Diversity Hiring Benchmarks: What to Track and How to Read Them

Diversity hiring benchmarks are reference points that show whether the people you apply, interview, hire and retain reflect the qualified labor market for each role. For a large U.S. employer, the useful set has four parts: representation by job group, selection rates at each funnel stage, availability data for the specific occupation and location, and early retention. A number only becomes a benchmark when it is compared with something: the labor market, your own history, or a peer group of similar employers.

This article covers what to measure, where the comparison data comes from, how to read results by occupation and region, and the data-quality limits that distort results.

TL;DR

• Track four groups of measures: job-group representation, stage-by-stage selection rates, offer acceptance, and first-year attrition, each split by demographic group.
• Compare against availability for the specific occupation and recruitment area, not a national workforce average.
• Use the four-fifths rule as a screening signal that shows where to investigate, not as a verdict.
• Treat low self-identification rates as a data problem first. Below roughly 60% completion, funnel figures have coverage gaps.
• Peer benchmarking among in-house recruiting leaders adds context that public data cannot provide.

Table of Contents

• What a diversity hiring benchmark is
• The measures that matter
• Reading the funnel stage by stage
• Choosing the comparison: availability data
• Interpreting benchmarks by occupation, location and group
• Data quality, privacy and legal limits
• Our perspective
• Peer benchmarking for diversity recruiting leaders
• A practical sequence for building your benchmark set
• Common mistakes
• FAQ
• Sources

What a diversity hiring benchmark is

Three kinds of comparison are in common use.

External benchmarks compare your numbers with the available labor market. Sources include Census Bureau data, the Bureau of Labor Statistics, and industry labor market reports.

Internal benchmarks compare your numbers with your own past results, or compare one business unit, location or recruiter team with another.

Peer benchmarks compare your results with organizations of similar size, industry and hiring profile. They answer a question public data cannot: what is a realistic result for a company like ours?

Each kind has a weakness. External data lags and describes who is available, not who could be reached with better sourcing. Internal data can lock in a past imbalance. Peer data is only as good as the comparability of the group and the consistency of definitions. A sound program uses all three and states which one each target is based on.

The measures that matter

Research on workplace equity metrics from the University of Massachusetts Amherst lists the measures that give a full picture: job-level composition, hiring, promotion, termination and departure flows, pay gaps within and between jobs, and discrimination complaints and outcomes. Hiring is one part of that list, and it is the part talent acquisition leaders control most directly.

For the hiring function, the working set looks like this.

One published framework lists diversity sourcing ratios, interview ratios and pass-through rates, offer acceptance by demographic, and retention of diverse hires among the metrics to follow. The measure that most often gets left out is attrition. A high first-year attrition rate among underrepresented groups signals that the environment may not support those hires, which a hiring-only view would miss.

Hiring metrics versus workforce metrics

Keep the two apart. Hiring metrics describe flows: who entered and left each stage in a period. Workforce metrics describe a stock: who is in the company on a given date. A company can hire a balanced group every year and still show little change in senior levels, because promotion and attrition move the stock. Report both, and name the owner of each.

Reading the funnel stage by stage

The most actionable benchmarks sit inside the funnel. The method is to track demographic selection rates from application to screening, screening to interview, interview to offer, and offer to acceptance, then compare each group's rate with the group that has the highest rate at that stage.

Using the four-fifths rule

The four-fifths rule is a screening tool. When a group's selection rate is below 80% of the highest group's rate, the stage is flagged for investigation. The same guidance recommends applying it to identify stages to look at more closely, not to conclude that a process is unlawful or sound.

A worked example with invented numbers, for illustration only:

• Group A: 200 candidates reach the interview stage, 40 receive offers. Selection rate 20%.
• Group B: 100 candidates reach the interview stage, 12 receive offers. Selection rate 12%.
• Ratio: 12% divided by 20% is 60%, below the 80% threshold.

That result says nothing about cause. The next steps are to check the sample size, look at which requisitions drive the gap, review interviewer panels and scorecards, and see whether the pattern repeats over several quarters. Small samples swing widely, so a flag on 20 candidates is a prompt to keep watching, not to act.

Where gaps usually appear

The stage where the rate drops points to different fixes.

• Application to screen: job description wording, screening criteria, résumé review practices.
• Screen to interview: recruiter judgment, knockout questions, referral weighting.
• Interview to offer: panel composition, unstructured questions, scorecard discipline.
• Offer to acceptance: compensation, candidate experience, competing offers, time to offer.

A policy brief from the National Institute for Women's Research discusses hiring structured around necessary skills and qualifications rather than subjective impressions as a way to make selection fairer. The same brief reports that employers with more centralized hiring contacts were less likely to fail to interview qualified Black candidates, and that employers with employee resource groups were more effective at recruiting qualified diverse candidates. These are findings about practices, so they suit use as design inputs when you decide what to change after a stage-level flag.

Choosing the comparison: availability data

The most common error in diversity hiring benchmarks is comparing a hiring result with the wrong denominator. A national workforce average is rarely the right one.

The better method is to compare applicant pool diversity with the qualified labor market for the role, using the relevant Standard Occupational Classification data from the Bureau of Labor Statistics. Harvard Kennedy School research on setting goals notes that employers may use internal availability estimates, such as current employee qualifications, or external labor force availability from Census Bureau data, local colleges, or industry labor market reports.

U.S. data sources

Workforce availability analysis commonly draws on the American Community Survey, degree and certification data, local university or college outputs, government reports, company workforce reports, and industry labor market availability. Practical notes for each:

• American Community Survey: gives occupation by geography and demographic group. Match the occupation code to the job, not to the department name.
• Degree and credential data: useful for roles with formal entry requirements, such as engineering, nursing or accounting. It reflects the supply entering the market, which is more forward-looking than current incumbency.
• Local college output: relevant for early-career programs and campus recruiting.
• Industry labor market reports: helpful for niche roles where census categories are too broad.
• EEO-1 data: the EEO-1 Component 1 report is a mandatory annual data collection for private sector employers with 100 or more employees, and for federal contractors with 50 or more employees meeting certain criteria. It reports workforce data by job category, sex, and race or ethnicity. It is the baseline snapshot many large employers already have.

Do not benchmark only against today's supply

Harvard Kennedy School cautions that relying solely on available talent pool data risks perpetuating historically institutionalized inequality. Its suggestion is to set forward-looking goals that factor in educational trends and the organization's own aspirations. In practice, report two numbers side by side: current availability and the target the organization has chosen, with the reasoning for the gap.

Interpreting benchmarks by occupation, location and group

National averages hide differences. Benchmarks should be tailored to occupations, industries and geographies, since a technology hub and a logistics center face different labor markets. Compare internal incumbency with availability in the reasonable recruitment area, and update the benchmark periodically.

Occupation matters in concrete ways. In 2018, women were 56.5% of U.S. retail workers, compared with 43.5% who were men, according to the Census Bureau. A company-wide gender benchmark that blends retail roles with engineering roles would be meaningless for both. Each role family needs its own comparison.

Group jobs carefully

Job grouping decides whether a benchmark means anything. Guidance from the U.S. Department of Labor's OFCCP describes effective job grouping as avoiding the mixing of jobs with different content, wages, or opportunities, and treats it as a prerequisite for a meaningful utilization analysis. The same document explains that grouping unlike roles together can mask underutilization in specific jobs. It also stresses that opportunity means upward mobility and access to career-advancing roles.

A short checklist for grouping:

1. Group roles by content, pay range and career path, not by department.
2. Keep entry-level and senior roles in separate groups, even when titles are similar.
3. Review groups when roles are redesigned or a business is acquired.
4. Document the logic so a year-over-year comparison stays valid.

Interpreting small numbers

Small groups produce unstable percentages. If a role family hires 15 people a year, a single hire shifts the rate by almost seven points. Pool several years, combine similar roles, or report counts next to percentages. Avoid ranking recruiters or hiring managers on small-sample rates.

Data quality, privacy and legal limits

A benchmark is only as reliable as the demographic data behind it. Collection should be voluntary and separate from the application materials evaluated for selection, with access restricted to HR analytics. Keeping evaluators from seeing the data protects both candidates and the integrity of the measurement.

Self-identification rates need their own metric. If fewer than 60% of candidates complete voluntary demographic forms, funnel data has significant coverage gaps, and response rates below 50% can distort funnel calculations. To improve completion:

• Explain why the data is collected and who can see it.
• Place the form after the application, not inside it.
• Include a decline option and make every field optional.
• Report completion rates by source channel to find where data is missing.

Legal context matters, and counsel should review any plan that sets targets or changes selection practices. Federal requirements on selection procedures and reporting apply to many large employers, and state laws differ. The approach in this article is analytical: measure, flag, investigate, and document. When a flagged pattern turns into a complaint or an internal concern, a documented process for workplace misconduct investigation steps helps HR and legal teams respond consistently under EEOC and NLRB rules.

Our perspective

We have operated peer networking and benchmarking groups for in-house corporate recruiting leaders for nearly three decades. Three observations come from that work.

First, public data answers "who is available." It does not answer "what do comparable employers achieve with the same roles, the same locations, and the same constraints." That second question is the one heads of talent acquisition are asked in budget and board discussions, and it needs peer-sourced, confidential figures.

Second, definitions decide the result. Two companies can report different offer acceptance rates for the same performance because one counts declined verbal offers and the other does not. Benchmarking among peers works when members agree on definitions first and compare second.

Third, the environment shapes the quality of the exchange. Recruiting leaders share more candidly when no vendor is present. Our communities are limited to in-house corporate leaders for that reason, and our diversity-focused group, DSIX, gives diversity recruiting leaders a place to compare approaches without commercial influence.

— Simon

Peer benchmarking for diversity recruiting leaders

Public data covers availability. Peer groups cover practice. IX Communities runs membership communities for corporate talent leaders: TLIX for talent acquisition heads, ESIX for executive recruiting leaders, and DSIX for diversity recruiting leaders. Membership is for in-house corporate recruiting functions, not for independent recruiters or vendors. Membership prices are not published, so details come from the membership pages.

For teams building skills alongside metrics, IX Academy recruiter training offers on-demand courses from $350, live online courses from $750, and team intact courses from $3,000, each as a one-off price. The membership overview also describes the ExecSmart database and talent acquisition books. Members can also hear from outside experts through the guest speaker sessions.

A practical sequence for building your benchmark set

1. Define job groups. Use content, pay and career path. Record the logic.
2. Pull your baseline. Use your most recent EEO-1 snapshot and current HRIS data.
3. Select availability sources per role family. Pair each family with an occupation code and a recruitment area.
4. Measure funnel rates. Calculate pass-through at each stage by group, and flag stages under the four-fifths threshold.
5. Check data coverage. Calculate self-identification rates before trusting any group-level figure.
6. Add retention. Report first-year attrition by group and by manager population.
7. Set targets with reasons. State whether each target rests on availability, history or peer results, and whether it looks forward.
8. Review on a fixed cycle. Update availability data and job groups at least annually, and whenever roles change.
9. Assign accountability. The UMass research recommends developing metrics and goals, making them transparent, and holding decision makers accountable like other business goals.

Common mistakes

• Using a national average for every role. It hides the occupation and local differences described above.
• Reporting only hiring, never attrition. Hires who leave in the first year do not improve representation.
• Reading small samples as trends. Pool years or roles before drawing conclusions.
• Treating the four-fifths rule as pass or fail. It identifies where to investigate.
• Ignoring missing data. A funnel built on 40% self-identification is not a funnel.
• Changing definitions mid-year. It breaks the comparison with prior periods and with peers.

FAQ

What are the most important diversity hiring benchmarks?

Start with representation by job group, selection rates at each funnel stage, offer acceptance rate, and first-year attrition, each split by demographic group. These show where representation changes and whether new hires stay. Add applicant pool composition against availability for the role.

How do I find availability data for a specific role?

Match the role to a Standard Occupational Classification code and use Bureau of Labor Statistics or American Community Survey data for your recruitment area. Add degree data, local college output or industry labor market reports where the census category is too broad. Document the source so the comparison can be repeated.

What does the four-fifths rule tell me?

It flags a stage where a group's selection rate is below 80% of the highest group's rate. The flag is a reason to investigate sample size, requisitions, panels and scorecards. It is not a finding of discrimination or of fairness.

What if candidates do not complete the demographic form?

Low completion makes funnel data unreliable. Published guidance says that fewer than 60% completion leaves significant coverage gaps, and below 50% can distort calculations. Make the form voluntary, separate it from the application, explain how the data is used, and track completion by channel.

Where can I compare results with similar employers?

Public data does not cover peer practice. Peer communities for in-house recruiting leaders, such as DSIX for diversity recruiting, let members compare definitions and results in a confidential setting. Details are on the membership pages.

Sources

• Metrics, Accountability, and Transparency, University of Massachusetts Amherst
• Diversity Hiring Metrics: Numbers to Track and Improve
• Crafting a Diversity Hiring Strategy: Key Metrics
• DEI Goals, Harvard Kennedy School
• Making Equal Opportunity Real, NIWR Policy Brief
• Alternative Sources to Analyze Workforce Availability
• EEO-1 Employer Information Report Statistics, EEOC
• Retail Jobs Among the Most Common Occupations, U.S. Census Bureau
• OFCCP job grouping guidance, U.S. Department of Labor