burnout is a job design problem, not a resilience problem

An organisation with a burnout problem usually responds by buying something for individuals. A meditation app, a resilience workshop, a wellbeing platform, an extra day of leave. The response is sincere, and the evidence says it does very little, because the causes it addresses are not the causes the research identifies.

The World Health Organization settled the definitional question in 2019. In ICD-11, burnout is classified as an occupational phenomenon rather than a medical condition. It is defined as a syndrome resulting from chronic workplace stress that has not been successfully managed, with three dimensions: energy depletion or exhaustion, increased mental distance from the job or cynicism about it, and reduced professional efficacy (World Health Organization, 2019). The definition places the cause in the work, and it explicitly excludes experiences outside the occupational context.

the model that explains it

The most useful framework is the job demands-resources model, set out by Demerouti, Bakker, Nachreiner and Schaufeli in the Journal of Applied Psychology in 2001. It sorts working conditions into two categories with different effects. Job demands, the aspects of work that cost sustained physical or mental effort, drive exhaustion. Job resources, the things that help people get work done and develop, protect against disengagement when present and produce it when absent (Demerouti, Bakker, Nachreiner and Schaufeli, 2001).

Two paths, two different fixes. Exhaustion is a demands problem. Cynicism is a resources problem. An organisation that answers both with the same intervention will fix neither.

The model got sharper in 2010. Crawford, LePine and Rich ran a meta-analysis across 64 samples and split demands into two kinds. Challenge demands, such as workload, time pressure and responsibility on work that matters, raise burnout but also raise engagement. Hindrance demands, such as role ambiguity, role conflict, politics and administrative obstruction, raise burnout and lower engagement (Crawford, LePine and Rich, 2010).

That distinction changes the instruction you give managers. "Reduce workload" is the wrong brief, because part of the load is the part people find meaningful, and cutting it removes the engagement along with the strain. "Remove hindrance demands" is the right brief. Hindrance load has no upside at all. Cutting it is the closest thing to a free move available to an operating leader.

what the intervention evidence shows

Individual interventions have now been tested at scale. William Fleming, at the University of Oxford's Wellbeing Research Centre, analysed survey data from 46,336 workers across 233 organisations in the United Kingdom. Participants in individual-level mental wellbeing interventions, including mindfulness, resilience and stress management training, relaxation classes and wellbeing apps, were no better off on subjective wellbeing measures than workers who did not take part (Fleming, 2024).

The contrast with organisational interventions is instructive. Panagioti and colleagues meta-analysed 19 controlled studies of burnout interventions covering 1,550 physicians. Interventions overall produced a small reduction in burnout, a standardised mean difference of 0.29. Split by type, organisation-directed interventions produced 0.45 and physician-directed interventions produced 0.18 (Panagioti et al., 2017). Changing the work outperformed changing the worker by roughly a factor of two and a half.

what organisations measure instead

Most burnout dashboards contain four things: an engagement score, employee assistance programme utilisation, wellbeing app adoption and sick days. Each has a problem.

Engagement scores are annual, aggregated and self-reported, and burnout appears in them as a dip that could mean six other things. Assistance programme utilisation measures help-seeking, which rises when conditions worsen and also rises when stigma falls, so the direction of the signal is ambiguous. App adoption measures marketing. Sick days lag by months, and across much of Asia and the Middle East they understate the problem badly, because presenteeism is culturally cheaper than absence.

None of these measures demands or resources. That is the gap.

a diagnostic you can run

Run this at team level. Not company level, where the numbers average into nothing, and not individual level, where the data becomes a surveillance problem. Team is where demands are set and resources are allocated, and it is the only level at which the output is safe to publish.

Build a ledger with three columns for each team.

• Challenge demands. Volume of work that matters, cognitive difficulty, responsibility, time pressure on committed deliverables.
• Hindrance demands. Role ambiguity, conflicting instructions, approval queues, rework caused by others, tool and access failures, meetings that end without decisions, priorities changed mid-cycle.
• Resources. Autonomy over how and when work is done, staffing adequacy, manager support, feedback quality, access to information, skill fit, recognition, social support.

Ask the team to rate every item twice on a five-point scale: how present it is, and how much it costs them. Multiply the two ratings. That produces a hindrance load figure and a resource adequacy figure per team, both comparable across quarters and across teams.

Then apply one rule. You may not add a resource until you have removed a hindrance. Adding resources is politically easy and generates visible activity. Removing hindrances requires somebody to give up an approval step, a control or a report, which is why it rarely happens unless a rule forces it.

Corroborate the self-report with data you already hold, at team level only. Five indicators do most of the work: the share of activity falling outside contracted hours, unplanned work as a proportion of total work, approval cycle time, rework rate, and leave taken against leave accrued. None is a burnout measure on its own. Together they tell you whether the ledger describes something real or something rehearsed.

Maslach and Leiter's areas of worklife model supplies the categories for the conversation that follows. They identify six areas where mismatch between person and job produces burnout: workload, control, reward, community, fairness and values (Maslach and Leiter, 2016). Fairness is the one most organisations skip, and the one the prevalence data flags hardest.

the prevalence numbers and what they imply

Gallup surveyed nearly 7,500 full-time employees and found 23 per cent burned out very often or always, with a further 44 per cent burned out sometimes. Burned-out employees were 2.6 times as likely to be actively seeking a different job, 63 per cent more likely to take a sick day and 23 per cent more likely to visit an emergency room (Gallup, 2018). The five causes Gallup identified were unfair treatment at work, unmanageable workload, lack of role clarity, lack of communication and support from the manager, and unreasonable time pressure. Four of the five are job design. The fifth is management behaviour. None is resilience.

Deloitte's survey of 1,000 full-time professionals in the United States found that 77 per cent had experienced burnout at their current job, and that 91 per cent said an unmanageable amount of stress or frustration harmed the quality of their work (Deloitte, 2018). The top driver named in that survey was lack of support or recognition from leadership.

Gallup also reports that employees who strongly agree they are often treated unfairly at work are 2.3 times more likely to experience high burnout. That is why fairness belongs in the ledger, even though it reads like a culture question rather than a design question. Unfairness is usually a design artefact: unclear allocation rules, inconsistent workload distribution, opaque promotion criteria.

who owns the output

The diagnostic fails if HR owns the result. Every lever in the ledger belongs to an operating leader: headcount, scope, sequencing, approval design, tooling and priority setting. HR's job is to produce the measurement, hold the definitions steady and put the numbers in front of the person who can change them, on a cycle short enough to matter.

That is the practical case for building demand and resource measures into the workforce data you already collect, rather than running them as a separate wellbeing exercise on a separate system. Tools such as entomo's wellness intelligence are useful to the degree that they connect what a team is being asked to carry with what it has been given to carry it. A wellbeing score sitting apart from workload and staffing data cannot make that connection, and a score that cannot make it will be read as a mood report.

Burnout responds to changes in demands and resources. It does not respond much to training people to tolerate the demands they already have. The organisations that improve are the ones that stopped asking employees to cope and started auditing what they were being asked to cope with.

see how entomo does this: wellbeing intelligence.