How to Improve Quality of Hire: 7 Evidence-Backed Ways to Hire Better (and Prove It)

TL;DR

• Quality of hire is the value a new hire delivers through job performance and tenure. You improve it by changing who you select and how – not by tracking the metric harder.
• The methods most teams lean on are among the weakest predictors of performance. Years of experience (ρ = .07) and unstructured interviews (ρ = .19) sit near the bottom of the latest meta-analysis. Structured interviews sit at the top (ρ = .42), with smaller differences between groups (Sackett et al., 2022).
• The seven ways, in order: define quality before you hire; replace the CV screen with a structured interview; assess experience instead of counting it; add job-specific tests where the role needs them; structure every later interview; protect the signal from AI-written answers and fraud; close the loop with post-hire data.
• Where to start: fix the front-of-funnel screen first, structure every interview after it, then connect hiring data to 6- and 12-month outcomes so the process keeps learning.

Quality of hire is the metric everyone wants and few can move. In an Aptitude Research survey sponsored by Crosschq, 75% of HR and talent leaders called it a leading priority, but only 38% said they consistently deliver quality hires (Aptitude Research, 2024 – vendor-sponsored). And only 20% of organisations track quality of hire at all, down from 27% in 2022 (SHRM 2025 Recruiting Benchmarking).

The pressure is growing. Application volumes keep rising, and 39% of candidates now use AI somewhere in their applications (Gartner, 2025). The CV was never a strong signal, and it gets noisier every quarter.

Most advice on how to improve quality of hire is a list of 15 or 31 tips treated as equals: better job ads, referrals, “culture fit”, reference checks, more interviews. This guide ranks seven ways by what research shows actually predicts performance and retention, names the tools that support each, and ends with a 90-day plan.

A disclosure: Sapia.ai makes one of the tools in this guide. We’ve run structured interviews for 10M+ candidates (vendor-reported) and publish an independent bias audit of our own scoring, so we’ve tried to hold this guide to the evidence bar it asks you to apply.

How we decided which strategies to improve quality of hire to cover in this article

We judged each strategy against five criteria:

1. Predictive validity – how well the method predicts job performance. We use Sackett, Zhang, Berry & Lievens (2022), the most current large-scale meta-analysis, rather than the older Schmidt & Hunter (1998) estimates most articles still quote.
2. Effect on retention – whether the method plausibly reduces early attrition, and whether there’s evidence for it.
3. Fairness – the size of the group differences the method tends to produce, and whether the tools behind it have an independent, published, role-level bias audit.
4. Candidate experience and integrity – completion, accessibility and resistance to AI-assisted answers.
5. Effort and time-to-value – what it takes to put in place this quarter.

Some popular tips didn’t make the list because the evidence is thin or the bias risk is high. Filtering on years of experience predicts almost nothing (more on that below). Using GPA as a general filter predicts job performance only modestly (ρ = .21) and turnover not at all (−.02), per Van Iddekinge et al. (2024). Adding more unstructured interview rounds adds cost, not signal. And unstructured “culture fit” judgements invite bias. Realistic job previews do make the cut, inside the first way – their effect on turnover is real but small, so treat them as cheap hygiene rather than a main lever.

7 ways to improve quality of hire, ranked by evidence

Validity and group-difference figures: Sackett et al. (2022, 2023), operational validity; experience figures also from Van Iddekinge et al. (2019). Reference-check figures are older estimates – see way 6.

1. Define quality for the role before you hire

What to do: before the job ad goes out, agree with the hiring manager what a quality hire looks like at 6 and 12 months – the four or five competencies that drive performance, the outcomes that matter and the experience level. Agree how you’ll report quality of hire too. SHRM Talent 2026 speakers defined it as the immediate and long-term value a new hire delivers through job performance and tenure (SHRM, 2026). A practical version uses three measures: performance rating, 12-month retention and hiring-manager satisfaction.

The evidence: every method downstream depends on this step. Structured interviews and job-knowledge tests predict performance because their questions come from an analysis of the job – “base questions on a job analysis” is the first of Campion, Palmer & Campion’s (1997) 15 components of interview structure.

How to implement it:

• Run a 45-minute intake with the hiring manager using a structured template.
• Cap the success profile at five competencies – more dilutes the signal.
• Set the experience level explicitly, so you know whether experience needs assessing at all.
• Add a realistic job preview. Its effect on voluntary turnover is small (r ≈ −.07, Earnest, Allen & Landis, 2011), but it’s cheap.

Where it falls short: hiring managers often describe the last person in the role rather than the job, so push for outcomes, not traits. And “culture fit” creeps in here – replace it with named, observable values, scored like any other competency.

How you’ll know it’s working: hiring managers rate shortlists more highly, and fewer requisitions need reworking.

Tools that help: Sapia’s Job Analysis Studio (JAS) turns a job description into a competency-mapped structured interview; SHL offers job profiling. See the tools section.

2. Replace the CV screen with a structured interview at the front of the funnel

This is the biggest single lever in this guide.

What to do: stop using the CV – years, titles, employers, schools – to decide who progresses. Give every applicant the same job-related structured interview, scored against the same anchored criteria. Keep the CV for eligibility checks only, such as right to work or licences.

The evidence: structured interviews have the highest operational validity in Sackett et al. (2022), at ρ = .42, with a Black–White difference of d = .23 – far smaller than cognitive ability tests (d = .79). The parts of a CV screen sit near the bottom: job experience at ρ = .07 and, in older estimates, points-based scoring of training and experience at .11 and years of education at .10 (Schmidt & Hunter, 1998). A review of 12 meta-analyses found structured interviews “much more reliable and valid” than unstructured ones (Levashina et al., 2014).

The CV screen is still the default because it’s fast, not because it predicts well. More in our guide to structured vs unstructured interviews.

How to implement it:

1. Write five or six questions from the success profile – behavioural (“tell me about a time…”) and situational (“what would you do if…”).
2. Write an anchored rating scale for each, describing weak, solid and strong answers.
3. Put the interview at the point of application, so every applicant gets a fair hearing – not just the CVs a recruiter had time to skim.
4. Pilot it on one high-volume role family and compare hiring-manager satisfaction and 90-day attrition with the old process.

Where it falls short: validity varies – the credibility interval is wide (roughly .18 to .66), so question quality and local validation matter. It adds a step for candidates, so untimed, mobile and short beats long and timed. And human-scored interviews don’t scale to thousands of applicants, which is why AI-scored versions exist – and those need their own evidence (see the tools section).

How you’ll know it’s working: the shortlist-to-offer ratio improves, you need fewer interviews per hire, and early attrition in the pilot group falls against baseline.

Tools that help: Sapia (text-first, CV-blind chat interview), HireVue (video and voice), SHL Smart Interview On Demand, Greenhouse Voice AI and BrightHire Screen.

In practice: Holland & Barrett reported an 89% reduction in early churn after moving to Sapia’s structured interview (customer-reported, before and after – not a controlled study).

3. Assess experience, don’t count it

What to do: for roles where experience really matters, ask about it directly and score the answer, rather than filtering on years in role or the employers on a CV.

The evidence: across 81 samples, Van Iddekinge, Arnold, Frieder & Roth (2019) found pre-hire experience correlates with job performance at ρ = .06 and with turnover at ρ = .00 (Personnel Psychology). Even experience in closely related jobs reached only ρ = .07, and the authors caution against selecting on such measures. What matters is the nature of someone’s experience – how their responsibility grew, what they delivered, how relevant it is – and a year count captures none of that.

How to implement it: add one or two structured experience questions, such as “What’s the most complex work you’ve been responsible for?” Score answers on depth and progression, quality and scale of outcomes, and relevance to the role, with anchors calibrated to the role’s level. Score experience separately from competencies so the two signals stay clean.

Where it falls short: answers are self-reported, like a CV – though structured questions make thin evidence easier to spot. It doesn’t test hard skills, so pair it with way 4 where needed. And any new scoring method needs its own local validation.

How you’ll know it’s working: fewer “great scores, but have they actually done it?” objections in shortlist meetings, and steady or better 6-month performance for experienced hires.

Tools that help: within JAS, Sapia’s Experience Scoring [confirm product name with Sapia] adds up to two structured experience questions and scores them on depth, outcomes and relevance – without using the CV as the assessment. It’s being validated in customer pilots now; it isn’t a hard-skills test or a lie detector, and people still make the decisions.

4. Add validated, job-specific tests where the role demands them

What to do: for roles with a real knowledge or technical core – engineering, finance, licensed trades – add a job-knowledge test or work sample after the structured interview. For roles without one, don’t add tests for the sake of it.

The evidence: job-knowledge tests reach ρ = .40, empirically keyed biodata .38, work samples .33 and cognitive ability tests .31 (Sackett et al., 2022). The fairness trade-off is real: cognitive tests carry the largest group difference (d = .79), work samples .67 and job-knowledge tests .54. Used as hard cut-offs early in the funnel, they can reduce diversity.

There’s an active debate here. Some researchers argue Sackett’s corrections are too conservative for cognitive ability (Oh, Le & Roth, 2023). But the finding that structured, job-specific methods rank at or near the top isn’t in dispute.

How to implement it: use tests later in the funnel and as one input among several, not a knockout. Check the vendor’s technical manual for criterion validity in roles like yours. And monitor pass rates by group from day one.

Where it falls short: adverse impact risk, especially with cognitive tests. And long test batteries cost you completions – long assessments are one of the most common complaints in user reviews of test platforms.

How you’ll know it’s working: time-to-productivity improves in the tested role family, and pass-rate ratios by group stay within your monitoring thresholds.

Tools that help: SHL (the deepest published psychometric documentation), Criteria, TestGorilla and Harver. For developer roles, look at specialist coding platforms such as HackerRank, Codility or CodeSignal.

5. Structure and calibrate every later interview

What to do: apply the same discipline to hiring-manager and panel interviews: the same questions for every candidate in a role, anchored rating scales, independent ratings submitted before any debrief, and interviewer training.

The evidence: unstructured interviews sit at ρ = .19, against .42 for structured ones, with a larger group difference (d = .32 vs .23). Anchored rating scales alone lifted validity from .26 to .35 in one study (Taylor & Small, 2002). Structure also dampens the effect of impression management on ratings (Barrick et al., 2009).

The practice gap is wide. In the studies Levashina et al. (2014) reviewed, the average interview used only 5.7 of Campion’s 15 components of structure, and only about half included interviewer training.

How to implement it:

• Build interview kits in your ATS, one per role.
• Write a rubric with example answers at each level – poor, borderline, solid, outstanding – as in Google’s re:Work guide.
• Lock each interviewer’s ratings before the debrief starts.
• Run a quarterly calibration session using anonymised scorecards.

Where it falls short: interviewers resist scripts, so explain that follow-up probes are fine within the structure. Structure without training decays fast, so plan refreshers. And interview-recording tools raise consent and privacy questions in some jurisdictions.

How you’ll know it’s working: interviewers agree with each other more often, and fewer candidates get overturned in the debrief on “gut feel”.

Tools that help: Greenhouse (interview kits and scorecards), BrightHire (interview intelligence and coaching) and SHL Smart Interview Professional. Metaview is popular for AI interview notes; note that its Application Review feature ranks applicants from CVs and applications, so it isn’t a structured assessment.

6. Protect the signal: integrity in the AI era, and references as a late check

What to do: make sure the answers candidates submit are their own, and that the person you interview is the person you hire. Use structured reference checks as a late-stage verification, not a way to rank candidates.

The evidence: in Gartner’s 2025 candidate surveys, 39% of candidates used AI in their applications – 29% of those for assessment answers – and 6% admitted to interview fraud, such as someone else sitting the interview (Gartner, 2025). Gartner predicts one in four candidate profiles will be fake by 2028.

Reference checks are weaker than their popularity suggests. Older estimates put their validity at .26 (Schmidt & Hunter, 1998) to .29 (Aamodt & Williams, 2005, a conference paper), and Sackett et al. (2022) found too little data to estimate it. Leniency is severe: in one dataset, 60% of references rated the candidate “excellent”. Structured, confidential multi-rater tools do better than a phone call (Hedricks, Robie & Oswald, 2013).

How to implement it:

• Ask every assessment vendor how it detects AI-written answers, for its published detection and false-positive rates, and for the false-positive rate among non-native English writers.
• Verify identity before final interviews for remote roles.
• Move reference checks to the final stage, in a structured digital format.

Where it falls short: AI-text detectors over-flag non-native English writers (Liang et al., 2023), so a flag should prompt human review, never automatic rejection. Heavy proctoring hurts candidate experience and accessibility. And most vendors publish no detection figures at all – treat that as an answer in itself.

How you’ll know it’s working: every flag gets a human review, and fewer early exits trace back to misrepresented skills – a factor in 33% of bad hires in a 2017 survey (CareerBuilder/Harris Poll).

Tools that help: Sapia (built-in detection of AI-generated answers, with vendor-reported figures), Criteria AI Proctoring, TestGorilla’s anti-cheating controls and Crosschq (fraud-resistant 360 references with ID.me verification).

7. Close the loop: connect hiring data to performance, retention and fairness

What to do: join pre-hire scores to 6- and 12-month performance ratings, retention and hiring-manager satisfaction. Use the results to adjust questions, weights and cut scores. And check adverse impact for each role, not just pooled across the whole company.

The evidence: with only 20% of organisations tracking quality of hire (SHRM, 2025), most hiring processes have no way to learn. SHRM Talent 2026 speakers called post-hire performance data an “underutilized goldmine”.

On fairness, a peer-reviewed 2026 study makes the case for role-level checks. Analysing around 4.1 million applications screened by pymetrics between 2018 and 2022, Bommasani and colleagues found racial disparities at the level of individual positions that pooled audits hadn’t shown (Stanford Report, 2026). That applies to every vendor, Sapia included: ask for adverse-impact results by role.

How to implement it:

• Agree a three-measure quality-of-hire index and keep it consistent between cohorts.
• Export assessment scores to your HRIS or BI tool, so pre- and post-hire data sit together.
• Run a 12-month check for each role family, and review selection rates by group for each role at each stage.

Where it falls short: performance ratings are noisy and lenient, so small samples mislead. You only see outcomes for people you hired. And the plumbing between ATS, assessment platform and HRIS is often the real blocker.

How you’ll know it’s working: you can say which interview competencies predict 12-month performance in your roles – and you act on it.

Tools that help: Crosschq’s Quality of Hire Analytics is the only dedicated quality-of-hire module among the tools we profile; a people-analytics platform such as Visier or your own BI stack can do the same job. Sapia’s Discover Insights covers hiring-funnel analytics, and the Talent Intelligence Assistant (TIA) reuses interview evidence to surface silver-medallists and draft 30-day onboarding plans.

Tools that help you improve quality of hire

Tools don’t improve quality of hire on their own – they make ways 2 to 7 practical at scale. What matters is what each one asks the candidate to do, what the score is built from and what evidence the vendor publishes.

Facts as of September 2026. Several vendors launched, acquired or re-priced products in the last 18 months – check each vendor’s current site before you buy.

Structured interview at the front of the funnel (way 2)

This is the lane where the biggest lever lives. Most vendors here now have candidates complete an interview, so the useful differences are what the score is built from and what audit evidence is published.

Sapia.ai

Best for: replacing the CV screen with a low-friction structured interview for every applicant – frontline and graduate volume, and increasingly experienced roles – on top of your existing ATS.

What it does: applicants complete an untimed, mobile, text-first structured chat interview built from a job analysis in JAS. Answers are scored against competency rubrics, with explanations written back to the ATS. Async video is available as a second stage, Experience Scoring [confirm product name with Sapia] adds experience questions where needed, and TIA answers questions across everyone interviewed, traceable to what they said.

Evidence profile: Candidate produces: written answers to structured questions. Score built from: answers only – not CV, school or employers. Published validity: published research, less extensive than SHL’s manuals. Independent audit: BLDS, late 2022 – 72 tests found no evidence of practically significant disparate impact; US and Canada models only (FAIR framework). Integrity controls: AI-generated-answer detection; Sapia reports 98% detection and a 1% false-positive rate (vendor-reported).

Strengths: CV-blind scoring removes the weakest predictors from the first cut; a published audit with a named auditor; no camera or voice – Woolworths saw 82.6% completion (customer-reported); detection figures published, not just claimed.

Where it falls short: thinner validity documentation than SHL; an audit from late 2022 covering US and Canada only, so ask for newer, role-level results; not a hard-skills test, ATS or analytics suite; and since AI-text detectors can over-flag non-native English writers, ask us for that group’s false-positive rate.

Pricing: quote-based.

Verdict: the lead option for way 2 when you want CV-blind scoring with a published audit. SHL leads on validity documentation, and you’ll still need an ATS and an outcomes tracker.

HireVue

Best for: large enterprises running very high-volume and campus hiring, including those that need FedRAMP.

What it does: video interviews, game-based and competency assessments, chat and scheduling. A two-way voice AI Interviewer launched in June 2026, after HireVue acquired Hireguide’s technology in March.

Evidence profile: Candidate produces: video or voice answers, assessment and game responses. Score built from: answer content – the AI “relies only on what is said by the candidate” (2024 explainability statement); CV exclusion not stated. Published validity: construct evidence, including r = .69 agreement with expert raters (vendor-reported); criterion studies mostly private. Independent audit: DCI Consulting, 2023 and 2025, pooled; none for the voice AI Interviewer. Integrity controls: fraud mitigation claimed; no figures.

Strengths: the widest range of formats on one platform; an early public explainability statement; scale – 1,150+ customers (vendor-reported).

Where it falls short: no audit for the voice product, and pooled rather than role-level audits; one-way video and voice add friction, and reviewers cite login and audio glitches.

Pricing: quote-based.

Verdict: the breadth leader for video, voice and assessments. Sapia fits better where an untimed text format and an audit of the interview product matter most.

Validated tests and assessments (way 4)

SHL

Best for: global enterprises that want psychometric science that stands up to scrutiny, plus structured interview support, from one supplier.

What it does: OPQ32 personality, Verify ability, situational judgement and skills tests. Smart Interview On Demand scores recorded answers with AI; Smart Interview Professional (February 2025) structures live interviews that humans score.

Evidence profile: Candidate produces: test responses; recorded spoken answers. Score built from: test scores; On Demand scores answer content against behaviourally anchored scales. Published validity: the deepest public documentation here for its tests, with construct and criterion evidence in technical manuals; none for AI interview scoring. Independent audit: none published for Smart Interview. Integrity controls: nothing published for On Demand.

Strengths: unmatched psychometric documentation – a real lead over Sapia; coverage from application to final interview; structured live interviews without recording.

Where it falls short: AI interview scoring is far less transparent than its tests; few public user reviews; cognitive tests carry the largest group differences.

Pricing: quote-based; SHL Online offers self-serve access for smaller teams.

Verdict: the pick when validity documentation is the deciding factor. Sapia wins on an untimed, CV-blind interview with a published audit.

Criteria

Best for: mid-market teams that want validated cognitive testing and interviews in one contract.

What it does: CCAT cognitive test, personality, emotional intelligence and skills tests; async and live video interviews; an AI Interview Agent launched in June 2026; and AI Proctoring.

Evidence profile: Candidate produces: test responses and video answers. Score built from: test scores; interviews get AI-supported scoring for human review. Published validity: published CCAT figures, including criterion validity of r = .33 for sales staff (n = 745, vendor-published). Independent audit: third-party audits claimed, but no auditor or year named. Integrity controls: AI Proctoring flags AI-generated answers, second devices and tab switching; no accuracy figures.

Strengths: publishes real criterion numbers; unusually explicit proctoring detail; tests and interviews under one roof.

Where it falls short: no identifiable independent audit; reviewers cite steep renewal increases and limited report customisation; cognitive-heavy batteries carry adverse-impact risk.

Pricing: quote-based.

Verdict: a solid validated-testing choice for way 4 – not a CV-blind front-of-funnel interview.

TestGorilla

Best for: SMB and mid-market teams that want self-serve skills testing on a transparent budget.

What it does: 350+ skills tests, job simulations, coding challenges, video questions and AI interviews, plus CV scoring and ID verification.

Evidence profile: Candidate produces: test answers, work samples, code, video. Score built from: test scores; CV scoring is a separate module, so check which inputs are switched on. Published validity: per-test fact sheets; quality varies by test. Independent audit: none published. Integrity controls: webcam snapshots, copy-paste blocking, question pools; no detection figures.

Strengths: transparent pricing; a broad test library; built-in anti-cheating controls.

Where it falls short: no independent audit; long assessments drive candidate drop-off; limited customisation of prebuilt tests.

Pricing: Free tier; Core $142/month billed annually; Plus from $400/month; Enterprise custom – on a credit model revised in January 2026 (pricing, September 2026).

Verdict: the budget-friendly way to add job-specific tests – not the choice when you need enterprise-grade fairness evidence.

Harver (including pymetrics)

Best for: high-volume frontline hiring in retail and contact centres.

What it does: personality, cognitive, situational judgement, simulation and language tests; pymetrics behavioural games; video interviews; and Checkster reference checks. Harver acquired Symphony Talent in September 2026.

Evidence profile: Candidate produces: test and game responses, video, references. Score built from: test and game scores; role models aren’t published. Published validity: validation approach described; no studies linked. Independent audit: BABL AI audits (2025) pooled across employers and roles; the independent 2026 study above found position-level disparities pooled audits hadn’t shown. Integrity controls: nothing published.

Strengths: a wide range of assessments; volume workflow built in; assessments and references from one vendor.

Where it falls short: pooled-only audits; reviewers cite limited self-service and integration friction.

Pricing: quote-based.

Verdict: a capable high-volume assessment engine. Ask for role-level adverse-impact results before you buy – the same question to ask every vendor, including us.

Interview structure and calibration (way 5)

Greenhouse

Best for: mid-market and enterprise teams standardising structured interviews inside their ATS.

What it does: an ATS with interview kits and scorecards, AI notetaking, fraud detection and a BI connector. Greenhouse Voice AI, from the Ezra acquisition completed in May 2026, runs two-way voice interviews on “the questions and scoring criteria your team defines”.

Evidence profile: Candidate produces: human scorecard ratings; voice answers for Voice AI. Score built from: human ratings; Voice AI scores answers against your rubric. Published validity: none published for Voice AI. Independent audit: Warden AI, confirmed for Talent Matching; monthly Voice AI audits are claimed, but we couldn’t verify published results. Integrity controls: fraud detection; no figures.

Strengths: the reference standard for structured scorecards; a large integration ecosystem; public AI governance, including ISO 42001.

Where it falls short: quality-of-hire analysis needs manual work or BI; features gated by tier; Voice AI’s evidence is new and thin.

Pricing: quote-based; Voice AI priced separately.

Verdict: the best way to make later human interviews consistent. For the first screen, compare Voice AI’s evidence with a CV-blind interview that has a published audit.

BrightHire (owned by Zoom)

Best for: teams improving interviewer consistency and calibration.

What it does: interview planning, recording, AI notes and summaries, and interviewer coaching; BrightHire Screen runs async voice or video AI interviews. Zoom completed its acquisition in December 2025; BrightHire still runs as a separate product.

Evidence profile: Candidate produces: recorded interviews rated by humans; spoken answers for Screen. Score built from: human ratings; Screen scores “using recruiter-defined rubrics… not resume information”. Published validity: none published. Independent audit: annual babl.ai audits claimed; results not published. Integrity controls: deepfake signals mentioned; no figures.

Strengths: evidence-based debriefs from recorded interviews; interviewer coaching; strong ATS integrations.

Where it falls short: reviewers cite recording sync and speaker-attribution errors; recording raises consent questions in some markets; audit results aren’t public.

Pricing: quote-based; Screen sold separately.

Verdict: strong for calibrating interviewers – not a CV-blind first screen with published audit evidence.

Measuring quality of hire and checking references (ways 6 and 7)

Crosschq

Best for: TA leaders who need to track quality of hire across the funnel and run fraud-resistant digital references.

What it does: Quality of Hire Analytics links source, interviewer ratings, reference scores and assessments to performance ratings, retention and hiring-manager surveys from your HR systems. It also offers 360 reference checks, AI interviews, Traitify assessments (acquired March 2026) and ID.me verification.

Evidence profile: Candidate produces: reference responses and assessment answers. Score built from: reference ratings and “outcome-trained models” (vendor wording). Published validity: none published. Independent audit: none published. Integrity controls: fraud and deepfake detection; no figures.

Strengths: the only dedicated quality-of-hire module among these tools; fraud-resistant references; broad HR system integrations.

Where it falls short: no published validity or audit; value depends on clean performance data, and reviewers cite integration friction with some HRIS platforms.

Pricing: quote-based.

Verdict: the practical route to closing the loop. Pair it with a structured, audited first screen so the data you analyse is worth analysing.

Where Sapia fits (and where it doesn’t)

Our lane: way 2 – a CV-blind, untimed, text-first structured interview for every applicant – supported by JAS for way 1 and Experience Scoring for way 3, with an independent, published audit (BLDS, late 2022, US and Canada models, 72 tests) and AI-answer detection with published, vendor-reported figures. It’s the lever with the strongest research case, which is why we lead the tools section.

Where we don’t fit, and who to choose instead:

• The deepest published psychometric validity → SHL.
• Hard-skills and coding tests → TestGorilla or a specialist coding platform.
• Structured human interviews inside your ATS → Greenhouse.
• Interviewer calibration from recorded interviews → BrightHire.
• Closed-loop quality-of-hire analytics and references → Crosschq, or your own BI stack.
• Native voice interviewing → HireVue or Greenhouse Voice AI.
• A system of record → your ATS. Sapia sits on top of it.

What you should still ask us: for criterion-validity evidence in roles like yours, and for a current, role-level adverse-impact analysis. Ask every vendor the same.

In practice (customer-reported before-and-after results via Sapia case studies, not controlled studies): Holland & Barrett reported an 89% reduction in early churn, and LNER maintained 30% ethnic-minority representation through each hiring stage (diversity hiring).

If you want your first screen built on what candidates tell you rather than what their CV says, book a demo.

How to improve quality of hire in 90 days – a 5-step plan

1. Weeks 1–2: set the baseline and definition. Agree a three-measure definition – 6- or 12-month performance, 12-month retention and hiring-manager satisfaction – and pull the last 12 months of data for one role family.
2. Weeks 2–4: rebuild the success profile. Five competencies, a clear experience level and a realistic job preview (way 1).
3. Weeks 4–8: pilot a structured interview at the front of the funnel. One high-volume role family, with the CV used only for eligibility. Add structured experience questions where the role needs them (ways 2 and 3).
4. Weeks 6–10: structure the later interviews. Interview kits, anchored rubrics, independent ratings and one calibration session (way 5). Set your integrity rules and move references to the final stage (way 6).
5. Weeks 10–13: review and commit. Compare completion, shortlist-to-offer ratio, 90-day attrition and selection rates by group against the baseline. Book the 6- and 12-month performance check in the diary (way 7).

Conclusion

You improve quality of hire by changing the selection decision, and the research is clear about where to start. The CV and years-of-experience screen most teams rely on is one of the weakest predictors of performance; a structured interview is one of the strongest. Structure every interview that follows, add tests only where the job needs them, protect the signal from AI-written answers and fraud, and close the loop with post-hire data.

By lane: Sapia for a CV-blind structured interview with a published audit; HireVue for video and voice at enterprise scale; SHL for psychometric depth; Criteria, TestGorilla and Harver for job-specific tests; Greenhouse and BrightHire for interview consistency; Crosschq for measuring the result.

If you want to see what your first screen looks like built on candidates’ answers rather than their CVs, talk to us.

FAQs

About Author

Barb Hyman

CEO & Founder