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

Diversity hiring metrics are the same thing risk management calls adverse impact analysis: measuring whether a hiring process is producing substantially different outcomes for different groups of applicants, using the same statistical methods regardless of what an employer’s broader diversity policy looks like. That framing matters because it means the metrics themselves are useful to any employer, whatever their position on diversity programs as a matter of policy, since an unmonitored process can create legal exposure and quietly filter out qualified candidates without anyone noticing which step is doing it.

This guide covers which metrics to track, how to calculate them, how to tell a real signal from statistical noise, and what to do when a number actually flags a problem.

Why Track These Metrics at All

The U.S. Equal Employment Opportunity Commission logged 88,531 new discrimination charges in fiscal year 2024, a 9.2% increase over the prior year, and recovered nearly $700 million for more than 21,000 people. Adverse impact, a statistically significant gap in selection rates between groups, is the pattern that triggers a large share of those cases. It doesn’t require intent. A facially neutral practice, a scored test, an education requirement, a resume-screening tool, can produce it without anyone deciding to discriminate.

The regulatory landscape shifted in 2025: an executive order directed the EEOC to deprioritize disparate-impact-only enforcement at the federal level. That changed enforcement emphasis, not the underlying law. Title VII itself wasn’t amended, private plaintiffs retain full standing to sue over disparate impact, and state laws in California, Illinois, Colorado, and New York City, among others, still apply the same disparate-impact standard regardless of federal enforcement priorities. For an employer weighing whether tracking these metrics is still worth the effort, the practical answer hasn’t changed: the legal exposure that adverse impact analysis is designed to catch is still live, just routed through different doors than it used to be.

The Core Metric: Selection Rate by Group

A selection rate is the share of applicants from a given group who advance to the next stage or get hired: the number selected divided by the number who applied. The EEOC’s four-fifths rule, sometimes called the 80% rule, compares those rates: if any group’s selection rate is less than 80% of the rate for the highest-selected group, the process may show adverse impact and is generally treated as such by federal enforcement agencies as a first screen.

Group B’s impact ratio of 50% falls well below the 80% threshold in this example, which flags the process for a closer look. The four-fifths rule is a screening test, not a legal verdict. It tells you where to look, not what caused what you find.

Where to Measure It: Every Stage of the Funnel

Running the calculation only on final hires misses most of the useful information. A full analysis calculates selection rates separately at each stage: application to screen, screen to interview, interview to offer, and offer to acceptance. Problem areas tend to cluster at one or two specific steps rather than showing up evenly across the whole funnel, which is exactly why stage-by-stage measurement matters more than a single end-to-end number. A gap that appears at the resume-screen stage points to a very different fix than one that first shows up at the offer stage.

How to Read a Four-Fifths Flag Without Overreacting

A four-fifths violation in a small sample can be pure noise. If five candidates from a given group applied and one was hired, that’s a 20% selection rate, and a single additional hire or rejection swings the ratio dramatically. This is the same statistical reality that makes it hard for a small business to build its own predictive hiring model: small numbers produce unstable rates.

The standard follow-up, once a four-fifths flag appears, is a statistical significance test, commonly a Chi-square or Fisher’s Exact test, which asks whether the observed gap is larger than what random variation alone would produce. A result is generally treated as statistically significant at a p-value of .05 or below. A flagged ratio in a small applicant pool that doesn’t clear this bar is worth watching, not immediately treated as proof of a problem. A flagged ratio in a large pool that does clear it is worth acting on.

Beyond the Four-Fifths Rule: Other Metrics Worth Tracking

Applicant pool composition. Compare who applies against the relevant labor market for the role and region. A selection-rate analysis can look clean while a sourcing problem keeps the applicant pool itself narrow before anyone is ever screened.

Time-in-process by group. Consistent delays for one group at a given stage are worth investigating even before they show up in a selection-rate gap, since slower processing can itself depress acceptance rates downstream.

Offer acceptance rate by group. A selection process can clear the four-fifths rule and still lose qualified candidates from a specific group at the offer stage, for reasons that have nothing to do with how they were evaluated.

Early retention by group. If a group is hired at a healthy rate but leaves within 90 days at a notably higher one, the problem may sit in onboarding or team fit rather than in selection itself.

What to Do When a Metric Flags a Problem

Start with the specific stage and the specific tool involved, not the process as a whole. If a skills assessment is the flagged step, check whether it has documented validity evidence for the role, since an unvalidated or poorly targeted test is one of the more common sources of adverse impact and also one of the more fixable ones. If a resume screen or an automated matching tool is the flagged step, review what criteria it’s actually filtering on, since proxy variables, a specific school, a gap in employment history, can produce group-level effects that have nothing to do with job performance.

A single flagged period is a prompt to look closer, not a verdict. A pattern that repeats across multiple hiring cycles at the same stage is a much stronger signal that something in the process itself, not chance, is driving the gap.

Best Practices for Tracking Responsibly

Separate demographic data from decision-makers. Where possible, keep the data used for adverse impact analysis away from the people actively screening or scoring candidates, so the metric measures the process rather than influencing it.

Document methodology consistently. Use the same calculation, the same grouping categories, and the same stages every review cycle, so changes over time reflect the process and not a change in how it’s measured.

Review on a fixed schedule, not just reactively. Quarterly or per-hiring-cycle review catches drift early. Waiting until a complaint triggers a look back means the data has already been accumulating unnoticed.

Keep the underlying assessments validated. The most durable way to keep selection-rate gaps small is to use assessments with real, documented validity evidence in the first place, rather than relying on after-the-fact analysis to catch a problem a better-built process would have avoided.

Frequently Asked Questions

What is the four-fifths rule in hiring?

It’s an EEOC guideline comparing selection rates between groups: if any group’s selection rate is less than 80% of the rate for the highest-selected group, the process may show adverse impact and is generally treated as a screening flag by federal enforcement agencies.

Is tracking diversity hiring metrics still required in 2026?

A 2025 executive order shifted federal enforcement emphasis away from disparate-impact-only cases, but it didn’t change Title VII itself. Private lawsuits and several state laws still apply the same disparate-impact standard, so the underlying legal exposure these metrics are designed to catch remains active.

How many applicants do you need before a selection-rate analysis is meaningful?

There’s no fixed cutoff, but small samples produce unstable rates where a single hiring decision swings the ratio dramatically. A statistical significance test, such as Fisher’s Exact test, helps distinguish a real gap from one driven by a small sample.

Does a four-fifths violation mean an employer discriminated?

No. It’s a screening test that flags where to look closer, not proof of intentional discrimination. Many flagged ratios turn out to be driven by sample size or a specific, fixable step in the process rather than broad discrimination.

What’s the most common cause of adverse impact in hiring?

Unvalidated or poorly targeted screening tools, including assessments that haven’t been checked for validity on the specific role, and proxy criteria in resume screening, such as education requirements or employment gaps, that correlate with group membership without predicting job performance.

Conclusion

Diversity hiring metrics are adverse impact analysis under a more familiar name: a way of checking whether a hiring process is producing the outcomes an employer thinks it’s producing, stage by stage, with enough statistical care to tell a real problem from noise. The legal and political conversation around diversity programs will keep shifting. The math behind the four-fifths rule, and the practical value of catching a selection problem before it repeats for four hiring cycles in a row, doesn’t depend on how that conversation goes.