Collective Labour Agreements (CLAs) and the EU Pay Transparency Directive: What Counts as a Category of Workers?

Every gender pay gap report, every employee information request, every joint pay assessment under the EU Pay Transparency Directive rests on one foundational decision: how you group your people into categories of workers. Get this right, and the rest of your compliance program follows a clear, defensible path. Get it wrong, and even a well-run pay equity analysis can produce numbers that don't hold up to scrutiny. For comp & benefits teams building their approach now, understanding how the Directive defines this concept, where collective labour agreements (CLAs) fit, and where local law adds its own twist, is one of the most valuable moves you can make.

Can you just use your collective labour agreement?

Sometimes, but not automatically. Many organizations with an existing CLA or CBA assume the classification levels already baked into that agreement can double as worker categories. Using CLA classifications can be appropriate, particularly where it's the only practical way to group employees with worker representatives' agreement. However, this only holds up if the CLA classification is itself grounded in gender-neutral criteria, since courts have repeatedly declined to treat CLA classifications or union agreements as automatically binding when assessing equal value.

In practice, this means a CLA grade structure is a strong starting point, not a free pass. Before adopting it wholesale, check whether the grading logic reflects the factors the Directive cares about and actually upholds the equal pay principle, or whether it reflects something else entirely, like seniority, historical bargaining outcomes, or role titles that no longer match what people actually do.

How are member states handling this differently?

The Directive sets the principle; national transpositions decide how prescriptive to get, and how much they add to or subtract from the core criteria. As of the transposition deadline, only a handful of member states had enacted legislation, and of those, Lithuania and Italy have been the most prescriptive about defining worker categories. 

Beyond that headline, though, the picture is less uniform than a simple restatement of the Directive's wording would suggest. Several member states have already adjusted the core skills, effort, responsibility, and working conditions test, either narrowing it or building it out with country-specific factors.

A few examples of how that plays out:

Italy anchors categories directly in the national collective bargaining agreement, and its definition omits "effort" as a standalone factor, focusing instead on skills, responsibilities, and working conditions. Under its transposition, internal job evaluation frameworks are allowed, but they have to work alongside the statutory CCNL classifications rather than substitute for them, making the CLA the primary reference point for what counts as equal value. 

The Czech Republic takes a similar reductive approach, dropping "working conditions" as a distinct category in favor of complexity, responsibility, and strenuousness.

Other member states have gone the other way, expanding the criteria rather than trimming them. Spain's definition would include factors like meticulousness, repetitive movement, isolation, forced postures, and care skills. Denmark, France, and Greece would explicitly require employers to evaluate non-technical and soft skills, such as communication, collaboration, and social or emotional competencies, alongside the traditional four factors. Finland goes furthest, breaking the criteria down into granular sub-factors that explicitly include psychosocial demands and interaction skills.

France is still drafting its own transposition (at the time of writing). The draft would expand the definition of work of equal value to explicitly include non-technical skills and working conditions, and would let employers establish worker categories through a company agreement or by relying on industry-level CBAs, with a unilateral employer decision, after consulting the works council, only where no such agreement exists. 
As the law hasn't been finalized, employers there should treat these details as directional rather than settled.

The throughline: there is no single, stable "four-factor test" you can apply the same way in every country. How much weight local law gives to existing bargaining structures, whether it adds or removes factors, and how much freedom employers get to define categories themselves, all vary considerably. That makes a one-size-fits-all category structure risky for any organization operating across multiple EU countries, and makes checking the local transposition a required step, not an optional one.

What typical challenges come up when defining categories?

Beyond the local-nuance question above, three practical issues tend to surface again and again once organizations start building out their category structure:

• Small and uneven groups: Splitting a workforce into narrow categories, by job family, grade, and location all at once, quickly produces groups with only a handful of people, sometimes with only one gender represented at all. Once a category drops below roughly 5 to 10 employees, or contains only men or only women, regression-based pay gap analysis loses statistical footing, and the numbers become hard to defend.
• Structural imbalance: When most women in a category sit in one job family or grade, it becomes difficult to separate a genuine gender pay effect from the effect of job structure itself, particularly in job families that are heavily male- or female-dominated to begin with.
• Multi-entity and multi-country reporting: Organizations with legal entities of different sizes across Europe face uneven reporting thresholds and inconsistent data granularity, which complicates any attempt to apply a single category logic group-wide.

None of these challenges are reasons to avoid the exercise. They're reasons to test your category structure before committing to it.

What's the best-practice approach?

Our recommendation is to aim for the balance point between several competing pressures at once: categories detailed enough to reflect genuine differences in skills, effort, responsibility, and working conditions (as locally defined); broad enough to remain statistically workable; and large enough to protect individual data privacy, since categories that are too small can make it possible to infer an individual's pay from a group average.

A few practical steps make that balance easier to find:

• Start broad: Where categories aren't already dictated by local law or an existing CLA, base them on grades or role levels grounded in a job evaluation methodology, rather than job titles or org-chart position alone.
• Combine where it's justified: Grades that reflect genuinely similar work on those four factors, such as adjacent sub-grades, can often be grouped together. Roles with fundamentally different pay logic, like sales and engineering, should stay separate even if pay levels happen to overlap.
• Simulate before you commit: Build a simple job family by gender by headcount table and flag any group under 5 to 10 people, or any group made up of only one gender, before finalizing your structure.
• Bring in worker representatives early: Since categories often need their buy-in, and since the Directive expects them to be involved in a non-arbitrary classification process, looping them in during the design phase avoids rework later.
• Validate: Compare your methodology against mandatory factors and official methodologies, where available.

Plan for the exceptions: an 80/20 rule

In practice, defining categories of workers is rarely straightforward, and a minority of categories will always fall short of what's needed for reliable statistical modeling. 

A useful way to plan for this is a practical 80/20 rule: aim for the majority of your categories to have sufficient headcount for regression-based analysis, which also keeps them above data privacy thresholds — below those thresholds, for example Article 7 data has to be rerouted through worker representatives rather than shared directly with an employee, since it carries a reidentification risk, a layer of complexity worth planning for early. For those larger categories, regression modeling lets you control for objective, gender-neutral factors, such as tenure or job family, to determine whether an unjustified gap crosses the Directive's 5% trigger for a joint pay assessment and remediation.

For the minority of categories with headcounts too small for the regression results to support remediation decisions, a manual, case-by-case average pay comparison, backed by documented reasoning for any gap, is the more defensible basis for action. The same applies at the extremes: in single-gender categories no pay gap can be calculated at all, and in heavily skewed ones the figure is technically calculable but too unreliable to act on. In all these cases, manual review remains the best way to catch and address individual pay outliers before they become bigger problems.

How beqom can help

Defining categories of workers is ultimately a judgment call. It's informed by the Directive, local law, your CLA, and your own job architecture. beqom's solution is built to support that judgment, not replace it.

beqom’s flexible configuration lets you map your Category of Worker grouping to internal grades, or to statutory classifications like Italy's CCNL. From there, you can simulate different category definitions side by side and see the resulting pay gap for each scenario. That makes it easy to spot where a proposed structure is too small or too skewed before you commit to it, and where remediation is needed once you do.