A talent acquisition manager at a mid-sized tech company spent the better part of Q1 rebuilding her sourcing strategy. She added HBCU job boards, refreshed the job description language, tapped into two professional associations she had not worked with before. By March, her pipeline looked different than it had in years. Qualified diverse candidates were applying. Her hire rate for those candidates was flat.

The problem was not where she was looking. The problem was what happened after candidates arrived.

Sourcing Is Not Where Diverse Candidates Disappear

Most DEI hiring conversations start at the top of the funnel: where are we posting jobs, who are we reaching, are our descriptions inclusive. These are real questions, and answering them well matters. But the data increasingly points to a different bottleneck.

A 2024 study from economists at UC Berkeley and the University of Chicago analyzed 83,000 fictitious applications sent to 11,000 entry-level positions at more than 100 Fortune 500 companies. The average employer called back white applicants roughly 9% more than Black applicants with equivalent credentials. At the worst-performing companies, that gap reached 24%. This disparity appeared before any conversation happened, before any hiring manager ever saw the name.

The resume screen is where many diverse candidates exit the process. Not because a recruiter made a deliberate call. Because resume review, done without consistent criteria, tends to replicate whatever pattern the reviewer already has.

The Screening Stage Has a Criteria Problem

Here is what unstructured screening looks like at scale. A recruiter opens a stack of 60 applicants for a req that needs a technical project manager. She has a job description, a salary band, and a vague sense of what good looks like from the last person who held the role. She applies judgment to each resume: does this person seem like a fit? Has she been given a rubric? Usually not. Has she aligned with the hiring manager on the actual criteria before reviewing a single application? Often not.

The result is that two equally qualified candidates can receive different outcomes based on factors unrelated to whether they can do the job.

The same problem follows candidates into the phone screen. When screening conversations are unstructured, every recruiter asks different things. One recruiter probes on technical depth. Another focuses on communication style. A third spends the back half of the call on career trajectory. None of them score against the same rubric, which means the hiring manager receives a slate that reflects who interviewed well with which person on which day, not who actually meets the role's requirements.

The U.S. Office of Personnel Management has documented this pattern for decades. Unstructured interviews show low to moderate levels of validity in predicting job performance, while structured interviews demonstrate a high degree of reliability, validity, and legal defensibility. That finding dates to 2008, but it has been replicated consistently in the research that followed. The underlying pattern has not changed: unstructured evaluation produces inconsistent outcomes, and inconsistency hits some candidates harder than others.

Criteria Drift Is the Mechanism

Even when a team agrees on what they are looking for, criteria shift from req to req and from interviewer to interviewer. "Strong communication skills" means something different to a hiring manager who values concise written updates than to one who values presence in a room. "Leadership potential" can mean that the candidate reminded someone of themselves at an earlier stage in their career.

None of this is necessarily deliberate. It is what happens when evaluation is left to individual judgment without a shared framework. And the candidates who absorb the cost of inconsistent criteria are disproportionately those already navigating a system that was not built with them in mind.

This is the mechanism. Sourcing brings diverse candidates into the funnel. Inconsistent screening criteria filter them out before the hiring manager has a chance to decide. The sourcing investment does not compound. It leaks.

What Structured Screening Changes

Structured screening sets criteria for a role before the first candidate is evaluated. Every candidate answers the same questions, scored against the same dimensions. The recruiter collects structured signal: not "does this person seem good?" but "on this specific dimension, does this candidate meet the defined threshold?"

This is what Eximius's screening agent, Sia, does. Sia conducts screening conversations across SMS,  and email, using criteria set for the specific req. Every candidate in the slate is screened against what the job demands and if the candidate brings that to the table. The recruiter sees responses organized by the criteria that matter for the role, not by who happened to have a strong 20-minute call on a Tuesday afternoon.

The recruiter still decides who advances. The hiring manager still runs the panel. The offer decision stays with the team. What changes is the quality of the input: a slate ranked on consistent, job-relevant criteria instead of a stack ordered by first impression.

This Is Not the Whole Answer

Structured screening does not guarantee more diverse hires. Bias is not something any single tool resolves. It is structural, embedded in how job requirements get written, in who sits in the debrief, in which candidates get the offer pushed through approvals quickly and which ones move through a slower process.

What structured screening does is remove one of the most common points where inconsistency enters: the evaluation stage, where criteria that were never made explicit get applied differently to different candidates.

The team still has to do the work on criteria design: what does "strong communication" actually mean for this role, what level of experience is genuinely required versus preferred, what does good look like before the first resume is reviewed. Sia runs the structured conversation. The recruiter sets what it measures. The judgment about what matters for the role stays where it belongs.

Where to Look If Your Numbers Are Flat

If your sourcing is working and your hire numbers are not moving, the first question is not "where do we find more candidates?" It is "at which stage are they exiting, and why?"

Pull your pipeline data by stage. Where does the funnel thin? Is it at resume review, at phone screen, at panel, at offer? If diverse candidates are entering the funnel but not reaching the hiring manager's slate, the sourcing investment is not the problem. The screening process is.

That is a different intervention. It requires criteria set before review begins, consistent questions applied to every candidate, and scoring that separates "did this candidate meet the threshold on this dimension" from "did this candidate feel like a fit." Recruiters deserve a framework that lets their judgment show up in the actual decision, not scattered across the criteria that determine who gets to the decision in the first place.

Diverse sourcing is necessary. It is not sufficient. The screening stage is where a diversity effort either compounds or collapses.

Want to see what structured screening looks like on your req volume? Book a pilot and we'll run your next role through the Eximius workflow.