EUVS KH 29 Workforce Disclosures

A controlled population, calculation and privacy guide for B8-B10

━━━━━━

B8-B10 often draw from different owners: HR holds contracts and gender data, payroll tests pay, health and safety records incidents, learning teams hold training hours, and industrial-relations teams know collective bargaining coverage. If each team chooses its own population, date and denominator, the final numbers can all be individually plausible yet mutually inconsistent.

The reporting team should therefore establish a controlled employee population and reporting calendar first, then reconcile each source system to it. Privacy review comes after the metric is technically correct but before publication, because small groups, health information and collective-bargaining data can create identification or legal risks even when the report contains only aggregate numbers.

Appendix A defines an employee as an individual in an employment relationship with the undertaking. This is narrower than “own workforce”, which can include certain non-employees. B8-B10 expressly refer to employees. Contractors, self-employed people and agency workers should not be inserted into the employee denominator merely because they work on site; they may be relevant to other disclosures or supplementary information.

B8 requires the number of employees in headcount or FTE by temporary/permanent contract, gender and country of employment contract where the undertaking operates in more than one country. These are three views of the same underlying population, so each total should reconcile to the approved employee total, subject to documented categories such as not reported or local legal constraints.

B9 requires both the number and rate of recordable work-related accidents. Appendix A defines a recordable accident as a discrete work-related occurrence causing physical or mental harm that results in death or more than three days of absence. Recordable work-related ill health is an illness caused or made worse by workplace factors that results in death or absence from work.

The 2026 Annex does not state a rate formula in paragraph 41. Historical EFRAG VSME guidance used: number of recordable work-related accidents ÷ total employee hours worked × 200,000. The multiplier represents the annual hours of 100 full-time workers. This remains useful implementation guidance, but the undertaking should approve its formula, disclose it where necessary and use it consistently rather than presenting the historical formula as wording of the 2026 requirement.

Keep the full incident population and the recordable subset separate.

Document treatment of telework, business travel and commuting under applicable law and method.

Do not infer work-related mental ill health without appropriate evidence and legal/medical controls.

Explain material estimation of hours worked and improve the source over time.

B10 asks whether employees receive pay equal to or above the applicable minimum wage for the country reported, determined by national minimum-wage law or a collective bargaining agreement. A group-wide “yes” should therefore be supported by country-level tests, not by comparing average pay with one corporate threshold.

The gender pay-gap datapoint is not universal under B10. It is required when the undertaking is already required by EU law or national regulations to report the percentage gap. The reporting team should therefore document the legal applicability test before calculating or omitting the metric.

Where the metric applies, the methodology must follow the relevant law. Historical VSME guidance illustrated a gap based on average gross hourly pay for male and female employees, but the applicable legal method, population, pay components, exclusions and reference period take precedence. Do not create a separate “VSME pay gap” that conflicts with the statutory calculation.

Collective bargaining coverage is the percentage of employees to whom the undertaking is obliged to apply one or more collective bargaining agreements. It can include non-union employees and may differ substantially from union membership or works-council representation. Count each covered employee once, even where more than one agreement applies.

A simple working formula is covered employees ÷ total employees × 100, using the same controlled population and reporting date or period. The agreement register should show country, entity, employee groups, effective dates and coverage logic.

Training hours should come from a defined learning population and activity taxonomy. The numerator normally includes completed training hours attributable to employees; the denominator should be the approved employee population or an average consistent with the method. The report should state how e-learning, mandatory induction, on-the-job training, partial attendance and multi-year courses are treated where those choices could affect comparability.

Figure 1. Workforce data from multiple systems must reconcile to one employee population before privacy review.

Workforce reporting can involve personal data and special-category data. GDPR principles such as data minimisation are relevant throughout preparation, while health and trade-union-related information receive enhanced protection. The sustainability reporting team should receive only the level of detail needed to calculate and review the aggregate metric, with identifiable source records held under appropriate access controls.

Paragraph 22(d) permits omission of information protected from unauthorised access or disclosure under EU or national law, or to safeguard the privacy or security of a natural or legal person. If used, the undertaking identifies the omitted disclosure in B1 and reassesses the exemption at every reporting date. This is not a general licence to omit any small or sensitive number.

Figure 2. Privacy and small-number controls are applied after the metric and legal trigger are understood.

The group creates a country mapping from legal contract types to temporary/permanent and retains the mapping. It removes contractors from the B10 training numerator, calculates average training hours using an approved average employee denominator, and reports the four recordable accidents and the approved rate method. Fatality information is assessed under the legal restriction and privacy process; the group reports only the permitted aggregate and documents any paragraph 22 omission in B1.

The gender pay gap is disclosed only for the entity or group scope for which a legal reporting requirement applies, using the statutory method. The report does not imply that absence of a group-wide B10(b) figure means no pay analysis has been performed.

Evidence needed: HR population extract, contract-country mapping, HC/FTE method, incident classification, hours-worked calculation, legal wage tests, pay-gap applicability memo and calculation, bargaining agreement register, learning extract, privacy assessment and sign-offs.

1. Using payroll recipients as the employee population. Payroll can include contractors, directors or inactive people. Reconcile to the employment-relationship definition.

2. Mixing year-end headcount with average FTE without explanation. Choose and document the basis for each metric and reconcile to B1.

3. Treating all incidents as recordable accidents. Apply the defined recordability test and retain the full-to-recordable bridge.

4. Presenting the historical 200,000 formula as mandatory 2026 wording. Label it as an approved methodology or historical guidance, not a paragraph 41 formula.

5. Reporting the pay gap because data are available, then calling it B10-required. First test the legal trigger and apply the relevant statutory method.

6. Using union membership for bargaining coverage. Use employees to whom the agreement must be applied.

7. Counting training enrolments rather than completed hours. Define completion and duration rules.

8. Publishing tiny cells without re-identification review. Apply privacy and legal controls before release.