How to run a pay equity review and close unexplained gaps
Separate the pay differences you can explain from the ones you cannot, then give every remaining gap an owner, a budget and a date.
Most employers that find a pay gap cannot say how much of it has a reason behind it and how much does not. A pay equity review exists to answer that question, and this guide shows how to run one with the data a typical payroll and HR system already holds. In the UK, employers with 250 or more employees must report gender pay gap data each year the threshold is met on the snapshot date, according to the gender pay gap guidance on gov.uk. In Canada, federally regulated employers with 10 or more employees must act proactively to review their practices and correct gender wage gaps, according to canada.ca. The duty is the reason to start. The review is what makes the duty something you can defend.
Before you start
Three things must be settled before anyone opens a spreadsheet. The first is authority. The person running the review needs access to individual pay records, job codes and start dates, and must be able to ask line managers why a decision was made. A review run by someone who cannot ask questions produces a tidy report and no change.
The second is definitions. Agree which pay counts (base salary only, or bonus, allowances and overtime), which snapshot date you use and which employees are in scope. Then check how your statutory figures were built. The UK guidance has been revised in places. Its calculation sections now refer to sex rather than gender, it says the sex field covers employees holding a Gender Recognition Certificate, and its section on recording employees is flagged as under review. Those changes affect how you set up your data fields, so take current advice from a lawyer before you fix them.
The third is ownership. Decide, before the analysis begins, who receives the result and who can approve pay changes. In our experience, reviews fail less often because of the arithmetic than because the findings land in a committee with no budget. The gov.uk guidance includes a section on closing the gap, covering understanding your gap and deciding actions, and a linked guide to creating an action plan for employers with more than 250 employees. Read both before the first meeting so that the review is built to feed them.
- A pay record for every employee in scope, with the pay elements you have agreed to count.
- A job code or grade for every role, and a written description of what each job involves.
- A start date, contracted hours and work location for each person, so the obvious explanations can be tested.
- A named sponsor on the board or executive team who will receive the result.
- Step 1
Fix the population and the pay
Decide which employees are in scope, which pay elements you will count (base salary, bonus, allowances, overtime) and which snapshot date you will use. Use the same definitions as your statutory figures, so the internal and external numbers can be reconciled line by line. What usually goes wrong: part-time staff or fixed-term contracts appear on one list and not another, and the gap moves without anyone noticing. Write the scope down and do not change it once results are visible.
- Step 2
Measure the raw gap
Calculate the simple difference in average and median pay between the groups you are comparing, and the share of each group in each pay quartile. This is the figure an outside reader will quote, so it must be right before anything else is done. Expect it to be larger than the final unexplained figure. The distance between the raw gap and the adjusted gap is the work the rest of the method does, and it is not a reason to stop early.
- Step 3
Group comparable work
Sort roles into groups where the work is comparable: same job family, similar grade, similar responsibility. Judge comparability on what the job involves, using factors you agree in writing before you look at pay, not on titles, which vary by company and by whoever wrote them. The Canadian duty is framed around work of equal value, so the comparison is about the work itself. What usually goes wrong: title-based grouping hides people doing similar work at different grades.
- Step 4
List the legitimate reasons
Record the factors that can legitimately explain pay differences: role and grade, length of service, performance rating, location and shift pattern. Record each factor for every employee, not only for the people it helps. A factor missing for one group cannot explain anything, and anyone reading the file will see it as selective. Decide in advance which factors you will treat as unreliable, such as a performance rating set by a manager whose judgements are themselves under review.
- Step 5
Test the factors statistically
Estimate how much of the gap remains once comparable employees are compared. A common approach is a regression of pay on the explanatory factors plus a marker for the group in question; the size of that marker is the part of the gap the factors do not account for. Neither government source sets out a method, so bring in a labour economist or statistician to choose and check the model. What usually goes wrong: a factor that the bias itself shaped, such as promotion history, is added to the model
- Step 6
Sort explained from unexplained
Put each remaining difference into one of two piles. Explained differences have a documented, job-related reason that applies in the same way to everyone it touches. Unexplained differences have no such reason, or the reason applies to one group more than another. Write the sorting rule down before the results come back. Changing the rule after you have seen the numbers is the most common way a review ends up explaining away the thing it was meant to find.
- Step 7
Trace each unexplained gap
Go to the managers responsible for starting pay, bonuses and promotions, and ask how those decisions were made. Look at where each group sits within its pay range, when pay rises were awarded, and who was put forward for stretch assignments. The cause is usually one of a few things: starting salary negotiation, discretionary bonuses, or who is nominated for promotion. Treat each finding as a fact about a process, and say so in the report.
- Step 8
Set the remediation plan
For each unexplained gap, write one action, one named owner, a costed budget and a review date. Cost the change before you commit to a target, and test it against the next pay round. Where the UK action-plan guidance applies, use it as the template for the document, because neither government source gives a remediation timeline of its own. What usually goes wrong: a percentage target with no mechanism behind it, which closes on paper and reopens at the next salary review.
- UK employers with 250 or more employees must report gender pay gap data (gov.uk)
- 250
- Federally regulated employers in Canada with 10 or more employees have a proactive pay equ
- 10
- UK deadline to publish gender pay gap information after the snapshot date (gov.uk)
- one year
What usually goes wrong
The most common failure is to run the analysis once, find a result that looks explained and stop. Three patterns recur. Reviewers add factors until the gap shrinks, then cannot say why those factors were chosen. They leave out factors that are recorded for only one group. And they treat the answer as a verdict on individuals when it is a finding about a process. In our experience, a gap that survives every fair test usually comes from how offers, bonuses or promotions are made, and the cure is a change to that process rather than a series of conversations with the people affected. A second failure is timing. A review published after the pay round has closed can only describe the past, so schedule it before budgets are set.
Pay equity aims to ensure that employers provide you with equal pay for doing work of equal value.
- Do we need a lawyer before we start?
Yes, in our view. The review handles individual pay data and may lead to legal and reporting questions. A lawyer or employment adviser should confirm the current UK guidance on recording sex and gender, and how your pay data may be held and shared. Ask them to read the definitions before the first spreadsheet is built.
- Which statistical model should we use?
Neither government source in our research sets out a method, so this outline is not a specification. Choose the model with a labour economist or statistician who can explain it to your board in plain terms. Write down the factors and the model before the results come back, so the choice cannot be shaped around the answer you hoped for.
- Does a company below these thresholds need to bother?
The duties in our sources do not apply below the thresholds: 250 employees for UK gender pay gap reporting, and 10 federally regulated employees for the Canadian proactive duty. Employers under Part II.1 of the Parliamentary Employment and Staff Relations Act fall under a separate arrangement, according to canada.ca. A smaller employer can still use the same method for its own purposes, though the statutory deadlines in these sources do not apply to it.
- How often should we repeat the review?
The UK reporting duty applies every year the 250-employee threshold is met on the snapshot date, according to gov.uk. Our judgement is that an annual review, run before pay decisions are made, is a sensible cadence for any employer in scope. Run it again after a major change to pay structure or grading, because the old factors may no longer describe the workforce.
- Who enforces the pay equity duty in Canada?
The Office of the Pay Equity Commissioner, which is housed within the Canadian Human Rights Commission, administers and enforces the Act and provides support and tools to covered organisations, according to canada.ca. Lori Straznicky was appointed federal Pay Equity Commissioner on November 1, 2023, as a full-time member of the Commission for a five-year term.
This week, do three things. Write down the decision the review must inform, such as the next pay round or a change to promotion criteria, and set the date by which it must be ready. Pull one extract for a single job family and check that the pay elements, the snapshot date and the sex field match your statutory figures; any mismatch is a finding in its own right. Then book thirty minutes with your chair or executive sponsor and agree who receives the result and who can approve changes. The rest of the method depends on those three items, so none of it should start until they exist.
