Reach More Qualified Candidates Without Demographic Scoring
Build a diversity recruiting process around wider outreach, job-related assessments, accessible hiring tools, and careful review of selection-stage disparities.

Diversity recruiting should broaden access to jobs and remove unnecessary barriers—not change an applicant’s score because of their demographic identity. For HR teams, that means expanding sourcing channels, checking that requirements reflect the work, and evaluating candidates consistently with appropriate accommodations. Measure where qualified candidates stop progressing, then investigate the selection step responsible.
Enter two groups’ advancement rates to check the four-fifths screening ratio.
Two-Group Advancement Check
Compare rates from the same role and selection stage. Enter percentages, not candidate counts.
Below 80%: investigate the disparity. This is not proof of unlawful discrimination.
40% ÷ 60% × 100 ≈ 67%.
Default example: Group A advances 30 of 50 candidates; Group B advances 20 of 50. These are illustrative figures, not study findings.
A result of 80% or more does not establish lawfulness. This check compares only two groups; use the highest rate across all groups in the relevant comparison. Counts, missing data, and statistical and practical significance still need review.
Source: article’s illustrative counts and the EEOC Uniform Guidelines Q&A; the four-fifths rule concerns race, sex, or ethnic-group selection rates.
The guidance below concerns U.S. federal law and is informational, not legal or employment advice. State and local requirements may add obligations; have qualified counsel review your program’s design.
Keep Outreach Separate From Selection Preferences
The label “diversity recruiting” does not determine whether a practice is lawful. The EEOC’s DEI-related discrimination guidance says employment actions motivated, wholly or partly, by race, sex, or another Title VII protected characteristic may be unlawful. Those protections apply to applicants of all backgrounds. Title VII generally covers employers with 15 or more employees, as well as employment agencies and other covered entities.
Keep three functions distinct:
| Function | Appropriate Focus |
|---|---|
| Outreach | Reach qualified people your existing channels miss, while keeping opportunities open. |
| Evaluation | Apply job-related criteria, consistent scoring, and appropriate accommodations. |
| Monitoring | Investigate disparities and potentially unnecessary barriers, not demographic bonuses. |
Do not improvise different test cutoffs to achieve a representation target. Title VII expressly prohibits adjusting employment-test scores or using different cutoffs based on race, color, religion, sex, or national origin. Specialized remedial arrangements require separate legal review.
Define Job Requirements Before Choosing Sourcing Channels
For each requirement, record what work it supports and how a candidate can demonstrate it. Separate legal or licensing requirements, essential skills at entry, and skills that can be learned after joining.
A support analyst role, for example, might require diagnosing customer problems and documenting resolutions. Unless a particular credential is necessary, consider allowing candidates to demonstrate those skills through relevant work, training, or a realistic exercise rather than automatically requiring a degree.
Review the same requirements in the job advertisement, application questions, ATS filters, and recruiter instructions. Removing a degree requirement from the advertisement changes little if the application still rejects everyone without one.
This is not simply a wording exercise. The EEOC explains that even neutral selection procedures can violate Title VII when they disproportionately exclude protected groups and are not job-related and consistent with business necessity.
Broaden Reach Without Restricting Eligibility
Choose additional channels around the role’s actual skills and location: community colleges, professional associations, workforce programs, accessible job boards, or community organizations. Ask partners to circulate the same opportunity and qualification criteria—not to screen people by protected characteristics.
Keep a publicly accessible application route. Treat employee referrals as one channel, not the whole recruiting strategy. The EEOC warns that discriminatory recruitment and job advertisements are prohibited and that reliance on word-of-mouth recruitment can create legal problems in some circumstances.
Measure more than clicks. For each channel, track completed applications, candidates meeting the predefined requirements, and progression to assessment. A large audience is not evidence that your outreach reached qualified candidates.
Standardize Interviews Without Blocking Accommodations
Write interview questions and scoring criteria before meeting candidates. Replace vague judgments such as “culture fit” with observable evidence: explaining a trade-off, resolving a customer issue, or coordinating a handoff.
The U.S. Office of Personnel Management describes structured interviews as using the same predetermined questions in the same order, with responses evaluated against the same rating scale and standards. Have interviewers record evidence and score independently before discussing candidates.
Consistency does not mean refusing accommodations. Provide a clear contact and enough information about an assessment for applicants to identify an access problem. DOJ’s AI and disability guidance explains that hiring technologies must measure relevant skills rather than unrelated impairments, and that reasonable accommodations are required for applicants with disabilities unless they create undue hardship.
Check What Hiring AI Actually Filters
A vendor’s “unbiased” claim is not evidence that its tool is suitable for your role. Before deployment, ask which inputs influence search visibility, rankings, scores, or rejection. Establish whether the tool compares candidates with previous hires or current employees, and what evidence supports that comparison for the particular job.
Ask which populations were included in accessibility and group-outcome testing. Confirm that the employer can inspect decisions, correct errors, and review changes to the tool. A report on a different population or use does not answer those questions for your hiring process.
A model that compares applicants with successful incumbents may reproduce historical exclusions; DOJ identifies this mechanism specifically in its disability guidance. Vendor documentation also does not transfer responsibility: the EEOC says employers remain responsible for ensuring their selection tests are valid for their use.
Use the AI hiring vendor validation checklist to turn these questions into procurement requirements.
Measure Where Candidates Stop Progressing
Review results by role, both for the overall selection process and for individual stages. Separate employer declines from candidate withdrawals; they suggest different problems. Investigate whether losses occur at an application filter, assessment, interview, or offer stage.
In the illustrative example used by the calculator, 30 of 50 candidates in one racial group advance at an assessment stage, compared with 20 of 50 in another.
| Group | Candidates Advancing | Advancement Rate |
|---|---|---|
| A | 30 of 50 | 60% |
| B | 20 of 50 | 40% |
Assuming these are the only groups being compared, the highest advancement rate is 60% and the other is 40%. The lower-to-highest rate ratio is 40 ÷ 60, or about 67%.
That falls below the Uniform Guidelines’ four-fifths rule of thumb, which compares selection rates for race, sex, or ethnic groups with the highest rate in the relevant comparison. It signals a need for investigation, not automatic proof of unlawful discrimination. Conversely, a ratio of 80% or more does not establish that a process is lawful; sample size and statistical and practical significance matter.
The calculator compares only the two rates entered. If your relevant comparison includes additional groups, identify the highest rate across that comparison rather than relying on a selected pair. Rates alone also cannot establish statistical significance: review the underlying counts and context.
Have counsel and privacy specialists define lawful demographic-data collection and access controls. Keep monitoring data separate from individual selection decisions, document missing data, and do not guess identity from names or photographs.
Assign an owner to investigate each flagged barrier and document the response: validate the requirement, test an equally effective alternative with less adverse impact, or correct an accessibility problem. For candidates, the visible signs of this work are concrete requirements, clear assessment instructions, and a usable accommodation route—not just a diversity statement.