Diversity Hiring Strategy:How to Build a More Inclusive Pipeline
Most diversity hiring efforts fail not because teams lack intent, but because they address the wrong problems. They run one unconscious bias training and call it done. The actual issues are upstream: exclusionary job descriptions, a short list of sourcing channels, and unstructured processes that favor familiar profiles over proven capability.
McKinsey's 2023 "Diversity Wins" report found that companies in the top quartile for ethnic diversity are 39% more likely to outperform peers financially. That number has grown every three years since 2015. The causal mechanism is not that diverse teams check a compliance box. It is that they bring different perspectives to decisions, which reduces blind spots.
The problem is that knowing diversity matters and knowing how to build it are different things. A lot of well-intentioned teams make the same set of mistakes: they optimize for awareness without changing their process, they track headcount at hire date but ignore pipeline drop-off, and they treat unconscious bias as a one-time education problem rather than an ongoing process design problem.
This guide is about process, not sentiment. Each section covers a specific lever you can pull: how you source, how you write job descriptions, how you structure screens and interviews, and how you measure what is actually happening in your pipeline. For teams that want to go deeper on reducing systematic bias, Google re:Work's unbiasing research is the best publicly available resource.
Before any of this, though, you need to understand where diverse candidates are actually leaving your pipeline. Without that data, you are guessing at solutions to an undiagnosed problem. Run a hiring process audit first if you have not done one.
The Core Problem
Where diverse candidates drop out of unstructured pipelines
The most common mistake in diversity hiring analysis is measuring only the end result: who got hired. That tells you where you landed, not where you lost people. When you map representation at each pipeline stage, a different picture emerges. Diverse candidates are often present at the top of the funnel but disappear at specific steps. Each drop-off point has a distinct cause.
Where diverse candidates drop out of a typical unstructured pipeline
Starting point varies by sourcing channel
Drop-off often tied to exclusionary JD language
Unstructured screens amplify interviewer bias
Homogeneous panels favor familiar profiles
Salary gaps persist without structured offers
Illustrative data based on SHRM and LinkedIn research patterns. Your numbers will vary.
The drop from 28% at job views to 16% at offer extended is not inevitable. It is a signal that process failures are compounding at each stage. Fix the right stage and the downstream numbers move. Fix the wrong one and nothing changes.
LinkedIn's Global Talent Trends data shows that companies with strong inclusive hiring practices see 22% lower early-tenure attrition among diverse hires. The pipeline problem and the retention problem are connected. Candidates who experience equitable processes are more likely to accept offers and stay.
Step 1
Source from channels where diverse talent actually is
Most teams post the same three places: LinkedIn, Indeed, and their own careers page. That reaches plenty of candidates, but it largely recycles the same networks. If your workforce is homogeneous, it is because your sourcing is homogeneous. The fix is expanding channels, not replacing them.
Employee referral programs are often the biggest blind spot. They feel efficient, and they are, but they systematically favor people who already look like your team. Harvard Business Review research shows that referral hires come from demographically similar networks to the referring employee in over 70% of cases. Referrals should be one channel among many, with active counterbalancing.
PowerToFly, Jopwell, DiversityJobs, iHispano
Howard, Spelman, Cal State LA, campus recruiters
National Society of Black Engineers, SASE, Out in Tech
LinkedIn sourcing with blind skills-based criteria
Partnerships with HBCUs and Hispanic-serving institutions take more time to build, but they produce lasting pipelines. Several companies, including Deloitte and Google, run multi-year campus programs with these institutions specifically because the relationship builds trust and brand recognition that job postings cannot replicate.
For active sourcing on LinkedIn, remove degree filters and company prestige filters from your search criteria. Use skills-based keywords instead. You will see a different candidate pool, often with equivalent or stronger relevant experience. Pair this with your passive candidate outreach strategy and the quality of your top-of-funnel improves across the board.
Step 2
Rewrite your job descriptions to stop filtering by accident
Job descriptions are the first filter in your pipeline. Most teams write them by copying an old posting and adding a few updates. That means old exclusions get inherited without anyone noticing.
Research from Textio and similar tools consistently shows that gendered, jargon-heavy language reduces application rates from women and underrepresented groups by 20-35%. But you do not need a paid tool to fix the basics. The patterns below cover the most common problems. Our full guide on writing effective job descriptions goes deeper on structure and language.
The single highest-impact change: add a salary range. Candidates from lower-income backgrounds and women are statistically less likely to apply to roles without listed compensation. Pay transparency also reduces negotiation gaps that widen compensation inequity after hire.
- Focus on outcomes: "drive revenue growth" not "must have 10 years experience"
- Use gender-neutral language: avoid terms like 'rockstar', 'ninja', 'crushing it'
- State the salary range explicitly in the posting
- List only requirements that are truly non-negotiable
- Mention specific flexibility: remote options, parental leave, async work
- Describe team culture honestly, not just aspirationally
- "Culture fit" as a qualification criterion with no definition
- Degree requirements for roles where a degree has no bearing on performance
- "5-7 years experience" requirements for junior-level compensation
- Lists of 15+ requirements that no single candidate will meet
- Jargon-heavy company speak that signals insider culture',
- "Fast-paced startup environment" without saying what that actually means
Degree requirements are the most common unexamined filter. A 2023 SHRM study found that 61% of job postings required a four-year degree for roles where internal data showed degree had no correlation with performance. IBM, Google, and Apple removed degree requirements from large portions of their job postings years ago, not as a PR move but because it expanded their candidate pool without degrading quality.
The right question for each requirement is: does this predict performance in this role? If you cannot cite specific evidence, it probably should not be in the posting. Cut the list to must-haves, move the rest to preferences, and watch your application funnel change.
Step 3
Build structure into your screening process
The screening stage is where implicit bias does the most damage, partly because it is low-stakes-feeling. Recruiters make quick decisions on incomplete information and those decisions compound across hundreds of candidates.
Structured screening means evaluating every candidate against the same criteria in the same order. This sounds obvious but most screening is improvised. A recruiter spends 15 minutes with one candidate and 22 minutes with another, asks different questions, and then makes a subjective call. The candidates who "feel right" advance.
Define pass/fail criteria before you start screening
Write down three to five must-have signals before you look at the first resume. Any candidate who meets those advances regardless of how their background was acquired. This prevents you from creating the criteria after seeing candidates and retroactively favoring certain backgrounds.
Use blind resume review for the first pass
Names, schools, and zip codes carry demographic information. Research from the National Bureau of Economic Research found that resumes with Black-sounding names receive 36% fewer callbacks than identical resumes with white-sounding names. Removing that information for the first review removes a substantial portion of that bias.
Use skills assessments where relevant
Work sample tests and skills assessments predict job performance more accurately than most interview techniques, according to decades of meta-analysis. They also shift the evaluation from where candidates went to school to what they can actually do. For technical roles, this can dramatically change who advances.
Track your screen-to-interview rates by demographic
If certain groups are advancing at significantly lower rates through screening, that is a process problem to investigate, not a candidate pool problem to accept. You cannot fix what you do not measure. Log this data every quarter and share it with the full recruiting team.
The goal of structured screening is not to slow down the process. It is to make your decisions defensible and consistent. An AI screening tool can support this by applying criteria consistently across high application volumes, as long as the criteria themselves are built without bias.
Step 4
Structure your interviews to evaluate capability, not comfort
Unstructured interviews are essentially conversations. People are better at conversations with people who are similar to them. That is why unstructured interviews systematically disadvantage candidates who are different in background, communication style, or cultural reference points, even when interviewers have every intention of being fair.
The fix is not training interviewers to try harder. It is changing what the interview looks like. Structured interviews ask every candidate the same questions in the same order, evaluated against pre-defined criteria using an interview scorecard. Google's internal research, documented in their re:Work guides, found that structured interviews predict job performance twice as well as unstructured ones, while producing more equitable outcomes.
Four non-negotiables for inclusive interview design:
Every interviewer uses the same question bank, calibrated before the loop starts
Scorecards are submitted independently before the debrief, preventing anchoring from the first voice in the room
Interview panels include at least one person from a different demographic than the hiring manager
Interviewers are trained to distinguish job-relevant signals from cultural familiarity signals
Diverse interview panels matter not just symbolically but practically. A 2017 study in PNAS found that panels with at least one woman reduced gender bias in outcomes by 25-46%. The same principle applies across other dimensions of diversity. When candidates see only one type of person evaluating them, it signals something about who fits in your organization. That signal affects who accepts offers.
My view on "culture fit" as an interview criterion: retire it. If it has no clear definition that all interviewers share, it becomes a proxy for comfort with familiar types. Replace it with "values alignment" and define the specific values you are assessing. See our guide on structured interviewing for how to build this into your process.
Step 5
Track pipeline diversity, not just hire diversity
The most common DEI metric is representation at hire. It is also the least useful one for improving your process. It tells you where things ended up, not what went wrong or where to intervene.
Pipeline diversity metrics tell you where candidates from different groups enter, where they drop out, and whether they are being treated consistently throughout. These are the numbers that give you something actionable. The four below are the minimum set for any team running more than 20 hires per year.
Application rate by demographic
What it is: % of applicants who self-identify by group
Why it matters: Shows whether your JDs and sourcing attract diverse talent
Target: matches or exceeds labor market representation
Screen-to-interview pass rate
What it is: % of screened candidates who advance, by group
Why it matters: Surfaces bias in recruiter screening decisions
Red flag: >10% gap between demographic groups
Offer acceptance rate
What it is: % of offers accepted, by demographic group
Why it matters: Low rates signal compensation gaps or poor candidate experience
Red flag: >15% gap between demographic groups
Time-in-stage by demographic
What it is: Days per pipeline stage, broken down by group
Why it matters: Longer wait times for some groups indicate process inequity
Target: no statistically significant variance by group
A note on data collection: voluntary self-identification is the standard. Candidates should always have the option to decline. The data should be stored separately from evaluation records and should never be accessible to interviewers. The EEOC's EEO-1 reporting guidelines set the legal framework for what data can be collected and how it must be handled for employers with 100+ employees.
For smaller teams without 100+ annual hires, statistical significance is harder to achieve. In that case, focus on qualitative audits: review a sample of rejected candidates each quarter and ask whether the rejection rationale holds up against stated criteria. You are looking for patterns, not statistical proof.
Share these metrics with hiring managers quarterly. Most diversity hiring problems persist because data lives in HR systems that hiring managers never see. When a manager sees their pipeline drop-off data compared to other teams, the conversation about changing process becomes much easier. Your recruitment metrics framework should include diversity indicators alongside speed and quality metrics, not in a separate report.
Common Mistakes
What most diversity hiring programs get wrong
Treating unconscious bias training as the solution
Awareness training is not ineffective. It is just insufficient on its own. The research, including a well-cited 2019 meta-analysis in the Journal of Applied Psychology, shows that training changes awareness but rarely changes behavior without process changes to support it. Train your team and then change the process. Do not treat training as the change.
Setting diversity hiring targets without pipeline analysis
If your goal is to hire 30% women in technical roles this year and only 12% of your current applicants are women, you have a sourcing and JD problem. Targets are useful. Targets without a theory of how to achieve them create pressure without a path, which sometimes leads to teams bending their process in ways that create legal risk.
Focusing entirely on demographic diversity while ignoring cognitive diversity
Hiring for demographic representation is necessary but not equivalent to building cognitively diverse teams. You can have a demographically diverse team that thinks identically because they all went through the same elite education pipeline and rose through the same organizations. The goal is teams that challenge each other's assumptions. That requires diversity of experience, perspective, and approach, not just identity.
Not fixing pay equity before publishing salary ranges
Pay transparency is good. Publishing ranges that reflect existing internal pay gaps is worse than not publishing at all, because it locks inequity into future hires. Before you add salary ranges to job postings, run a pay equity analysis. Correct the gaps. Then publish the ranges. Doing this in the wrong order creates liability and damages trust with your team.
Frequently Asked Questions
What is diversity hiring and how is it different from affirmative action?
Diversity hiring refers to intentional practices that remove barriers in recruiting so that qualified candidates from underrepresented groups have equal access to opportunities. It is not the same as affirmative action, which involves specific numerical targets or preferences mandated by law or contract. Most diversity hiring practices are about fixing process problems: exclusionary job descriptions, narrow sourcing channels, unstructured interviews that favor familiar profiles. The goal is equitable access, not quota fulfillment.
Does diversity hiring mean lowering the bar?
No. The honest answer is that if your hiring bar is producing homogeneous results, the bar itself may be miscalibrated. Requiring an Ivy League degree for a role that does not need it is not a high bar. It is a proxy that filters based on socioeconomic background, not capability. Genuine diversity hiring means defining what actually predicts performance in a role, then assessing all candidates against that criteria consistently. That typically raises the quality of hiring, not lowers it.
How do I get more diverse candidates into my pipeline?
Start with sourcing channels. Post on diversity-focused job boards like PowerToFly and Jopwell. Build relationships with HBCUs, Hispanic-serving institutions, and professional affinity organizations. Audit your job descriptions for exclusionary language and degree requirements that lack justification. Employee referral programs often perpetuate homogeneity because people refer people like themselves. Balance referrals with proactive outreach to new networks.
What does an inclusive interview process look like?
Three things matter most. First, use structured interviews with the same questions for all candidates evaluated against pre-defined criteria. Second, build diverse interview panels so candidates see themselves reflected in your process. Third, debrief using scorecards before discussion to prevent the loudest voice in the room from dominating. Research from Google re:Work shows that structured interviewing reduces bias and improves prediction of job performance simultaneously.
How do we measure progress on diversity hiring?
Track representation at each pipeline stage, not just at the point of hire. Calculate pass-through rates by demographic group from application through offer. If diverse candidates apply at comparable rates but advance at lower rates, the problem is in your process, not your sourcing. At minimum, track: application rate, screen-to-interview pass rate, offer acceptance rate, and time-in-stage. Run the numbers quarterly and share them with hiring managers.
Can small companies with limited resources build diverse pipelines?
Yes, and often more effectively than large companies. Small teams do not have the inertia of legacy processes. The biggest levers cost nothing: rewrite your job descriptions to remove exclusionary language, add salary ranges, post on two or three diversity-focused job boards, and implement structured interview questions. Each of these is a half-day of work. The ROI compounds because better process attracts stronger candidates across all groups.
Resources & Further Reading
Related Guides
- How to Reduce Unconscious Bias in Hiring
Tactical steps to de-bias screening and evaluation
- Structured Interviews: The Complete Guide
The interview format that predicts performance and reduces bias
- How to Write Job Descriptions That Attract Better Candidates
Full guide to writing inclusive, effective JDs
- 15 Recruitment Metrics Every Hiring Team Should Track
Where diversity metrics fit into your broader analytics framework
External Sources
- Google re:Work: Unbiasing Research
Google's internal research on structured hiring and bias reduction
- McKinsey: Diversity Wins and How Inclusion Matters
Longitudinal research on diversity and financial performance
- EEOC: EEO-1 Data Collection Guidelines
Legal framework for workforce diversity data collection
- SHRM: Talent Acquisition Research
HR benchmarks on inclusive hiring practices
Build a more equitable hiring process
Prepzo gives you structured screening, consistent scorecards, and pipeline analytics so you can see where diverse candidates drop out and fix it. Free to start.
Try Prepzo free