What it costs, and whether it nets out

A worked example for a 50-person firm, the break-even arithmetic, and how the tax actually computes. The case for doing it at all is at /corporate-case-regulatory; this page is the money.

The conclusion first, so the arithmetic has something to land against: across the plausible range of inputs, only the low-stress, low-effect corner of the model leaves a firm worse off, and at those inputs a firm would not be buying tier-2 provision in the first place. Everywhere else the programme breaks even or comes out ahead. The rest of this page is the working, including the inputs a sceptical reader is most likely to disagree with.

A worked example

The arithmetic below works through a 50-employee professional-services firm taking Tier 4 weekly delivery: Emma on-site each Friday for a half-day visit, four 1-hour sessions per visit, recipients booked into a six-session monthly course before the cohort cycles. The numbers are illustrative; the structure of the calculation matters more than any individual figure, and the sensitivity tables below show how the conclusion holds up against inputs a reader can plausibly disagree with.

These figures are for the sustained programme, not the pilot. A first-time firm usually starts with a pilot: a few weeks of visits to see whether staff book in, come back, and ask when Emma is next on site. It is the cheap way to find out whether the appetite is there before committing to a year of it. A pilot runs at one or two sessions a head, below the dose these figures rest on, so what it proves is take-up, not the savings on this page. The numbers below assume that question is already answered and the firm is running the weekly course.

Programme cost

Per-visit rate (Tier 1 floor, applied to weekly cadence)	£250
Working weeks delivered per year	48
Gross programme cost	£12,000
PSA gross-up (mixed-rate workforce, weighted)	×1.43 to ×2.06
Corporation tax relief (BIM47070, 25% main rate)	×0.75
Net cost to firm, range	£12,900 to £18,500

Net midpoint ~£15,000. Exact figure depends on PSA banding mix across the workforce. This is welfare expenditure under BIM47070, not a discretionary employee benefit; the firm absorbs the tax via PSA rather than pushing it onto the employee’s P11D, which would undercut the duty-of-care framing the spend rests on.

Reach

Slots per visit	4
Sessions per recipient (monthly × 6 months)	6
Concurrent recipients (one cohort)	16
Annual cohort cycles	2
Distinct annual recipients	32
Reach as proportion of workforce	64%

The cost being defended against

Deloitte’s 2024 sectoral breakdown of UK mental-health-related employer cost puts professional services at £2,357 per employee per year, averaged across the workforce (not just the mental-health-affected sub-cohort, who carry a disproportionate share).¹ For a 50-employee firm at the sector average that is £117,850 of absorbed cost per year, decomposing roughly 47% to presenteeism, 39% to turnover, and 14% to absenteeism. Sector spread runs from around £1,500 at the low end to £5,179 in finance and insurance.

Break-even threshold

To net zero at the mid net-cost of around £15,000, the programme would need to reduce firm-level mental-health-related cost by 12.7%, or 20% on a per-recipient basis at 64% reach. Pre-tax, the same threshold is 10.2% firm-level / 16% per recipient. The per-recipient figure is firm-size invariant; the firm-level figure scales inversely with workforce size, so the same programme would require a smaller firm-level reduction at larger headcount, but the same per-recipient reduction either way.

Packheiser et al. (2024)’s adult touch-intervention meta-analysis reports effect sizes of g=0.59 for depression, g=0.59 to 0.64 for anxiety, g=0.69 for pain (137 randomised trials, n=12,966).² A six-session monthly course sits at the modal dosing of the trials in the meta-analysis, so effect-size transfer is well-supported at this dose. Standard mapping puts g=0.59 at roughly a 25% improvement on symptom-severity scales. Symptom-severity improvement is not identical to absorbed-cost reduction; the sensitivity tables below run the cost-reduction figure across a range that accommodates that uncertainty.

Sensitivity: effect translation × sector cost

Annual saving from the programme, with recipients drawn at random from the workforce (no concentration assumption). Rows are the per-recipient cost-reduction rate at end of the six-session course; columns are Deloitte sector cost per employee.

	£1,500/emp	£2,357/emp	£3,500/emp	£5,179/emp
15% per-recipient	£7,200	£11,300	£16,800	£24,900
25% per-recipient	£12,000	£18,900	£28,000	£41,400
35% per-recipient	£16,800	£26,400	£39,200	£58,000

Random-selection saving is firm-size invariant in absolute terms (recipients × effect × per-employee cost), so this grid would look the same at a 200-employee firm. What changes with firm size is the proportional impact: at 50 employees the same £18,900 saving represents 16% reduction of total firm absorbed cost, against the 4% reduction it would represent at 200 employees. Against the mid net cost of about £15,000, the grid splits at roughly a 20% per-recipient reduction at the sector average: above it the absorbed-cost column nets positive, below it the firm is out of pocket on that column. The 15% row runs at a loss in professional services (£11,300 against ~£15,000), and the £1,500 low-cost column only clears at the 35% rate. So on absorbed cost alone the spend pays for itself once the translation rate clears about a fifth, which is the open question flagged above; below that it leans on the higher-cost sectors and on the three returns the grid does not count.

The grid above measures one column of return: reduction in absorbed mental-health cost. Three other returns the same programme delivers sit outside those numbers and do not depend on which cell the firm lands in. The firm gets a documented tier-2 provision artefact (booking log, uptake report, written policy) that contributes directly to the deliberative-quality evidence portfolio under Working Minds, performed before any HSE inspection rather than after one. It gets a visible staff-engagement signal that lands every Friday and is not interchangeable with a wellbeing newsletter or an EAP login. And in a saturated professional-services talent market, “access to on-site massage” is materially more than a token benefit when hiring or retaining qualified staff. None of these are in the sensitivity grid; all are part of what the firm is buying for the same £15,000 a year.

Three scenarios at the professional-services sector cost

The sensitivity grid above assumes recipients are drawn randomly from the workforce. In practice they are not. Self-referral concentrates voluntary health-utilisation toward the symptomatic population (the standard finding in Andersen-model behavioural-health-utilisation literature), and the disclosure-bypass property of massage relative to counselling amplifies the concentration further. At small-firm scale the concentration headroom is limited (random reach is already 64%, so the multiplier ceiling is around 1.5× before all mental-health-affected employees are reached); the scenarios below reflect that ceiling.

Scenario	Concentration	Effect	Saving	Net of £15k
Floor (low effect)	None (random)	15%	£11,300	−£3,700
Conservative	None (random)	25%	£18,900	+£3,900
Central	All MH-affected reached	25%	£25,700	+£10,700
Illustrative ceiling	All MH-affected reached	50%	£51,400	+£36,400

The floor is what a hostile reader will quote, so here it is at its real value: at a 15% per-recipient reduction with no concentration, the absorbed-cost column comes in about £3,700 under net cost. It clears once the reduction passes roughly a fifth, the conservative row. The central case assumes the seven or eight mental-health-affected employees in a 50-person firm are reached over the year, which 64% annual reach makes likely, at the same lower-bound effect; the ceiling holds that reach and lifts only the translation rate. From the conservative row up, and counting the three returns the grid leaves out, the spend covers itself.

Operational constraint

The six-session monthly course is the dose at which Packheiser’s effect-size transfers cleanly. Inconsistent delivery (fewer than four slots filled in some weeks, holiday gaps, partner-blocked Fridays) drops per-recipient dose below the six-session floor and weakens the dose-response argument. Cohort rotation discipline is a service-design constraint, not a marginal optimisation, and the programme’s financial case depends on it.

One distinction from the EAPA reach figure

At 64% workforce reach per year, this worked example covers a substantially higher fraction of the firm than EAPA UK reports for its national counselling reach (around 10% of the mental-health-affected workforce annually, 5% at clinically-meaningful dose, against an EAP-industry headline of higher utilisation that bundles non-counselling contacts). What matters is not the relative size of the two reach figures but what each is being asked to carry. EAPA UK’s reach figure is load-bearing for a legal-sufficiency claim, where single-digit clinical-dose reach is structurally fatal; tier-2 massage provision makes no such claim. It is the additive layer NICE Guideline NG212 §1.6 calls for, sized for the population the EAP does not reach. The argument for the page /corporate-case-regulatory works through this in full.

How the tax computes

Workplace massage paid for by the employer is allowable against trading profits as employee welfare expenditure. HMRC BIM47070: “You should allow expenditure incurred directly by a trader on the welfare of their employees, unless of a capital nature.” Corporation-tax relief at 25% (main rate) or 19% (small profits rate) applies to the gross spend in every route below.

The PAYE Settlement Agreement (Tier 3 Pilot, and Tier 4 by default) is the mechanism by which the firm absorbs the income tax and Class 1B NIC on the employee’s behalf in one annual settlement, without the benefit appearing on individual payslips or P11Ds. Approximate gross-up at basic rate is 43%: a £3,000 pilot becomes approximately £4,290 to the firm after settlement. Higher-rate-payer-heavy workforces gross up further. PSA application must be in place by 6 July following the relevant tax year, per Regulation 105 of the PAYE Regulations 2003.

Net of both. The two adjustments pull in opposite directions. PSA gross-up raises the bill, because the firm is absorbing the income tax and Class 1B NIC the employee would otherwise pay: that is the ×1.43 to ×2.06 in the worked example, depending on the rate mix across the staff. Corporation-tax relief then takes a quarter back off. Applied to the £12,000 gross, the two leave a net cost of £12,900 to £18,500, which is roughly 1.1 to 1.5 times the headline figure. The relief narrows the gap; it does not bring the cost below the sticker price.

Alternatives to PSA. A sustained programme can instead use a P11D-reported benefit-in-kind, where the firm pays full freight and each employee declares the benefit annually and pays their own tax (a £69 session times 12 sessions a year is an £828 benefit, roughly £166 in tax for a basic-rate employee). A co-pay arrangement has the employee contributing a portion and the firm subsidising the balance, with the benefit reportable on the subsidised element only. Both shift the tax onto the employee, which undercuts the duty-of-care framing the spend rests on, so PSA is the default the model assumes.

The trivial-benefits route used at the one-off, small-per-head end of the offer is covered on the pricing page, where it belongs with the tiers it applies to.

Sources: Corporation Tax Act 2009 s.54; ITEPA 2003 s.323A, s.323B; HMRC manuals BIM47070, EIM21864, EIM21865; Regulation 105, PAYE Regulations 2003 (SI 2003/2682).

References

Deloitte, Mental Health and Employers: The case for investment (2024), Figure 16, p.18 (cost per employee by SIC sector). Per-employee cost of poor mental health: professional services £2,357, information and communications £2,571, finance and insurance £5,179. National total £51bn, weighted toward presenteeism. Figures reported as-published in 2024 prices.
Packheiser et al., “A systematic review and multivariate meta-analysis of the physical and mental health benefits of touch interventions,” Nature Human Behaviour 8, 1088–1107 (2024). 137 randomised trials, n=12,966. Adult effect sizes: depression g=0.59, state anxiety g=0.64, trait anxiety g=0.59, pain g=0.69. The meta-regression found session count, not session duration, correlated with effect size, so a six-session course sits within the supported dose range. Effect-size transfer to a self-selecting workplace population is directionally supported rather than a precise point estimate; the sensitivity grid runs a range to accommodate that.

Emma Burgess

Owner-operator, Emma’s Luxury Mobile Massage