What are DEI metrics?

DEI metrics help an organization examine diversity, equity and inclusion across its workforce and employment experience. They may cover representation, recruitment, access to development, progression, retention, pay and employees' reported experience.

The value comes from a clear question and a defensible interpretation. A representation percentage can be useful; it is not a complete picture of inclusion. An experience score can reveal a concern; it does not identify the cause of every employment outcome.

This guide helps people teams choose measures, define denominators, review gaps and protect confidentiality. Use DEI dashboards for presenting the measures, and equity metrics for a broader treatment of access and outcome differences beyond employment.

Choose measures across the employee journey

Start with the decisions the organization needs to make and the information it can appropriately collect. Not every organization needs every metric, and a sensitive field should not be collected simply because it appears in a template.

Scroll horizontally to see all columns →

Keep administrative outcomes and survey experience distinct even when reviewing them together. A workforce record may be identifiable for HR operations while a listening survey is anonymous or restricted to approved aggregates.

Create a definition sheet before a dashboard

For each metric, specify the purpose, source, eligible population, numerator, denominator, reference period, group definition, exclusions, missingness and access rules. Name the person responsible for maintaining it.

For representation, decide whether the denominator is the full workforce, employees with a known response to an optional field, or a particular job level. Those choices answer different questions. Report the coverage rather than quietly treating “not disclosed” as a demographic category or excluding it without explanation.

For progression, define who had the relevant opportunity. Comparing promotions to all employees may obscure differences in role, tenure or eligibility. A refined analysis still requires judgment about which factors are legitimate to consider and which may themselves reflect barriers.

CIPD's disability workforce-reporting guidance emphasizes communicating how employee information will be kept confidential. Build that explanation into collection and reporting, not only a dashboard access setting.

Worked example: three measures, three different questions

Fictional example. An organization has 200 employees. Eighty identify with Group A, 100 with Group B and 20 have not provided the relevant optional group information. These generic labels illustrate calculation only.

Scroll horizontally to see all columns →

Do not switch between 40% and 44.4% without explaining the denominator. Neither tells you how employees experience the organization.

Now suppose 40 Group A employees and 50 Group B employees were eligible for a defined development opportunity. Twenty-four and 40 respectively participated. The participation rates are 60% and 80%, a difference of 20 percentage points. Review eligibility, access and context before assigning a cause.

If an anonymous survey reports a concern about development, it can inform a group-level review. It does not authorize identifying the commenters or claiming they are the same individuals who did not participate.

Interpret pay measures carefully

A pay-gap measure summarizes a distribution; it is not interchangeable with an assessment of equal pay for comparable work. Specify whether the measure uses hourly pay, annual earnings, base pay or another defined amount, and whether it uses means or medians.

For example, ONS's 2025 UK gender pay-gap methodology defines its headline measure using median hourly earnings excluding overtime. This is a source-specific definition, not a universal formula for every organization's reporting obligations.

Do not combine incompatible pay measures or turn an unadjusted group difference into a legal conclusion. Role mix, hours, seniority and other factors may matter to the analysis; whether and how to account for them requires suitable expertise and the actual question. Applicable requirements need a separate, current review for the jurisdiction.

Measure inclusion through appropriate employee listening

Ask about specific experiences the organization can understand and address. Examples might concern whether people can raise a work-related concern, understand how opportunities are allocated or feel their contributions are considered. Use an appropriate instrument where needed and test wording with the intended workforce.

A broad “I feel included” score may be useful but can mean different things to different respondents. A limited number of optional open-ended questions can help clarify the experience without creating a long mandatory writing exercise.

Keep the promise made to respondents. If the survey is anonymous, do not join its comments to HR identities. If it is confidential and identifiable for a stated purpose, restrict access and explain that purpose accurately. Group-level analysis is often sufficient for learning.

Use employee survey software requirements to evaluate the collection and reporting controls, and the Employee Experience course to plan the complete listening workflow.

Read comments without turning them into individual judgments

Use clearly defined themes, inspect examples and retain uncertainty. Look for positive, critical and contradictory accounts. A quotation can illustrate an experience; it cannot establish why an entire group has a particular promotion, retention or pay outcome.

With larger datasets, Sopact's approach supports applying human-owned definitions across eligible responses, revising the analysis when definitions change and reading themes alongside appropriate quantitative context. This reduces repeated coding and manual assembly while keeping review necessary.

See the visual and illustrated effort comparison in qualitative and quantitative analysis. Use that workflow with employee confidentiality controls; permission to inspect a theme is not automatically permission to read every original comment.

Do not use sentiment or a theme label as an automatic employee performance judgment, individual attrition prediction or employment decision. Organizational listening and individual HR decisions have different purposes and evidence requirements.

Handle small groups and overlapping filters

Disaggregation can reveal patterns hidden by an overall result, but it can also expose people. Review small group sizes, distinctive quotations and combinations of filters. A seemingly safe table may become identifying when compared with another table.

Choose reporting thresholds and disclosure methods appropriate to the setting, and apply them consistently to dashboards, exports, search results and AI summaries. There is no single minimum group size that guarantees anonymity in every context.

Intersectional analysis may be useful where the question, data and safeguards support it. Do not fragment a dataset into tiny groups merely because the interface allows another filter. Document when the evidence is too limited to support a useful comparison.

Compare over time without losing context

Distinguish a snapshot from a cohort measure. Workforce representation on two dates reflects hiring, departures and transfers. It is not the same as tracking the progression of the employees present at the first date.

For surveys, an anonymous repeated sample can describe changing reported experience, while an appropriately designed panel can examine within-person change. State who is included and how the design affects interpretation.

Retain the definition and collection version for each period. If a question or group category changes, decide whether the trend remains comparable. Document corrections instead of silently revising history.

Align a shared core across locations

Growing organizations may have different local processes, languages and applicable requirements. Agree common measures where comparison is meaningful, then retain local questions and definitions where necessary. Do not force sensitive categories or one entire questionnaire across every location without assessing the context.

A shared dictionary should define the limited common core, sources, periods and treatment of missing values. Where categories or eligibility differ, show the distinction. Aggregation should follow a defensible mapping rather than a common label alone.

Use the findings to improve a process

1. Confirm the result. Reproduce the calculation and check coverage and definitions.
2. Investigate the experience. Review relevant evidence and perspectives without assuming the cause.
3. Select a process to improve. For example, clarify access to a development opportunity rather than promising an outcome from a dashboard change.
4. Assign responsibility. Name the owner, implementation step and review point.
5. Report back. Explain what was learned, what will change and what remains uncertain.

Choose a cadence that fits decisions and evidence availability. A periodic reviewed report may be appropriate. Continuous collection is not automatically more trustworthy and should not become unnecessary individual monitoring.

Evaluate a manageable, governed workflow

HR systems, spreadsheets and BI tools can support valid calculations and reporting. The additional work often lies in reconciling sources, maintaining definitions, reviewing open text, enforcing access across outputs and rebuilding each reporting cycle.

Sopact's connected approach brings recurring collection, appropriate context, analysis and evidence review together. A practical evaluation should include an anonymous survey, a restricted administrative source, missing group data, a definition change and a small-group filter. Test whether trained staff can maintain routine work and reproduce approved results.

Compare implementation and recurring review effort across the whole workflow. Keep authoritative HR records and specialist analysis where they are needed. Verify actual product controls before treating a demonstration as proof of suitability for sensitive employee data.

For a clear narrative, use the report-writing guide and report examples, adapting the structure to the employee audience and confidentiality requirements.

Watch: connect dashboard measures to stakeholder evidence

This companion video discusses evidence behind DEI dashboards. Apply it with the purpose, confidentiality and interpretation boundaries in this guide; identifiable employee disclosure is not required for every metric.

Frequently asked questions

What are useful DEI metrics?

Useful measures can include representation, recruitment and development access, progression, retention, defined pay differences and employee experience. Select measures that serve a clear decision and can be collected appropriately.

Is a diversity percentage enough?

It can answer a representation question, but it does not describe the full employee experience. Review the relevant opportunity, progression and listening evidence alongside it.

Do DEI metrics require identified employee comments?

No. Administrative measures and listening evidence can have different access designs. Anonymous and approved group-level feedback can be valid; do not break the confidentiality promise to create a link.

Does a pay gap prove unequal pay?

A pay-gap statistic and an equal-pay assessment are different analyses. Interpret the chosen measure in context and use appropriate expertise for employment or legal conclusions.

Can an AI summary explain why a group has a lower outcome?

It can help organize relevant accounts, but it does not establish the cause. Review sources, uncertainty and other evidence, and retain human responsibility for interpretation and decisions.

Choose your next step

Put the workflow to work.

Solution

Employee experience

Explore collection, analysis and governance for this workflow.

Explore the solution →

Build your plan, step by step.

Academy course

Build an employee-listening program

Work through practical lessons and prepare a plan your team can use.

Start the course →

Customer story

King Center

Related practice: using participant feedback for program review. This story does not validate employee confidentiality controls or demonstrate a DEI outcome.

Read the customer story →