Keyword Matching vs Evidence-Based Resume Screening

Two resumes can use the exact same phrase from a job ad and represent two completely different levels of capability. One candidate wrote it because the work is genuinely theirs. The other wrote it because they read the job ad closely before submitting. A system that checks only for the presence of the phrase cannot tell these two candidates apart, but a system that checks what stands behind the phrase can.

That is the difference between keyword matching and evidence-based screening. It sounds like a narrow difference in method. In practice, it produces different shortlists.

Key Takeaways

Keyword matching checks whether specific words or phrases appear on a resume. Evidence-based screening checks whether the resume demonstrates that a requirement has actually been met.

A 2025 study published in Academy of Management Proceedings found that resume assessments carried validity coefficients as low as 0.04 to 0.07, with no significant relationship to job performance or turnover.

A Harvard Business School and Accenture study found that 88% of employers acknowledged their screening systems exclude qualified candidates who do not use the exact language of the job description.

Evidence-based screening addresses this by breaking a role into discrete requirements before any resume is read, then scoring each candidate’s demonstrated experience against each requirement individually.

Talentranx is built on the evidence-based model, scoring candidates requirement by requirement rather than by word overlap.

What keyword matching actually tests

Keyword matching works by comparing the words on a resume against a list of terms drawn from the job description. If enough of those terms appear, the resume passes. If they do not, it gets filtered out or ranked low, often before anyone reads it.

What this measures is linguistic overlap: how closely a candidate’s vocabulary matches the vocabulary of the person who wrote the job ad. It is fast and cheap to run, and it produces the same result every time. It is not a test of whether the candidate can do the job. Those are two different questions, and keyword matching only answers the first one.

The method rewards candidates who happen to describe their experience in the same words the job ad uses, whether or not that experience runs deep. It penalises candidates who did the same work but described it differently, a common outcome for specialists who picked up the terminology of a previous employer, industry, or era.

What evidence-based screening tests instead

Evidence-based screening starts from a different premise, one covered in more depth in the article what evidence-based resume scoring actually means. Rather than searching for words, it works by breaking the job description into a list of discrete, assessable requirements. Each candidate is then measured against each requirement separately, using the specifics of what the resume actually describes: what the candidate did, at what scale, with what outcome, and whether that maps to the requirement fully, partially, or not at all.

The wording a candidate happens to use no longer determines the outcome. A candidate who describes leading containment and remediation across three teams during a ransomware event earns credit for incident response experience even if the phrase “incident response” never appears on the page. The requirement gets scored, not the phrasing.

Same resume, two different verdicts

Consider a cybersecurity analyst role requiring incident response experience. Candidate A’s resume states plainly: “Incident response, SIEM, threat hunting.” Candidate B’s resume reads: “Led containment and remediation for a ransomware incident, coordinating three internal teams against an eight-hour service level agreement, and rebuilt the detection rules that missed it the first time.”

A keyword match scores Candidate A higher. The required term is right there, stated in plain words. Candidate B’s resume does not contain the phrase “incident response” at all, so a strict keyword filter may not credit that requirement, or may rank the resume lower than one that states the term outright.

An evidence-based assessment reaches the opposite conclusion. Candidate A’s resume offers nothing beyond the listed term: no scale, no outcome to weigh it against. It could describe someone who watched a dashboard during a single drill. Candidate B’s resume describes a specific, high-pressure incident and the leadership role played inside it, with a concrete outcome attached. That is evidence the requirement was exercised under real conditions, not just claimed.

Both resumes are real, but only one demonstrates the requirement.

What the research says about each approach

Direct, controlled studies pitting keyword-matching software against evidence-based scoring software head to head, using identical resumes and identical hiring outcomes, do not appear to exist yet. That is a genuine gap in the research. What does exist is strong, converging evidence on each half of the comparison separately.

On the keyword side, a Harvard Business School and Accenture study surveying more than 2,250 employers across the US, UK, and Germany found that 88% acknowledged their automated screening systems exclude qualified, high-skilled candidates simply because those candidates did not use the exact language of the job description. The same study found that 49% of employers were filtering out candidates for lacking a specific degree in roles that did not functionally require one. The filtering mechanism enforced the presence of a word or a credential, regardless of the capability behind it, a pattern covered in more depth in the limitations of ATS keyword matching.

On the resume-reading side more broadly, a 2025 study published in Academy of Management Proceedings examined recruiter ratings of resumes against actual job performance and turnover across two independent samples: nurse applicants in the United States and sales applicants in China. Both samples converged on the same finding. Resume assessments carried validity coefficients between 0.04 and 0.07, with no statistically significant relationship to how those candidates went on to perform or how long they stayed. A rating built primarily from reading a resume, whatever method produced it, predicted almost nothing about the outcome it was meant to predict.

Structured, criterion-anchored assessment fares considerably better in the broader selection literature. Schmidt and Hunter’s 1998 meta-analysis of 85 years of selection research found that structured methods substantially outperform the loosely defined judgments that typically drive resume screening. Google’s own re:Work research reaches a related conclusion from a different angle: structure only produces valid comparisons when it is defined before candidates are seen and applied consistently to every one of them, which is the discipline evidence-based screening is built around, unlike a flat keyword list.

None of this proves that a specific evidence-based tool will outperform a specific keyword tool in every case, but it does establish that the mechanism behind evidence-based screening, structured, per-requirement, evidence-anchored assessment, sits on considerably firmer ground than keyword overlap.

Where the difference matters most

For high-volume, entry-level roles with a large talent pool and simple requirements, the gap between the two methods matters less. Vocabulary tends to be standardised and requirements are few, so the cost of missing an individual candidate stays low.

For specialist roles, the gap widens. Requirements are layered and vocabulary is fragmented across industries and career histories. On top of that, the talent pool is already narrow before any screening happens. A keyword filter applied to a role like this can do more than miss a few edge cases: it can systematically remove the candidates whose experience is genuinely deepest, simply because they described it in language the job ad did not anticipate.

How Talentranx applies this

When a job description is uploaded, Talentranx extracts and structures the underlying requirements rather than generating a list of terms to search for. Hiring managers can review and adjust this framework before scoring begins, so it reflects what the role actually needs rather than which words appeared most often in the ad.

Each candidate is then scored against every requirement individually, with the evidence drawn from the specifics of what the resume describes. The output shows the reasoning behind each rank: where a candidate’s experience fully meets a requirement, where it partially meets one, and where the evidence is missing altogether, regardless of whether the resume happens to use the exact terminology of the job ad. The mechanics of running this comparison, requirement by requirement, are covered in how to compare resumes against job requirements.

Summing up

Keyword matching answers a narrow question: whether this resume used the right words. Evidence-based screening answers a different one: whether this candidate has actually done what the role requires. The two questions sound close enough to substitute for each other. The research on each suggests they are not, and the gap between them tends to be widest for exactly the roles where getting the shortlist right matters most.

Sources

Roth, P. L., Andrekovich, M. S., & Harrison, J. T. (2025). Resumes for Selection: Ubiquitous in Use but Little Evidence of Criterion Validity. Academy of Management Proceedings. https://journals.aom.org/doi/abs/10.5465/AMPROC.2025.10899abstract
Fuller, J. B. & Raman, M. (2021). Hidden Workers: Untapped Talent. Harvard Business School and Accenture. https://www.hbs.edu/managing-the-future-of-work/research/hidden-workers-untapped-talent
Schmidt, F. L. & Hunter, J. E. (1998). The validity and utility of selection methods in personnel psychology: Practical and theoretical implications of 85 years of research findings. Psychological Bulletin, 124(2), 262–274.
Google re:Work. A guide to structured interviewing for better hiring practices. https://rework.withgoogle.com/intl/en/guides/a-guide-to-structured-interviewing-for-better-hiring-practices