For years, pay equity was treated as a goodwill gesture: important, but something you could put off. That has changed. With the EU pay transparency rules coming into force, not knowing whether men and women are paid fairly for the same role becomes a concrete legal and reputational risk — not just a matter of principle.
The good news is that pay equity is, at heart, a data problem. The company already has the salaries, the roles, the tenures and the levels; what usually is missing is the method to cross-reference them rigorously and separate legitimate differences from differences that cannot be explained. That is exactly where a well-run audit separates the noise from what matters.
This guide walks, step by step, through how to run a pay equity audit with data: from the essential distinction between gap and inequity, through preparing the data and the statistical analysis, to the hardest part — fixing what you find and stopping it from coming back.
Pay gap and inequity: a distinction that changes everything
The most common mistake is to confuse two different numbers. The unadjusted pay gap compares the average salary of all women with that of all men, without looking at the role. It mainly reflects the structure of the organization: if the top jobs are held mostly by men, the unadjusted gap will be large, even if each person is paid fairly within their role.
Pay inequity, or the adjusted gap, is something else: it measures whether, for equal work or work of equal value, comparable people are paid differently for reasons that should not count — such as gender. This is the number the law targets and the one a serious audit has to isolate. Reporting only the unadjusted gap, without this distinction, creates needless panic or false reassurance.
Why the topic is no longer optional
Directive (EU) 2023/970, on pay transparency, was due to be transposed into national law by 7 June 2026. At the time this article was published, several member states — including Portugal — had not yet completed that transposition, but the obligations are already defined and companies would do well to prepare now. Among the points with a direct impact on people management are:
- Transparency before hiring: giving the candidate the pay or pay range, typically right in the job ad, and not asking about salary history.
- Right to information: workers can ask for their own pay level and the average by sex for comparable work; pay-secrecy clauses are no longer allowed.
- Pay gap reporting: larger companies will publish gender pay gap indicators, with deadlines that apply first to organizations with 250 or more workers and then step down in the following years.
- Joint pay assessment: when the gap in a category exceeds 5% and is not justified by objective, gender-neutral criteria, the company is required to carry out a joint assessment with worker representatives and to correct it.
The 5% threshold is why the analysis cannot be a rushed annual exercise: it is a number the company has to know before it is required to explain it.
The starting point: what data you need to gather
An audit is only as good as the data that feeds it. The minimum needed usually includes:
- Full compensation, broken down: base salary, allowances, bonuses, commissions, benefits and variable pay — not just base salary.
- Role data: title, level or grade, job family, area and location.
- Legitimate factors: tenure, relevant experience, performance, time in level and working pattern (full-time or part-time, converted to full-time equivalent).
- Gender and, if possible, hire date and progression history.
Before any analysis, cleaning pays off: standardize values to full-time equivalent, annualize bonuses, and check for duplicates and gaps. A single badly filled field — a bonus booked in the wrong month, a part-timer counted as full-time — is enough to distort the whole result.
Define "work of equal value" before comparing
You do not compare salaries at random: you compare people doing equal work or work of equal value. Defining those comparable groups is the most subtle step and the one that most influences the result. Grouping too broadly mixes roles that are not really equivalent; grouping too narrowly creates groups so small that no difference is statistically reliable.
The foundation is a minimally consistent job architecture — families and levels — based on objective criteria such as the skills required, effort, responsibility and working conditions. Many organizations discover, at this stage, that they never had a coherent job grid, and that tidying it up is half the value of the exercise.
From the unadjusted gap to the adjusted gap: where statistics come in
With the groups defined, you first compute the unadjusted gap — useful as a portrait, not as a verdict. The decisive step is the adjusted gap: a regression that explains salary from the legitimate factors (level, tenure, performance) and includes gender as an additional variable. The gender coefficient measures how much of the salary is left unexplained after accounting for everything that should count.
If, once the legitimate factors are controlled for, a systematic gender-linked difference remains, that is where the real inequity lies. That is also the number to compare against the 5% threshold. It is wise to pair the estimate with a measure of uncertainty: in a small group, an apparent gap may be just statistical noise, and acting on noise is as bad as ignoring a real problem.
From analysis to correction: what to do with the gap that remains
Finding the problem is the easy part; fixing it takes method and budget. The usual approaches combine:
- Targeted adjustments: correct the unjustified individual cases, starting with the most glaring, with a dedicated equity budget.
- Structural fixes: review the pay bands and progression rules that produced the drift, so you are not bailing out the boat without plugging the hole.
- Prevention at the source: pay ranges set before hiring, raise decisions reviewed for gender impact, and an end to the question about previous salary — which perpetuates inequalities carried over from past jobs.
An audit that ends in a report left in a drawer fixes nothing. The cycle has to close: measure, correct, and measure again in the next period to confirm the gap has not returned.
Common mistakes in a pay equity audit
- Confusing the two gaps: reporting the unadjusted gap as if it were discrimination, or using the adjusted one to deny a structural representation problem. They are different questions.
- Controlling for contaminated factors: if the level or title itself already reflects past discrimination, "adjusting" for it hides the problem instead of revealing it.
- Groups that are too small: drawing strong conclusions from comparisons with a handful of people, ignoring the uncertainty.
- Treating it as a one-off project: an annual snapshot is not enough when every new hire and every raise can reopen the gap.
Mini-case: the invisible gap at a services company
A services company with about 300 employees was relaxed: it had never received complaints about pay. Preparing for the new rules, it ran its first serious audit. The unadjusted gap was 14%, which caused alarm — but much of it was explained by the under-representation of women in senior roles, a real problem, though of a different nature.
The adjusted gap, controlling for level, tenure and performance, came to about 6% and therefore above the 5% threshold, concentrated mainly in two job families. Correcting those cases cost roughly 0.4% of the annual payroll — far less than management feared — and reduced the adjusted gap to under 2%. The most valuable effect, however, was another: the company began reviewing every hiring and raise proposal with an equity check, avoiding reopening the gap with each decision.
In practice
Pay equity has stopped being a debate about values and become a data-based management exercise with a fixed deadline. The path is clear: separate the unadjusted gap from real inequity, gather complete and clean compensation data, honestly define what counts as work of equal value, isolate the adjusted gap with statistics and — the step many avoid — correct and monitor continuously. You do not need an expensive system to start; you need rigor and the will to look at the numbers. Companies that do this work now will not just comply with the law: they will reach fairer decisions sooner, and the trust that builds is worth far more than the cost of the adjustment.
