Nobody samples the market.
Participants choose themselves, and the set changes with every edition. Year-on-year movement is partly composition, not market.
Build your pay system · Using market data
Market data is a reference, not a reason.
Salary surveys tell you what some employers paid for jobs they matched in their own way. That can inspire how you differentiate pay – it cannot tell you what a job is worth, and it cannot justify a gap between women and men.
01
Why survey data is never neutral
Matching is an assumption.
Every result depends on how dozens of employers matched their jobs to generic descriptions.
History is built in.
Market medians reflect decades of lower pay for work done mainly by women. Anchoring to them carries that gap into your structure.
Someone chose the numbers.
Results may be raw, cleaned of outliers, or modelled. You always buy one provider's view.
The European Commission's services put it plainly: market conditions are unlikely to reflect the intrinsic value of a job and should be treated with caution.
European Commission services, FAQ on the Pay Transparency Directive – preliminary views, not an official position. Only the CJEU interprets EU law. European Commission FAQ (via EU action for equal pay)
02
What drives pay – and where each driver belongs
What is the job worth?
Job evaluation on skills, effort, responsibility and working conditions sets the grade.
What is the person worth?
Position inside the band, through written, gender-neutral progression criteria.
| Driver group | Examples | Where it belongs in a compliant system |
|---|---|---|
| Job | Level of requirements (grade), specialism, location | Grade from job evaluation; location as a documented regional factor |
| Company | Industry, size, public or private sector | Which external sources you compare against |
| Person | Relevant experience, qualifications, performance | Position in the band, only through written criteria |
Sex is a driver of pay in every labour market. It is never a legitimate one.
03
Build from the inside out
- 01
Value first
Job evaluation sets the grades. Every job in grade 10 is work of equal value – a nurse, a lecturer, a medical technology engineer, a sales specialist.
- 02
Your pay line
Set the midpoint of each grade from your own pay policy, affordability and internal actual pay.
- 03
Check the level
Compare the overall level with several external references – collective agreements, public pay scales, more than one survey, starting salaries, recruiting data. Adjust the whole line if it is out of step; never re-rank jobs because of it.
Step 3 in practice: grade 10 midpoint check
Result: References range from €67,000 to €73,500. The internal midpoint sits inside that range – no change to the pay line.
04 · One category per grade
A job family group can explain a difference. It cannot define equal value.
Art. 4 · EUR-Lex Art. 9 · EUR-Lex Art. 10 · EUR-Lex Art. 18 · EUR-Lex
- The Directive groups workers by the value of their work. Different jobs of equal value belong in the same category and the same pay range (Art. 4; Commission FAQ on categories of workers).
- Additional factors under Art. 4 (4) must be relevant to the job, justified, gender-neutral and proportionate. A family label or market scarcity says nothing about job content.
- Splitting grade 10 into “engineering grade 10” and “nursing grade 10” would make the gap disappear from the report – and reappear in the first individual claim, because employees can compare themselves with any work of equal value.
- The robust route: keep one category per grade, let differences show, and justify them with evidence (Art. 9, 10, 18).
European Commission services, FAQ on the Pay Transparency Directive – preliminary views, not an official position. Only the CJEU interprets EU law. European Commission FAQ (via EU action for equal pay)
05
Differentiated structures – done properly
Job-specific market pricing is not defensible. A small number of evidence-based structures can be.
A – One structure for hard-to-recruit roles
Attached to roles by evidence, not to a family label. A scarce nursing specialism qualifies as much as a scarce engineering role.
B – A few job family group structures
For example sales, production, engineering, nursing or faculty. More demanding: each needs its own evidence and gender check.
Vary the pay mix before you vary the pay level. A sales role can carry a lower base and a variable target at the same target total pay as the common band.
Grade 10: one category, several structures (fictional organisation)
All roles sit in grade 10: work of equal value, one worker category. Shares of women are fictional examples.
| Role / family group | Women (example) | Structure | Base midpoint | Target variable | Target total | vs. common band |
|---|---|---|---|---|---|---|
| Business support | 78% | Common band | €70,000 | – | €70,000 | – |
| Nursing (ward) | 85% | Common band | €70,000 | – | €70,000 | – |
| Faculty (lecturer) | 48% | Common band | €70,000 | – | €70,000 | – |
| Sales specialist | 30% | Common band, different mix | €62,000 | €8,000 | €70,000 | – (monitor payouts) |
| Medical technology engineer | 15% | Hard-to-recruit structure | €74,500 | – | €74,500 | +6.4% |
| Intensive-care nursing specialist | 90% | Hard-to-recruit structure | €74,500 | – | €74,500 | +6.4% |
Common band: €56,000 – €70,000 – €84,000. Hard-to-recruit band: €59,600 – €89,400 (±20% around €74,500).
Medical technology engineer
- Vacancy duration: 9 months
- Declined offers: 4 of 10
- Turnover: above organisation average
- Market data: two independent sources
- Share of women and men: 15% women · 85% men (example)
- Approved: executive board · Next review: in 12 months
Intensive-care nursing specialist
- Vacancy duration: 9 months
- Declined offers: 4 of 10
- Turnover: above organisation average
- Market data: two independent sources
- Share of women and men: 90% women · 10% men (example)
- Approved: executive board · Next review: in 12 months
- The structure is attached to roles by evidence, not to a family label – which is why a mostly female nursing specialism sits next to a mostly male engineering role.
- The +6.4% stays visible as a gap inside the grade 10 category. Above 5%, the report must name the reason; without a documented justification, a joint pay assessment follows (Art. 10).
- Sales has a different pay mix at the same target total. Because reports use actual pay, payouts by sex must be monitored.
Eight guardrails
- One grade structure for everyone; differentiated structures only sit on top.
- As few structures as possible; the common band is the default.
- Pay mix first; a different total needs stronger evidence.
- Evidence from two directions: market data from at least two sources plus internal data (vacancy duration, declined offers, turnover).
- Share of women and men recorded for every differentiated structure; a lower structure for a female-dominated group, or a higher one for a male-dominated group, needs the strongest justification.
- A written maximum difference to the common band; anything beyond needs executive sign-off.
- Annual review; when the evidence goes, the difference goes – for new hires and in the progression rules.
- Monitor actual payouts by sex, because reporting uses actual pay, not target pay.
Any difference must be justified in proportion by the employer (Enderby, C-127/92 · EUR-Lex ), who carries the burden of proof (Art. 18).
Where a single role cannot be filled and no differentiated structure applies, a separate premium may be used. Before paying it, the employer must be able to show three things (Commission FAQ, following Enderby, C-127/92 · EUR-Lex ):
Relevant and justified
The factor is linked to a real, documented recruiting need.
Gender-neutral
It is free of bias and open to anyone in the role.
Proportionate
Its size matches the evidence, no more.
Five rules
- Documented evidence.
- A separate, named pay component.
- Time-limited, with annual review.
- Available to anyone in that role, regardless of sex.
- Shown and explained in Art. 7 answers and reports.
The employer carries the burden of proof (Art. 18).
European Commission services, FAQ on the Pay Transparency Directive – preliminary views, not an official position. Only the CJEU interprets EU law. European Commission FAQ (via EU action for equal pay)
07
If you use surveys: how they match jobs
Every survey result depends on how your job was matched to a survey job. Three methods exist.
| Method | How it works | Strengths | Weaknesses |
|---|---|---|---|
| Benchmark positions | A full description per reference job; you match to the closest one | Precise matches | Few jobs covered; “close enough” matching distorts results |
| Level descriptions | Generic career paths × levels × job families | Wide coverage; can be mapped to your own grades | Descriptions vary in quality; data spreads thinly across cells |
| Point-factor evaluation | Scored job size combined with specialism | Precise for management roles | Needs method expertise; weaker for specialist roles |
Match by content, not by title. Map your grades to the survey's levels once and document the mapping.
08
Reading the numbers
- Percentiles show how much employers disagree.
- A wide P10–P90 spread means there is no single “market rate”.
- The median is the most stable value.
- Read each job's row as a whole.
09 · Survey quality checklist
Choosing a survey
If you buy market data, ask these questions first.
Result per group
No overall score – each group stands on its own. Nothing is stored.
Purpose
0 of 2 answered · Yes 0 · Partly 0 · No 0 · Don't know 0
Matching
0 of 4 answered · Yes 0 · Partly 0 · No 0 · Don't know 0
Data
0 of 5 answered · Yes 0 · Partly 0 · No 0 · Don't know 0
Currency
0 of 1 answered · Yes 0 · Partly 0 · No 0 · Don't know 0
Effort & cost
0 of 2 answered · Yes 0 · Partly 0 · No 0 · Don't know 0
10
Taking part in a survey
- Plan the effort: matching takes roughly 15 minutes per job with level descriptions, 20–30 minutes with point-factor methods, plus data preparation (rough guidance from practice).
- Keep one master file; copy only the needed fields into each submission.
- Give every job an ID and store its survey matches.
- Focus on jobs that matter for your structures; an 80% fit is a usable match.
- Submit pseudonymised data only, and check where the provider processes it (GDPR).