The £24bn problem hiding behind your absence figures

The Office for National Statistics can tell you that the UK lost 148.8 million working days to sickness or injury in 2025. It can tell you that this is 2.0% of all working hours, and 4.4 days per worker. It counts to one decimal place.

It cannot tell you the other number. Nor can you, yet, about your own organisation.

Absence is measured because it leaves a record. Somebody rang in, somebody logged it, and a line appeared in a system. Working while unwell leaves nothing. The person was there, the timesheet is full, and no exception report fires. The cost is real and the trace is absent.

Why does nobody count the days you half-lose?

Start with what is known. ONS put the sickness absence rate at 2.0% of working hours in 2025, unchanged on the year, across 148.8 million working days. That is 4.4 days per worker.

CIPD and Simplyhealth, surveying over 1,100 employers, put the average at 9.4 days per employee, the highest in more than fifteen years. Their series ran at 5.8 days before the pandemic and 7.8 days in 2023.

Two national figures, and why they disagree

4.4 days and 9.4 days are both correct. They measure different things, and a finance director will ask, so it is worth saying plainly.

ONS derives its figure from the Labour Force Survey, where people report absence in a reference week. CIPD asks employers what their own absence records show across a full year. One is a household survey, the other is a payroll fact.

The useful point is not which is right. It is that both are estimates of the same visible thing, and both are counted carefully. Nothing counts the invisible one at all.

What does poor workplace health actually cost an employer?

In November 2025 the Keep Britain Working review published its arithmetic. The Department for Work and Pensions and the Department of Health and Social Care commissioned it. Its technical note puts the annual cost to employers of poor workplace health at around £85bn, and breaks it into four parts.

Read the sick pay line again. Of the £85bn, only the £10bn is the part a business actually sees leave its bank account.

The rest is output that did not happen. The business carries it either way. It simply never appears as a cost, because nothing bills you for work that was not done.

Why do two credible estimates of presenteeism disagree?

Keep Britain Working costs presenteeism at £21bn a year, across all causes. Deloitte, in its fourth report on mental health and employers, costs presenteeism at around £24bn, and that figure covers mental ill health alone.

The narrower estimate is the larger one. That should stop you, and most articles quoting either figure never mention the other.

What explains the gap

Three things, and none of them is that one party is wrong.

• They were built in different years, on different economies, from different surveys. Deloitte published in May 2024, Keep Britain Working in November 2025.
• They model lost productivity differently. Deloitte works from a survey of working adults; the review builds a bottom-up cohort model with published assumptions.
• They price a lost hour differently, and presenteeism is almost entirely a question of how you price a partly productive hour.

What survives the disagreement

Both models, built independently, land presenteeism in the same order of magnitude as absence and sometimes above it. Neither can be checked against a record, because no record exists.

For a board, that is the finding. The precise number is contested. The scale is not.

Why does presenteeism leave no record?

Every reporting system in an organisation is built to record exceptions. Presenteeism is not an exception. It is an ordinary day that produced less than an ordinary day should.

Four things keep it hidden, and only the last one is uncomfortable.

Absence has an owner. Presenteeism has none

Somebody reports the absence rate monthly and it appears on a dashboard. No job description anywhere contains the words "reduced capacity while present". Nothing without an owner gets measured.

It looks exactly like ordinary work

Nothing crosses a threshold. No system flags anything. The person is at their desk or on the line. The work comes out slower, or comes out wrong, which reads as a bad week.

The people affected are the least likely to raise it

Saying it out loud means saying you are not coping. In most organisations that is a career risk, whether or not anyone intends it to be. So it is not said, and the record stays clean.

A manager under pressure would rather have someone there

This is the one nobody writes down. Faced with a target and a short team, a present person is worth more than an absent one today. Tomorrow is somebody else’s problem. The incentive quietly runs the wrong way.

Where does presenteeism hide in your own numbers?

It hides in the gap between what a team should produce and what it does. Some of the pattern is visible in national data. The Health and Safety Executive counts 40.1 million working days lost in Great Britain in 2024/25, to work-related ill health and injury. It publishes the causes by condition.

HSE also reports that each person suffering took around 16.4 days off work on average. That is the tail of the problem, not the whole of it. The days before somebody stops are not in the figure.

Four patterns are worth checking, and you already measure all four.

• Output per head drifting down, with no change in demand and no obvious cause.
• Error and rework rates climbing in one department while the others hold steady.
• Overtime rising while headcount stays flat, which is a team buying back its own lost capacity.
• A team that only hits its numbers when everyone stays late, which is capacity borrowed rather than earned.

None of these proves presenteeism on its own. Machines wear out, demand shifts, a supplier slips. Several of them together, in one team, at one time, is a signal worth following.

Where it concentrates

ONS reports that workers in process, plant and machine operatives occupations had the highest sickness absence rate of any occupational group in 2025. It also puts part-time workers at 2.8% against 1.9% for full-time.

For manufacturing, logistics and warehousing that matters twice over. The visible absence is already higher there. The invisible part sits under the same conditions.

CIPD found that 35% of organisations report presenteeism has risen because of homeworking. Note what that is: employers reporting a rise in something none of them can measure.

A test you can run this week

You do not need a project to start. You need one team and three numbers you already hold.

Take one team and three numbers

1. Output per head for that team, monthly, for the last twelve months.
2. Overtime hours for the same team, over the same months.
3. The team’s absence rate, over the same months.

Put the three on one axis. You are looking for a period where absence stayed flat or fell while overtime rose and output per head did not.

That shape is the interesting one. It says the team held its attendance and lost its capacity, which is the definition of the thing nobody is recording.

Then ask the question the data cannot answer

Take the chart to the person who runs that team. Ask one question: in those months, how many of your people were at work and not able to do the work?

They will know. Line managers always know. What they have never had is anywhere to put the answer.

What a useful answer looks like

A useful answer is a number and a reason, at team level. "Roughly four of my twenty, mostly through the winter, mostly the same people." That is a diagnosis you can act on.

An unusable answer is "everyone is fine". Not because it is dishonest, but because nothing has ever asked the question, so nothing has ever been thought about it.

Which objections are worth taking seriously?

Four come up every time somebody raises this in a boardroom. Three of them are fair.

"Everyone has an off day"

True, and a day is noise. Nobody should try to cost a bad Tuesday.

The claim is about pattern, not incident. One team, several months, output per head down while attendance holds. A day is noise. Two quarters is a signal, and it is a signal you already have the data to see.

"This is engagement research with a new name"

It overlaps, and it is not the same question. Engagement asks how people feel about the work. This asks what the organisation actually got.

The two are connected, and Gallup measured the connection at unit level. What is proposed here starts from output rather than from sentiment, because output is what a board is accountable for.

"We already run a wellbeing programme"

Then you have already done the harder half, which is getting the budget agreed. What is usually missing is the number it moved.

Attendance at a session is not a result. It is an input. The result is what happened to output per head, to rework, to overtime, in the teams the programme reached, against the teams it did not.

"Our people would not tell us"

Some would not, at first, and that is a reasonable thing to expect. It is also why this is not built on asking people how they feel.

The organisational measures do not require anybody to disclose anything. Output per head, rework and overtime are already recorded, already anonymous, and already yours.

Does acting earlier really return more?

Yes, and it comes from the same research as the cost. Deloitte reports an average return of £4.70 for every £1 invested in workforce mental health and wellbeing. That average comes from a review of 26 studies published since 2011.

Read that carefully, because it is not the usual claim. It does not say spend more. It says spend sooner, which is only possible if you know where the problem is before it becomes an absence.

The £4.70 is an average across a body of research, not a promise about your organisation. What it establishes is direction, and direction is enough to justify measuring first.

What does a diagnostic measure, and what must it never measure?

Organisational metrics, never individual ones. Gallup, in its Q12 meta-analysis. It compared top-quartile and bottom-quartile business units. It found a 23% median difference in profitability and a 78% median difference in absenteeism.

Both are unit-level comparisons. Nobody had to look inside a single person’s health record to produce either one, and that is the point rather than a footnote.

Health data about an identified person is special category data under UK GDPR, and it carries obligations well beyond an ordinary staff survey. Our approach is built on aggregates for that reason, which is both the defensible position and the more useful one.

In practice that means a baseline before anything else. Workforce Diagnostics establishes where performance is leaking and what it costs. What follows comes from that answer, not from a menu.

What changes for each seat at the table?

The same number lands three ways, and it should. A board decision needs all three.

• For the managing director it is a mandate question. Capacity you already pay for is not arriving, and nobody currently owns the number that would prove it.
• For the operations director it is a frontline question. The drift shows up as rework, overtime and missed throughput, and it will already be familiar.
• For the finance director it is a business case. £4.70 returned per £1 is the published average, and the point of a baseline is to replace an average with your own figure.

That is also why this is sold as advisory and not as a platform. We have been delivering workplace health to UK organisations since 2010, and the work that moves a number is diagnosis first.

What should you do with the absence report you already have?

Do not replace it. It is measuring something real and it is measuring it correctly. Put a second question next to it.

Triage what you already produce

Look at your last twelve monthly reports. For each one, ask a question. Would anything on that page have moved if a third of the workforce had been at work and at half capacity all year? If not, the report cannot see it.

Then ask three things, in order

1. What is our absence rate, and what does it cost us in output rather than in sick pay?
2. Where does output per head diverge from the plan, by team, over the last year?
3. Is that divergence spread evenly, or concentrated in two or three places?

If the answer to the third is that it is concentrated, you have found where to look. If nobody can answer any of them, that is the finding, and it is a common one.

The number underneath the number

You already know your absence rate. It is reported, it is trended, and somebody owns it.

The more useful question sits underneath it and almost nobody reports it. How much of your workforce is at work today and below capacity, and what is that costing you?

If you cannot answer it yet, that is the place to begin. Tell us what your absence rate is and we will tell you what it is costing, and what taking it down is worth.