How to Conduct a Pay Equity Audit, Step by Step

What is a pay equity audit, and how is it different from a pay gap report?

A pay gap report compares the average pay of two groups across the whole organisation, usually men and women. It tells you who sits where in your pay distribution. A pay equity audit asks a harder question: are people doing equal work, or work of equal value, paid equally, and where they are not, is there an objective reason that has nothing to do with sex or another protected characteristic?

You need both. The overall gap is what many reporting rules ask for, and it shows structural issues such as too few women in senior roles. The like-for-like audit is what tells you whether any individual is being paid unfairly. Mixing them up causes most of the confusion in pay conversations: a 16% overall gap does not mean women are paid 16% less for the same job, and a small overall gap does not prove that nobody is underpaid.

For a sense of scale, the UK's Office for National Statistics reported in October 2025 that the gender pay gap in median hourly earnings was 6.9% among full-time employees and 12.8% among all employees in April 2025. One organisation's figures measure something narrower, so use national numbers only as context.

Why audit pay now?

Expectations have shifted. More jurisdictions now ask employers to show pay ranges to candidates, stop asking about previous pay, let employees request pay information and report gaps. The European Union's Pay Transparency Directive is the best-known example, and rules in parts of North America and Australia go in a similar direction. Because these rules change quickly, check the current position with your legal adviser for every country where you employ or advertise.

Research suggests transparency narrows gaps, though not always in the way people expect. Michael Baker and colleagues found that public salary disclosure laws reduced the gender pay gap among Canadian university faculty by roughly 20 to 40 percent (American Economic Journal: Applied Economics, 2023). Morten Bennedsen and colleagues found that a Danish rule requiring firms to report wage statistics by gender narrowed the gap by about 2 percentage points, mostly because men's wages grew more slowly (Journal of Finance, 2022). Zoë Cullen's 2024 review in the Journal of Economic Perspectives adds that transparency among co-workers can lead employers to bargain harder, lowering average wages. How you do it matters.

The practical argument is simpler. Pay that can be explained survives being discovered. Pay that cannot be explained does not, and salary websites, published ranges and conversations between colleagues make discovery a matter of time.

What data do you need for a pay equity audit?

Ask payroll for one row per employee at a fixed snapshot date, with a pseudonymous ID rather than a name. Agree every definition before the extract: what counts as base pay, which allowances are ordinary pay, what counts as bonus and over what period.

Then clean it. Check for duplicate IDs, missing gender, FTE values outside 0.1 to 1, contracted hours that disagree with FTE, levels that do not exist in your bands, and dates that cannot be right. Confirm the headcount with payroll. Do not read a single gap until every check is clean or explained.

• Employee ID, job family and level (ideally from a job evaluation)
• Location, if a written location pay policy applies
• Gender as recorded by the employee, including options such as Not disclosed
• FTE and weekly contracted hours
• Start date and date into current level
• Annual base salary, allowances and bonus paid in the last 12 months
• Calibrated performance rating, only if ratings have been checked for bias

How do you calculate mean and median pay gaps, quartiles and bonus gaps?

Use hourly pay for headline gaps so part-time and full-time staff compare fairly: annual base plus ordinary allowances, divided by weekly hours times 52. Check how your reporting rules define hourly pay, because definitions differ.

Read the mean and median together. When the mean gap is much larger than the median, a small number of very high earners, usually in the most senior roles, is pulling the average up. Then look at quartiles: the share of each group in the top quarter of earners is often the clearest explanation of the overall gap.

How do you compare like with like?

Build comparator groups. Equal work groups contain people in the same or broadly similar jobs, typically the same job family and level. Equal value groups contain people in different jobs that demand a similar overall level of skill and knowledge, responsibility, effort and working conditions. The International Labour Organization set out the equal value principle in its Equal Remuneration Convention of 1951, and many countries now include it in law.

Equal value comparisons are only as good as your levels. If levels come from a documented, gender-neutral job evaluation, the comparison is meaningful. If they come from titles or history, treat the results as a reason to evaluate jobs properly.

Set a threshold for flags (5% is a common screening level; check what any local rule uses) and a minimum group size. Review every flagged group on one page: why the jobs belong together, the numbers, the factors you checked and the people furthest below expected pay. Groups too small to compare still matter: review those people one by one.

How do you separate explained from unexplained gaps?

A gap is explained only by factors that are written in your pay criteria, applied the same way to everyone, unrelated to sex directly or indirectly, and large enough to account for the difference. Level, location under a written policy, relevant time in role, calibrated performance and a documented market premium can qualify. Previous salary, negotiation and 'it was approved at the time' rarely do.

Specialists often use regression, which estimates how pay rises with each factor and whether a difference remains after all of them. It is powerful but easy to get wrong with small groups, outliers or the wrong factors, and it should be built and checked by a specialist before any decision with legal consequences.

A transparent alternative works for most small and mid-size firms. Build each person's expected pay from your own structure: band midpoint for their level, times the location factor, adjusted for time in level and calibrated performance. Compare actual pay with expected pay. The difference between the groups after each factor shows how much that factor explains; what remains is the unexplained gap. Keep the order of factors the same every year, because the order changes how the explained part is split.

In the composite case in the Pay Transparency and Equity Audit Kit, a 16.6% raw gap in full-time-equivalent salary breaks down as 14.6 points explained by level, almost nothing by location, time in level and performance combined, and 2.3% unexplained, concentrated in two engineering groups where starting offers had been anchored on previous salary.

How do you fix a pay gap fairly?

Raise pay; never cut it to close a gap. Set a fair floor, such as 95% of expected pay, and raise everyone below it whatever their gender: it is fairer and easier to defend than adjusting one group. Fix the unexplained part of the gap through pay, and the structural part, such as fewer women in senior roles, through hiring and promotion.

Cost the plan person by person, add employer social costs, and phase it only if you must, worst cases first and later phases committed in writing. Keep adjustments separate from the annual pay review so they are not quietly absorbed. Then fix the cause: remove pay history from offers, reallocate accounts on documented criteria, include part-time staff in bonus pools, catch up pay for people returning from leave.

How do you publish pay ranges and answer pay questions?

Audit before you publish. Build each advertised range from the band and your location policy, check that it is not so wide it says nothing, and run an internal equity check: count current staff in the role who are paid below the minimum you are about to publish, and fix their pay first. Tell people in the role before the advert goes live.

When questions come, managers should never reveal a named colleague's pay. They can explain the criteria, the band, where the person sits and, for groups large enough to protect privacy, the average pay of people doing the same work. Name a date for the answer and keep it, and never penalise anyone for asking or for discussing their own pay. Formal requests and anything that sounds like a discrimination concern go to HR the same day.

AI tools can help draft adverts, letters and reports, map job titles and spot unusual patterns in group averages, using an approved tool and anonymised data. They must not decide anyone's pay. To check how well your managers and HR team judge pay decisions, AssessAll's Pay and Reward Decision Assessment for Managers and HR Teams is one option, and Jobulary can turn the gaps it shows into a development plan.