A screening team can show that applicants with stronger resume-fit ratings are more likely to move forward. That still does not tell you whether those people perform better in interviews.

The reason is simple. Interview evidence exists only for the applicants who reach an interview. Most rejected applicants never produce a structured interview score, so the team cannot see how many of them would have cleared the same bar.

Metaview's application data shows a clear relationship between the fit rating assigned during screening and the recruiter's decision at that same stage. Among applications with a recorded recruiter decision, 17.2% of those rated great advanced, compared with 12.1% rated good, 7.8% rated okay, and 5.6% rated poor.¹ That is useful evidence about how screening decisions are made. It is not evidence about interview performance.

So the honest answer to the title is that the current data cannot settle it. It can show whether a screening system and a recruiter tend to agree. It can also help a team find where its criteria are vague or applied inconsistently. To learn whether the screen is rejecting people who would interview well, the team has to collect some of the missing interview evidence deliberately.

Why rejected candidates vanish from your data

The blind spot begins at the screening decision. Applicants who move forward can generate interview notes, scorecards, work-sample results, and later-stage decisions. Applicants who are rejected usually generate none of those records inside that hiring process.

This is a form of selection bias, often described as range restriction in assessment research. The company observes later-stage performance only among the group that passed the earlier filter. It can study differences within that selected group, but it cannot assume the same pattern would hold among the people who were screened out.

The Metaview application figures illustrate the distinction. Among applications with a recorded recruiter decision, the advancement rate rose with the Application Review fit bucket:

That gradient shows that higher fit ratings and recruiter decisions move in the same direction. It does not show that great-fit applicants scored higher in interviews, received more offers, or became stronger hires. The rating and the decision were both produced during screening, and the dataset cited here does not provide a standardized interview outcome for every applicant.

It also does not reveal the false-negative rate. A poor-fit applicant who was rejected has no later-stage result in the process, so the team cannot tell whether the screen was right, whether the criteria were too narrow, or whether the resume simply failed to show evidence the candidate could have supplied in conversation.

What a resume can and cannot tell you

A resume can provide useful, job-related evidence. It can show where someone worked, the scope they claim to have owned, the technologies or markets they have encountered, and the broad direction of their career. When the role criteria are clear, that is enough to decide who deserves closer review.

But “strongest resume” is not an objective category. A resume can look strong because the candidate has a familiar title, a recognizable employer, or a polished description of their work. None of those automatically establishes fit for the role in front of you.

Even relevant claims need testing. “Managed a team of 12” does not explain what changed under that person's leadership. “Grew revenue by 30%” does not show the baseline, the market conditions, or the candidate's individual contribution. “Owned the migration” may describe technical leadership, project coordination, or simply membership in the team.

A structured interview or work sample can ask for the missing detail. It can test how the candidate reasons, what they personally did, how they handled constraints, and whether the example meets the role's competency standard. Official U.S. Office of Personnel Management guidance defines structured interviews around job-related competencies, predetermined questions, and consistent rating standards for all candidates.²

That does not make the interview a perfect ground truth. Interview ratings can also be noisy or biased. The point is narrower: a resume and an interview observe different evidence. A screening audit should not treat the resume as proof of future interview performance, or the interview as proof of future job performance.

How to test a screen that hides its own misses

A useful audit separates two questions. The first is whether reviewers applied the written screening criteria consistently. The second is whether those criteria select people who clear the next assessment stage. The first can be tested from applications alone. The second requires new evidence.

1. Test consistency before testing prediction

Choose one role and pull a random sample of advanced and rejected applications. Give the same materials to a second reviewer without showing the original recruiter decision or fit bucket. Ask them to apply the profile that was in force when the decision was made.

Record where the reviewers disagree and why. Repeated disagreement on the same criterion usually points to one of three problems: the requirement is vague, the evidence standard is unclear, or the criterion is being used as a proxy for something the role does not actually require.

This exercise measures repeatability. A high agreement rate does not prove the screen is valid. Two reviewers can apply the same bad rule consistently.

2. Compare screening predictions inside the advanced group

For applicants who already advanced, compare the screening fit bucket with a clearly defined first-stage result. Use the same structured scorecard or work-sample standard across the role. Look for recurring mismatches, such as applicants rated great on paper who repeatedly fall short on one competency, or applicants rated okay who consistently clear the interview bar.

This can show whether the screen is useful within the group it selected. It still cannot estimate how rejected applicants would have performed, because they remain outside the observed range.

3. Collect evidence from below the cutoff

To investigate false negatives, plan a prospective audit before screening begins. Select a small, predefined sample of applicants who fall below the normal cutoff but still meet the role's hard eligibility requirements. Give them genuine consideration for the open role through the same first-stage assessment used for other candidates.

Set the rules before anyone sees the results:

• which fit buckets or score range can enter the audit sample;
• how candidates will be sampled;
• the structured questions, work sample, or scorecard used at the next stage;
• the result that counts as clearing the stage;
• whether interviewers can be kept unaware of the original fit rating;
• how the team will handle candidates who pass;
• which legal, privacy, accessibility, and record-keeping requirements apply in the markets where the company hires.

The candidates must be part of a real hiring process, not invited into an interview the company has no intention of honoring. If someone clears the same bar as the advanced group, they should receive the same consideration for progression.

A random sample across all rejected applications estimates misses across the full rejected pool but can consume a great deal of interview time. A sample concentrated near the cutoff is cheaper and more likely to find borderline errors, but it answers a narrower question. State which question you are testing before choosing the sample.

Candidate participation creates another limitation. Some rejected applicants will decline or fail to respond, and those who accept may differ from those who do not. Report the invitation and participation rates rather than treating the interviewed sample as representative of every rejection.

4. Change criteria, not just thresholds

The useful output is not one grand “screen accuracy” number. It is a list of specific criteria that need attention.

• A requirement that reviewers interpret differently should be rewritten with clearer evidence.
• A criterion that rejects candidates who later clear the structured assessment may be too strict, too broad, or based on a weak proxy.
• A high-fit signal that repeatedly fails at the next stage may be overweighted or easy to overstate on a resume.
• A hard constraint that is genuinely required should remain a hard constraint, even if otherwise strong applicants fail it.

Repeat the review within a single role or closely related role family. Pooling unrelated jobs can create a clean-looking average that hides completely different criteria and assessment bars.

Where Metaview fits

Metaview Application Review provides the record a screening audit needs, but it does not validate the screen on its own.

The product pulls a role from the connected ATS and uses the job description and any supporting documents to draft an Ideal Candidate Profile. The recruiting team reviews and edits that profile. Application Review then evaluates applicants against it, assigns match ratings, and shows written reasoning for each evaluation. When the profile changes, candidates can be re-evaluated against the updated criteria.³

That makes several parts of the process easier to inspect:

• the criteria the system used;
• the fit rating assigned to each application;
• the reasoning and evidence behind the rating;
• the recruiter's progress or reject action;
• how the pool changes when the profile is refined.

The human boundary is explicit. Recruiters decide who progresses and who is rejected. Metaview never auto-rejects a candidate.

Application Review cannot tell you how a rejected applicant would have performed in an interview, whether a screening criterion is legally appropriate, or whether a later hire succeeded on the job. It can give the team a visible, editable record of the screening logic so those questions can be tested rather than reconstructed from memory.

Sources

¹ Metaview canonical aggregate data, APPLICATIONS-8: advancement among applications with a recorded recruiter decision by Application Review fit bucket. Great 17.2%, good 12.1%, okay 7.8%, poor 5.6%. July 22, 2026 data pack.

² U.S. Office of Personnel Management, Structured Interviews, accessed July 31, 2026.

³ Metaview Help Center, Application Review Overview, accessed July 31, 2026.