teaching coping skills.

Someone who waits weeks for a diagnostic appointment while symptoms continue is in a materially different position from someone seen quickly, and that is a logistics problem technology is genuinely good at solving.

That claim needs careful handling. The access model is less well-evaluated than the app layer that has just failed its test, not better evaluated.

Employer-purchased health screening in particular has a mixed evidence base, and broad screening of asymptomatic people has never produced the outcome improvements its advocates expected.

Providers in this space, Spital Clinic included, should be generating outcome data rather than assuming they sit outside the critique that has just landed on their neighbours.

What employers should actually demand

Three questions would improve procurement in this market considerably, and none requires clinical training to ask.

What outcome do you measure, and how? If the answer is engagement, satisfaction or utilisation, the vendor is reporting activity. Ask what changed for the people who used it, measured how, against what comparison.

Where is your denominator? Case studies describe people the service worked for. The interesting number is what happened across everyone who was offered it, including those who never engaged, because that is the population the employer paid for.

What would failure look like? A vendor that cannot describe a result that would count as its product not working has not designed an evaluation. It has designed a marketing exercise.

Health technology sold into workplaces has had an unusually easy ride on evidence, largely because the buyer is not the patient and the person paying rarely sees the clinical result.

That gap is closing. The Oxford work is the first large UK study to put a well-funded category of workplace health products to a serious test, and the category did not survive it.

The sensible conclusion is not that employers should stop spending on health. Given 40.1 million working days lost a year, that would be an odd response.

It is that the next decade of spending should be conditional on evidence in a way the last decade was not, and vendors who object to that are telling you something useful.