Pay Equity Audits: How to Run One

The first pay equity audit at a 600-person fintech did not begin with regression. It began with an argument about job levels. Product managers in London had levels 4 to 7. Product managers in Berlin had "associate, senior, lead." Lagos had no levels at all. The data could not answer fairness questions because the job architecture was not ready.

Pay equity work is analytics, but it is also job design, data hygiene, legal judgment, and leadership discipline.

Define the audit question

A pay equity audit asks whether employees doing comparable work are paid fairly after accounting for legitimate factors. Common focus areas:

• Gender.
• Race or ethnicity where lawful to collect.
• Disability.
• Age.
• Nationality.
• Other protected characteristics under local law.

The first question is usually simple:

"Do employees in comparable roles receive comparable pay after controlling for legitimate factors such as level, tenure, performance, location, and department?"

Clean the data before analysis

You need:

• Employee ID.
• Country and location.
• Role family.
• Job level or grade.
• Department.
• Manager.
• Base salary.
• Bonus or incentive.
• Allowances.
• Equity value, if included.
• Full-time equivalent.
• Tenure.
• Performance rating, if used.
• Protected characteristic fields where lawful.

Group comparable work

Do not compare everyone to everyone. Compare employees doing substantially similar or equal-value work.

Common groupings:

• Same role family and level.
• Same job grade.
• Same country or labor market.
• Similar department where work differs materially.

Small sample sizes are a real constraint. If only three employees sit in a group, regression may be meaningless. You can still review individual cases qualitatively.

Use regression where sample size supports it

Regression estimates whether pay differences remain after controlling for legitimate factors. Controls may include:

• Level.
• Role family.
• Country or city.
• Tenure.
• Performance rating.
• Education or credential where job-related.
• Manager or business unit.

Do not control for factors that may themselves reflect bias without legal review. For example, prior salary can preserve historical discrimination. Performance ratings may contain bias if calibration is weak.

Interpret gaps carefully

Raw gap:

• Median men earn USD 90,000.
• Median women earn USD 78,000.
• Raw gap: 13.3%.

Adjusted gap:

• After controlling for level, department, tenure, and performance, women earn 4.2% less on average.

Both matter. The raw gap may reveal representation problems: women clustered in lower-paid levels. The adjusted gap may reveal unequal pay within comparable work.

Fix root causes, not only salaries

Remediation may include:

• Salary adjustments.
• Promotion review.
• Starting salary governance.
• Salary range cleanup.
• Hiring range discipline.
• Calibration review.
• Manager training.
• Bonus allocation review.
• Equity grant review.
• Job-level corrections.

If women are underpaid because they enter at lower salaries, banning salary-history questions and using hiring ranges may matter more than a one-time adjustment.

Communicate with discipline

Do not overpromise. Do not hide forever. A good message:

"We reviewed pay for employees in comparable roles and made adjustments where pay was not aligned with our compensation principles. We are also improving job levels, hiring ranges, and manager decision processes."

Do not share protected data or individual comparisons.

Key takeaways

• Pay equity audits need clean job architecture and data before statistics.
• Compare comparable work, not everyone against everyone.
• Regression helps when sample size and data quality support it.
• Raw gaps and adjusted gaps answer different questions.
• Counsel should guide privilege, legal risk, and remediation.
• Fix root causes such as starting pay, promotion, leveling, and bonus decisions.