Most DEI reports I see start and end with one chart: the share of women and men in the company, sometimes split by country or age. That chart is useful, but it answers only one question — who is here. It says nothing about who gets hired, who gets promoted, who gets the biggest raises and who quietly leaves.
In 21 years in HR I have learned to treat diversity the same way I treat corporate culture: you change it through leadership, education and management practices, and you prove it with numbers. This guide gives you the metrics I use, the formula that turns them into a clear signal, and examples you can copy into your own report.
If you already have your data, go straight to the free DEI gap calculator.
In 21 years in HR I have learned to treat diversity the same way I treat corporate culture: you change it through leadership, education and management practices, and you prove it with numbers. This guide gives you the metrics I use, the formula that turns them into a clear signal, and examples you can copy into your own report.
If you already have your data, go straight to the free DEI gap calculator.
What are DEI metrics?
DEI metrics are numbers that show two things: how diverse your workforce is and whether different groups have equal outcomes in the processes that matter — hiring, promotion, pay, development and retention.
In my Diversity & Inclusion course I list five qualities an organization needs to manage diversity successfully: leadership, reliable and credible performance measurements, education, cultural and managerial changes, and follow-up activity — ongoing evaluation and performance management. Measurement is the second item for a reason. Without it, leadership commitment stays a statement, and training has no result anyone can check.
The second point from the course: diversity work is a process that is complex, continuous and measurable. Data collection and target indicators are used to find gaps and close them, and review committees validate plans, set goals and measure progress.
In my Diversity & Inclusion course I list five qualities an organization needs to manage diversity successfully: leadership, reliable and credible performance measurements, education, cultural and managerial changes, and follow-up activity — ongoing evaluation and performance management. Measurement is the second item for a reason. Without it, leadership commitment stays a statement, and training has no result anyone can check.
The second point from the course: diversity work is a process that is complex, continuous and measurable. Data collection and target indicators are used to find gaps and close them, and review committees validate plans, set goals and measure progress.
Which groups to measure: primary and secondary dimensions
Before you choose metrics, decide which differences you are going to look at. The course uses two levels:
One rule protects people: report only aggregated groups, and only when a group is large enough that no individual can be recognised. If a department has two people in a group, merge it with the next level up.
- Primary dimensions — age, ethnicity, gender, physical abilities, race, sexual orientation.
- Secondary dimensions, which can change over a person’s life — education, place of residence, social class, marital status, religious beliefs, employment, life experience.
One rule protects people: report only aggregated groups, and only when a group is large enough that no individual can be recognised. If a department has two people in a group, merge it with the next level up.
Representation metrics: who is in the company
Representation = employees in the group ÷ total employees × 100%
Calculate it for the whole company, then for management levels, departments and job families.
Representation is the starting point, and the most informative version of it is by level. A company can be 40% women overall and still have very few women among senior managers. That pattern — balance at the bottom, imbalance at the top — is usually the first thing a real DEI analysis shows.
Representation also becomes a goal. When I was HR director at Preply, diversity was not a slogan for me, it was a KPI: we were an international company, our clients were outside the country, but our team came mostly from one country. My goal was to increase the share of employees from other countries. Later our CTO set himself a goal of his own — to hire more women into the technical team, which was a real challenge because there were far fewer women developers on the market than men.
Representation also becomes a goal. When I was HR director at Preply, diversity was not a slogan for me, it was a KPI: we were an international company, our clients were outside the country, but our team came mostly from one country. My goal was to increase the share of employees from other countries. Later our CTO set himself a goal of his own — to hire more women into the technical team, which was a real challenge because there were far fewer women developers on the market than men.
The lesson I took from that goal: a representation target only moves when you change a process — in our case, where and how we searched for people. That is why the next group of metrics matters more than the headcount chart.
Outcome metrics: what happens to each group
In the course I show this on the simplest example. Say the company is 40% women and 60% men. Fine. Now look at the people who were promoted this year — what is the ratio there? Among those who got the highest pay raises? Among those who got the highest bonus percentage? Among those who left — is it women who leave, and not men?
Each of these questions becomes a rate you can calculate for every group:
Each of these questions becomes a rate you can calculate for every group:
- Hiring rate = hired from the group ÷ applicants from the group.
- Promotion rate = promoted in the group ÷ headcount of the group.
- Top pay raise rate = people in the group who got a raise in the top band (for example, the highest 20% of raises) ÷ headcount of the group.
- Top bonus rate = people in the group in the top bonus band ÷ headcount of the group.
- Exit rate = leavers from the group ÷ average headcount of the group.
The parity ratio: one number that shows a gap
Parity ratio = lower rate ÷ higher rate
1.00 — equal outcomes. Below 0.80 — a gap worth investigating.
The 0.80 threshold is the well-known four-fifths rule used in adverse impact analysis of selection: if one group’s selection rate is less than four-fifths of the rate of the group with the highest rate, the process deserves a closer look. I apply the same screening logic to promotions, raises and bonuses. For exits, flip it: the question is which group leaves more often.
Example. A team of 200: 80 women and 120 men. During the year 6 women and 14 men were promoted. Promotion rate: 6 ÷ 80 = 7.5% for women, 14 ÷ 120 = 11.7% for men. Parity ratio: 7.5 ÷ 11.7 = 0.64. In the same year 14 women and 12 men left: exit rate 17.5% against 10%, so women left 1.75 times as often.
Representation in that team looks reasonable. The rates show a group that is promoted less and leaves more — and those two facts are probably connected.
Example. A team of 200: 80 women and 120 men. During the year 6 women and 14 men were promoted. Promotion rate: 6 ÷ 80 = 7.5% for women, 14 ÷ 120 = 11.7% for men. Parity ratio: 7.5 ÷ 11.7 = 0.64. In the same year 14 women and 12 men left: exit rate 17.5% against 10%, so women left 1.75 times as often.
Representation in that team looks reasonable. The rates show a group that is promoted less and leaves more — and those two facts are probably connected.
Inclusion metrics: how people experience the company
Diversity you can count in your HR system. Inclusion you measure through people’s experience, and the easiest way is to reuse what you already collect:
- Engagement or pulse survey results by group. Compare the average score and the share of favourable answers between groups. A gap here usually appears before a gap in exit rates.
- Exit reasons by group. If one group names the manager or the lack of career growth more often, that is your first hypothesis.
- Policy review by the people it affects. A practical step from the course: give your policies to employees from minority groups and ask them whether they see anything discriminatory. The person who wrote a policy rarely notices it.
Process and programme metrics: is the strategy real?
Outcome numbers tell you where the gap is. Process metrics tell you whether the company is doing anything about it. The self-assessment questions from my course work as a checklist — each “no” is a missing metric:
- Is there a diversity strategy, and is it viewed as a business strategy?
- Has a budget been allocated to implement it?
- Does it depend on management performance indicators — is it in someone’s KPIs?
- Is there an official process for implementing it?
- Is it owned by a senior person who is responsible for implementation?
- Are similar measures taken for gender, disability, age, sexual orientation, faith and religion — not only for one dimension?
How to measure the results of DEI training
Training is often the biggest line in a DEI budget, and the hardest to prove. The course framework is simple: training is needed only when the goal is an improvement in DEI metrics, and you agree with business leaders in advance how the change will be confirmed and by whom. Then you pick the evidence by result type:
- Change in behaviour — interviews and observation; sources are clients, colleagues, direct reports, participants and their leaders.
- Stakeholder opinion — satisfaction surveys, interviews, focus groups.
- Business indicators — data from company systems and reports: the hiring, promotion and exit rates above.
- Work product — for example, job ads or policies compared against standards and evaluation criteria.
Seven mistakes in DEI reporting
- Reporting only headcount shares. Representation without outcome rates hides the gaps in promotion, pay and exits.
- Dividing by the total headcount. Every group rate must be divided by the size of that group.
- Ignoring levels. The company-wide average hides a balanced base under an unbalanced top.
- Treating 0.79 as a verdict. The threshold is a signal to look at the process, especially when groups are small.
- Reporting tiny groups. If people can be recognised, you break confidentiality and lose trust in the survey next year.
- Targets without owners. A number nobody has in their KPIs will not move. At Preply the goal sat in my KPIs, and the CTO set his own.
- Measuring once. DEI is a continuous process. Run the same cut every review cycle and track the trend.
For the formulas behind headcount, hiring and exit rates, see HR Metrics: 27 formulas with examples. The full diversity programme — how to persuade the company to address the topic, design and evaluate training, facilitate difficult conversations and support minority groups — is in my DEI Course: Diversity and Inclusion Training.
