The EU Pay Transparency Directive is built on one idea: equal pay for work of equal value. To compare jobs that are different on paper, a sales coordinator and a payroll administrator, say, you need a way to score them. That’s job evaluation.

Many employers assume they need a licensed method, the Hay method being the best known, and a consultant to apply it. The law doesn’t require either. Article 4 of the Directive sets the criteria, and leaves the method to you, as long as it’s objective and gender-neutral. Recital 26 refers to gender-neutral job evaluation and classification systems in general terms. Malta’s L.N. 173 of 2026 follows the same line in regulation 3(3).

Here’s how to build a scheme of your own that holds up.

The four required criteria

Article 4(4) names four criteria. Every scheme has to include them:

Criterion	What it covers	Sub-factors you might score
Skills	What the job needs you to know and be able to do	Knowledge and qualifications, problem solving, communication and people skills
Effort	The mental, physical and emotional demands	Mental effort, physical effort, emotional effort
Responsibility	What the job is accountable for	People, money and assets, decisions, care of customers and others
Working conditions	The environment the work happens in	Physical environment, schedules and travel, exposure to risk

You can add other factors where they’re relevant to the job. You can’t use anything based directly or indirectly on sex, and the Maltese text adds that soft skills have to be taken into account. In practice that means communication, emotional effort and care for others get their own sub-factors, so roles that rely on them aren’t scored down.

Five steps to build the scheme

1. Score the job

Work from the job description and ask what the role requires. How well the current holder does it belongs in pay progression, under its own documented criteria.

2. Give each sub-factor clear levels

Use a scale of 1 to 5 with a short description of each level, written in plain terms. “Level 3 responsibility for money: approves spending up to a set limit.” Two people scoring the same job should land on the same level.

3. Weight the criteria

Decide how much each criterion counts. A company of engineers might weight skills more; a care provider might weight effort and working conditions more. There’s no single right answer, but there must be a reason, recorded when you set the weights. If you change them, keep the old version and record why. Where employees’ representatives exist, agree the criteria with them.

4. Turn scores into grades

Add up the weighted scores, then set the point ranges for each grade. Every role in a grade counts as work of equal value, across departments. That’s your category of workers for the gap report.

5. Give each grade one pay band

Set a band per grade and check where everyone’s pay falls in it. People below or above their band are the first places to look for unexplained gaps. The band is also what goes in your job adverts under Article 5.

Using AI to draft scores

Reading every job description and proposing a level for each sub-factor is slow work. A model can draft those levels from the job description alone, with a short reason for each, as long as:

it only sees the job title and description, with names, contacts and pay figures removed
it never sees pay, gender or any data about the person in the role
a person reviews, edits and approves every level
the approval is recorded, with the model and the weights used

That keeps the evaluation objective and keeps a human accountable for it.

Checking the result

Before you rely on the grades, test them:

Sense check. Do roles you’d intuitively call similar land in the same grade? If not, is it the weights or the levels?
Gender check. Are roles mostly held by women scored lower on criteria where they shouldn’t be? Look hardest at effort and responsibility for care.
Spread check. Does each grade have a sensible spread of roles, or has one grade swallowed half the company?

How Squadra HR does it

The pay transparency module scores each role on the four criteria with sub-factors you can rename, add or retire. Weights are saved as versions, each with its reason and who saved it, and an agreement with employees’ representatives can be recorded against a version. A model can suggest levels from the job description, with names and pay removed, and a person approves each one. Grades and bands follow from the scores, and the gap report, the 5% flags and the answers to pay requests all name the weight version they used.

No Hay terms, tables or methods are used.

See the pay transparency module, or read The 5% rule and joint pay assessments.

Sources

Directive (EU) 2023/970, Article 4 and recital 26, EUR-Lex
L.N. 173 of 2026, regulation 3, legislation.mt

Job evaluation without Hay