Years of Experience: Why It Keeps Failing Hiring Teams and What Actually Predicts Performance

TL;DR

Years of experience is a weak predictor of job performance, yet it remains a common screening criterion across the hiring process. This article explains what the research says about experience, where years-based requirements can narrow candidate pools or introduce bias, and which assessment methods give hiring teams better evidence of who can perform the work. It also looks at how structured interviews, recruitment technology, and Sapia’s CV-blind experience scoring can help organisations assess experience more consistently without simply counting years.

Imagine you’re in a shortlist meeting. One candidate gave the strongest interview answers and the clearest evidence of the outcomes they’ve delivered. Another has one more year on their CV. The first candidate is cut, because the job description says five years’ experience and they only have four.

It sounds crude, but it’s a familiar reflex. Years of experience is cheap to screen on, easy to defend in a meeting, and simple for an applicant tracking system to apply across thousands of applications. Above all, it feels objective.

Years of experience is the total time a person has spent in a role, field or industry, usually counted from a CV and used as a minimum screening requirement.

The problem is that easy to measure doesn’t mean it’s useful for prediction. Counting years is one of the weakest predictors in a century of selection research. It also shrinks candidate pools before anyone has been assessed, creates age and career-break exposure, and is now being applied at scale by AI CV screeners. As Sapia CEO Barb Hyman puts it:

“The CV rewards whoever writes the best summary of their career, not whoever did the best work in it.”

This article gives you three things: the evidence, with numbers you can take to a hiring manager; a ranked comparison of 13 selection methods that predict performance better; and a practical way to keep assessing experience for the roles that genuinely need it. It draws on the selection research and on Sapia’s experience of running structured interviews with more than 10 million candidates.

Why Years of Experience Keeps Failing as a Hiring Signal

The problem with years of experience isn’t that experience has no value, it’s that time is a poor way of measuring its value.

The largest meta-analysis focused specifically on pre-hire experience makes that clear. Van Iddekinge, Arnold, Frieder and Roth (2019) analysed 81 independent samples and found a corrected correlation of .06 between pre-hire experience and job performance (based on 44 samples), .11 with training performance, and .00 with turnover. Even when the experience was in the same type of work, the relationship with performance only reached .07.

In practical terms, how long someone had spent in a line of work told employers almost nothing about how well they would perform, and nothing at all about how long they would stay. But the authors were careful about what that means. As Van Iddekinge put it: “Don’t ignore experience, but the way employers measure it now does not appear to be very effective.”

That finding isn’t an outlier. Schmidt and Hunter’s widely cited 1998 review put job experience at .18. When Sackett, Zhang, Berry and Lievens (2022, Journal of Applied Psychology) re-estimated the evidence using more recent methods, the figure fell to .07, the second-lowest of the 25 predictors they examined.

The drop is largely because the older estimate leaned on studies of employees’ experience in the job they were being rated in, while the newer figure reflects pre-hire experience – how long someone worked before you hired them. That’s exactly what a CV screen measures.

So why does counting years perform so poorly? The research points to five reasons.

Time Isn’t the Same as Quality

Two people can both have five years of experience and have done completely different things with them. One might have taken on steadily greater responsibility, solved difficult problems and delivered measurable results. The other might have done broadly the same job, at the same scope, five times over.

Both have five years, but only one has necessarily built the kind of experience the role requires.

Tesluk and Jacobs (1998) describe experience as multidimensional: what someone has done, how often they’ve done it, how challenging it was and what they learned all matter. The calendar captures none of that.

Task-Level Practice Matters More Than Elapsed Time

Quiñones, Ford and Teachout (1995) compared different ways of measuring experience and found that the method made a big difference. Measured as tenure within an organisation, experience had a relationship of around .16 with performance. Measured as the number of times someone had actually performed a task, it rose to around .43, and measured at the level of specific tasks rather than whole jobs, around .41.

These were mostly samples of existing employees, so they shouldn’t be read as a direct estimate of how well a CV screen predicts future performance. But they illustrate the underlying principle: what someone has actually practised tells you far more than how long they’ve been employed.

That changes the question a hiring team should be asking. Instead of:

“Does this candidate have seven years of experience?”

Ask:

“Has this candidate actually done the work this role requires, at the level of complexity and responsibility we need?”

The Returns from Experience Flatten

More experience doesn’t keep adding value indefinitely. Schmidt and Hunter found the experience–performance relationship was strongest when people had less than five years in a job, and Sturman (2003) found it approached zero at around 15 years in many jobs.

That doesn’t mean someone with 15 years’ experience is no better than someone with one. It means the number becomes less and less useful as a stand-in for capability the higher it goes. Treating “15 years” as inherently better than “10 years” tells a hiring manager very little, particularly if the extra five years weren’t spent doing work relevant to the role.

Experience Can Even Carry Unhelpful Habits

Experience isn’t automatically an asset, however. Dokko, Wilk and Rothbard (2009, Organization Science) found that prior experience improved performance only through the knowledge and skills it built. Once those were accounted for, more prior experience was actually linked to lower performance. 

The likely explanation is that experience brings assumptions, routines and habits along with expertise, and not all of them transfer well. That’s particularly relevant when someone is moving between organisations with different cultures, processes, technologies or customer expectations.

Titles and Tenure Don’t Transfer Cleanly

A job title doesn’t mean the same thing from one company to the next. “Head of Sales”, “Product Manager” or “Engineering Manager” can describe very different scopes of responsibility depending on the organisation’s size, structure and maturity. As Van Iddekinge and colleagues have noted, the context in which someone gained their experience doesn’t move with them when they change jobs.

A title tells you what someone was called. It doesn’t tell you what they actually did. The problem, then, isn’t experience, it’s measuring experience by the calendar.

The Hidden Costs of Years-Based Screening

Using years of experience as an early screening rule can look harmless because it’s so easy to apply. But the consequences extend beyond the candidate who gets filtered out.

You Shrink the Talent Pool Before Assessing Anyone

A requirement such as ‘five years’ experience’ creates a hard boundary before the recruitment team has assessed what applicants can actually do, which means some of the best candidates can be filtered out before anyone has read their application.

That matters when qualified candidates have taken unconventional career paths, changed industries, returned after a career break, or developed relevant skills outside a traditional progression.

It also happens more often than many hiring teams realise. A TalentWorks analysis (2018, reported by SHRM) found that 61% of full-time “entry-level” job postings asked for three or more years’ experience – screening out the very candidates those roles were supposedly designed for.

The market is moving away from rigid experience requirements in some areas, but not evenly. In fact, Hiring Lab reported that 30% of US job postings specified a number of years of experience in May 2024, down from around 40% in 2022. Yet among technology postings that mentioned experience, the proportion asking for five or more years increased from 37% to 42% between Q2 2022 and Q2 2025.

LinkedIn’s Economic Graph has also reported that skills-based hiring can expand candidate pools substantially – by 6.1 times globally, 8.4 times in the UK and 7.7 times in Australia – with larger gains for women and for workers without degrees. These are LinkedIn’s own figures, based on its platform data, but the direction is clear: years-based requirements narrow candidate sourcing before assessment has even begun.

None of this means every organisation should remove every experience requirement. It does suggest hiring teams should ask whether a particular requirement is genuinely necessary before turning it into a screening rule.

Years of Experience Requirements and Age Discrimination 

A requirement for a minimum number of years can indirectly disadvantage younger candidates. A maximum-years requirement can create the opposite problem for older workers. How much legal exposure that creates depends on where you’re hiring, so the summary below is an overview rather than legal advice.

UK

Under the Equality Act 2010, a years-of-experience requirement can amount to indirect age discrimination if it cannot be objectively justified. Acas advises that it’s “better to set out the type or types of experience needed rather than ask for a number of years’ experience”. In Rainbow v Milton Keynes Council (2008), an employment tribunal found that an advertisement aimed at teachers in the first five years of their careers amounted to indirect age discrimination.

US

Maximum-years requirements have been challenged as potential age proxies under the Age Discrimination in Employment Act (ADEA). Whether external job applicants can bring disparate-impact claims is contested: the Seventh Circuit said no in Kleber v. CareFusion (2019), while the court in Mobley v. Workday allowed them in 2026. Federal enforcement of disparate-impact claims has been deprioritised under Executive Order 14281, but private litigation and state law mean the risk hasn’t gone away.

Australia

The Age Discrimination Act 2004 covers recruitment and job advertising. Indirect discrimination turns on whether a requirement is reasonable in the circumstances, so a years-of-experience threshold that can’t be clearly linked to the job is difficult to defend.

Across all three jurisdictions, the practical point for hiring managers is the same: if a number isn’t genuinely necessary for the job, it’s worth asking why it’s there.

What About Career Breaks?

Career breaks deserve a more nuanced approach. A 2021 Behavioural Insights Team/Government Equalities Office CV trial found no significant callback penalty associated with a 2.5-year career gap, and summarising experience as total years actually increased callbacks.

That doesn’t prove career breaks never affect hiring outcomes. But it does mean the case against counting years rests on prediction and proxy risk, not on a simple assumption that gaps are always penalised. Either way, the answer is the same: look at what candidates have done, rather than assuming a conventional, uninterrupted timeline is inherently better.

How Applicant Tracking Software and AI Screeners Scale a Weak Signal 

Applicant tracking systems and applicant tracking software make it easy to manage large numbers of job applicants. That’s useful, but automation doesn’t make a weak hiring signal stronger.

If an applicant tracking system or AI screening tool ranks candidates using tenure, previous job titles or employer history, it can simply apply the same questionable assumptions to more people, more quickly, potentially filtering out qualified applicants before a hiring manager has the chance to assess them. This is one reason hiring teams need to understand what candidate data their screening tools actually use.

AI also introduces risks of its own. Research by Wilson and Caliskan (AIES, 2024) found that LLM-based résumé screeners favoured white-associated names 85% of the time. And in Mobley v. Workday, a job applicant alleges that AI-driven screening disadvantaged older candidates; the court conditionally certified a nationwide ADEA collective action in May 2025, and the case is ongoing.

The lesson isn’t that organisations shouldn’t use recruiting software or automated screening, but rather that automation should be applied to good selection practices, not used to scale weak ones.

Applicant Tracking System (ATS) Software: Where It Helps and Where It Doesn’t 

Most applicant tracking system software is designed to automate the administrative work around hiring; they can make the hiring process more efficient, but they can’t turn a weak hiring signal into a strong one. An ATS platform can help teams understand how applicant tracking systems work in practice: by bringing applications, candidate information, communication, workflows, and recruitment activity into one system.

Most ATS software is designed to automate the administrative work around hiring, giving recruiters and hiring managers more time to focus on evaluating candidates. They can also create a more consistent, positive experience for candidates by reducing delays and making candidate communication more predictable.

The best ATS software can automate repetitive tasks, organise candidate information, support candidate communication, and give hiring teams a shared view of recruitment activity. It should make better hiring easier, not simply make old screening habits faster.

Applicant Tracking System Benefits: What an ATS Can Automate

An applicant tracking system ATS can support much of the administrative work involved in moving candidates through the recruitment process. Essential features can include:

• Job distribution: post job openings to multiple job boards and social platforms from one place, rather than posting jobs individually.
• Candidate tracking: keep job candidates, applications, CVs, notes, and other candidate information in a central database or single database.
• Workflow management: move candidates through customisable recruitment stages, from application and screening to the interview stage and offer.
• Interview scheduling: automate scheduling interviews and allow candidates to select suitable timeslots, reducing back-and-forth communication.
• Candidate communication: send automated updates, reminders, and emails so candidates know where they stand throughout the process.
• Team collaboration: give hiring managers and recruitment teams one platform to leave feedback, ratings, and notes, rather than relying on spreadsheets, email, or manual tracking.
• Reporting: track recruitment activity, hiring performance, time to hire, and recruitment costs across roles and teams.
• Integrations: modern ATS platforms integrate with job boards and HR systems, as well as specialist assessment tools, so candidate data flows between them without manual re-entry.
• Compliance: an ATS helps maintain compliance with data privacy regulations like GDPR by managing consent, retention periods and candidate data requests in one place.

These applicant tracking system benefits can save time for HR teams and recruiters, particularly when organisations are hiring at high volume. They can also create a more consistent, user-friendly candidate experience by reducing delays and making candidate communication more predictable.

Modern ATS platforms may also include AI features such as resume parsing, automated screening, and natural language processing. These tools can extract information from incoming CVs, categorise applications, and help recruiters screen candidates and identify potential candidates more quickly.

But this is where the quality of the underlying hiring criteria matters. If an ATS is configured to screen candidates based on a requirement such as “five years of experience”, automation doesn’t solve the problem identified in this article. It simply applies the same weak proxy more efficiently. A candidate with four years of highly relevant, high-impact experience can still be filtered out before a hiring manager ever sees their application.

Building an Inclusive Workforce: Why Automation Should Support Judgement 

The same principle applies to AI screening. AI can compare large numbers of applications faster than a recruitment team can manually review them, but speed isn’t the same as predictive validity. Hiring teams still need to decide which evidence actually matters for the role and how candidates should be assessed against it.

That distinction is particularly important when organisations want to build an inclusive workforce. Automated workflows can help reduce bias in administrative processes by applying the same process consistently, but they don’t automatically remove unconscious bias from the criteria themselves. A biased or overly simplistic screening rule can be applied consistently and still produce an unfair outcome.

The strongest approach is therefore to use an ATS for what technology does well: organising candidate data, automating repetitive tasks, supporting candidate communication, coordinating interview scheduling, and giving hiring teams a shared view of the recruitment process. Assessment methods should then focus on evidence that has a meaningful relationship with performance.

In other words, automation can help you process more candidates. It can’t tell you whether the right candidates are being assessed in the first place.

When Years of Experience Requirements Still Make Sense

Experience does matter, and in some situations a requirement belongs in your hiring criteria.

Licensed and Safety-Critical Roles

Some jobs have statutory minimums, such as required flight hours or professional licensure. In these cases, the requirement should remain a clear eligibility gate: keep it, state it precisely, and don’t stretch it to roles that don’t need it. Just don’t treat it as evidence that more experience automatically means better performance.

People New to a Type of Work

Experience tends to matter more early on. Van Iddekinge et al. (2019) found a somewhat stronger relationship for people new to a job, although this was based on a small subset of studies. It’s consistent with the earlier finding that the returns from experience are greatest in the first few years.

Highly Task-Specific Experience

As the Quiñones research shows, experience that closely matches the tasks involved does predict performance. That’s the whole point: assess that experience directly, not the years around it.

Senior and Complex Professional Roles

Domain knowledge, networks and exposure to difficult situations can genuinely matter. But these are precisely the situations where counting years is often the least informative option. If you’re hiring someone to lead a complex technology migration, “seven years of experience” tells you less than knowing whether they’ve actually owned a migration of comparable scale, what went wrong, what decisions they made and what happened as a result. Job knowledge tests and structured questions about scope and outcomes will get you that evidence; a tenure threshold won’t.

The research case isn’t: Experience doesn’t matter. It’s: Measure experience better.

What Actually Predicts Performance: 13 Selection Methods Ranked

If years of experience is weak, what should hiring teams use instead? To answer that, we compared 13 selection methods against six criteria.

How We Scored and Ranked the Methods 

Predictive validity – how strongly the method is related to job performance. We used Sackett, Zhang, Berry and Lievens (2022) as the primary source, with Schmidt and Hunter (1998) for context, and say clearly where a method hasn’t been re-estimated using modern corrections.

Fairness and defensibility – including subgroup differences (the standardised difference in average scores between groups, shown as d, where lower is fairer) and potential proxy risks such as age, socio-economic background or career breaks. Pooled figures can hide disparities at role level, so buyers should ask any vendor for role-level analysis.

Candidate experience and accessibility – whether applicants accept the method and can complete it fairly. Research on applicant reactions (Anderson, Salgado & Hülsheger, 2010) found candidates rate work samples and interviews most favourably, and honesty tests and personal contacts least favourably.

Integrity in the AI era – how vulnerable the method is to AI-generated answers or other forms of faking, particularly in unproctored settings (Hickman et al., 2025). So far, evidence suggests real-world score inflation has been small (Dabdoub & Pool, 2026), but the risk varies considerably by method.

Scalability and fit – whether the method works practically for the role and volume involved, including whether it suits entry-level candidates. Work samples and job knowledge tests, for example, generally don’t (Sackett et al., 2023).

Explainability – whether hiring teams can clearly explain to a candidate, hiring manager or regulator why someone received a particular score, and whether that reason is job-related.

Validity comes first, but it isn’t the only consideration. A highly predictive method that creates substantial subgroup differences or is particularly vulnerable to manipulation may be less attractive than a slightly less predictive method with a stronger overall fairness and integrity profile.

The rankings below are an editorial assessment of the published evidence, rather than a universal rule for every recruitment process. Assessment centres (.29 in Sackett et al., 2022) aren’t ranked separately because they combine several of the methods below.

Validity figures are corrected correlations with job performance. Subgroup d values are Black–White differences from Sackett et al. (2022) where available.

Tier 1: Strongest Evidence 

1. Structured Interviews

Best for: almost any role or level.

A structured interview gives every candidate the same job-related questions, derived from a job analysis, and scores their answers against predefined, behaviourally anchored criteria. Behavioural questions ask candidates to describe something they have done (“Tell me about a time when…”), while situational questions ask how they would respond to a hypothetical scenario (“What would you do if…”). Taylor and Small (2002) found past-behaviour questions somewhat more predictive (.56 vs .45 with anchored scales).

The important part isn’t simply asking good interview questions. It’s standardisation: the same questions, the same scoring approach and job-related anchors.

The evidence: Sackett et al. (2022) placed structured interviews first among the 25 predictors they re-estimated, at .42, compared with .51 in Schmidt and Hunter’s 1998 estimate. The range across studies is wide, though (roughly .18 to .66), so the quality of the build matters. Structured interviews also showed a subgroup difference of .23, the lowest of the top-tier predictors, and applicants rate interviews among the most acceptable selection methods.

Where it falls short: a badly designed “structured” interview loses much of its advantage. Live interviews are expensive to deliver consistently at scale and drift without interviewer training. Unproctored text and asynchronous formats are exposed to AI-generated answers, so they need active controls such as job-specific scenarios, follow-up probes and AI-answer detection, and those detectors carry fairness questions of their own.

Tools that deliver it (as of October 2026): Sapia.ai offers an untimed, text-first mobile chat interview, scored on the candidate’s answers rather than their CV, with experience scoring for roles that need it. SHL’s Smart Interview offers AI-scored asynchronous video, and SHL publishes the deepest validity documentation in the category. HireVue offers video interviews and, since June 2026, a voice AI Interviewer; we found no published independent audit of the voice product, although earlier audits cover 2023–24. Greenhouse now offers a native voice interviewer through its acquisition of Ezra, scored on the customer’s own rubric, and publishes monthly independent audits. It’s worth checking what each tool’s score is built from: some AI screeners, such as Bullhorn’s Amplify Screener, blend interview answers with résumé data, which reintroduces the CV signals this article argues against.

Verdict: The strongest general-purpose replacement for a years-of-experience screen, provided the interview is genuinely structured.

2. Structured Assessment of Experience Quality

Best for: experienced and professional roles where the real question is “Have they actually done it?”

Instead of giving candidates points for the number of years they’ve worked, ask them to describe relevant work they’ve owned:

• What did they do?
• What changed because of their decisions?
• How complex was the work?
• What was the outcome?

Trained raters, or a scoring model, then score those answers against consistent, job-related anchors.

This is known in the research as the behavioural consistency or accomplishment record method. It’s the opposite of the “point method”, which awards credit for years and credentials – essentially what a CV sift does.

The evidence: McDaniel, Schmidt and Hunter (1988) found a substantial difference between the behavioural consistency approach (.45) and the point method (.11). This is older evidence and wasn’t re-estimated by Sackett et al. (2022), so it should be treated as strong but dated. On fairness, the accomplishment record method has generally shown little or no difference by sex or race (Hough, 1984), while point-method scoring has been found to disadvantage women (Alonso et al., 2009).

Where it falls short: the headline validity figure hasn’t been re-checked with modern methods. Candidates are still reporting their own experience: structured questions make thin evidence easier to spot, but they don’t independently verify someone’s claims, and narrative answers are now easy to draft with an LLM. Traditional versions also take weeks to build and can be burdensome for applicants, which is why relatively few organisations have used them.

Tools that deliver it (as of October 2026): Sapia.ai’s experience scoring adds up to two structured experience questions to its chat interview, scored on depth and progression, quality and scale of outcomes, and relevance and transferability. Anchors are calibrated to the role’s experience level, and experience is scored separately from competencies. The capability applies the same logic as the research above – structured questions and consistent anchors – and is currently being validated in customer pilots. Beyond that, this method is typically built in-house or with I/O psychology consultancies, and the US Office of Personnel Management publishes guidance on the accomplishment record method.

Verdict: The evidence-backed answer to “But experience matters” – assess it rather than count it.

3. Job Knowledge Tests

Best for: experienced hires in roles with a defined body of knowledge, such as accounting, engineering or regulated trades.

Job knowledge tests directly assess what candidates know about a specific domain, usually through multiple-choice or short-answer questions. They answer the question that years of experience is often used to approximate: does this person actually know the field?

The evidence: Sackett et al. (2022) estimated validity at .40 (.48 in 1998). They’re also inexpensive to administer once built. The subgroup difference is larger than for interviews, at .54.

Where it falls short: they’re generally unsuitable for entry-level candidates and career changers (Sackett et al., 2023). Unproctored tests are highly exposed to outside assistance, including AI, so they need proctoring or question rotation.

Tools that deliver it (as of October 2026): SHL and Criteria; for software roles, HackerRank, Codility and CodeSignal.

Verdict: The sharpest substitute for tenure when the role really is knowledge-heavy.

4. Empirically Keyed Biodata

Best for: very high-volume hiring with substantial historical outcome data.

Empirically keyed biodata uses a structured questionnaire about candidates’ backgrounds and past behaviour, with each response weighted according to how well it has predicted outcomes in your own data.

The evidence: Sackett et al. (2022) estimated validity at .38 (.35 in 1998), with a moderate subgroup difference of .33.

Where it falls short: it needs large samples to build and periodic re-keying to stay accurate. The weightings can be difficult to explain to candidates and hiring managers. And if the underlying data isn’t carefully audited, items can encode proxies such as location or background.

Verdict: Powerful in the right environment, but impractical for many organisations.

Tier 2: Useful, With Caveats

5. Work Samples and Job Simulations

Best for: experienced hires and technical roles where candidates can demonstrate the work itself.

Work samples ask candidates to complete a task that resembles the job. They have strong face validity because applicants can see the connection between the assessment and the work they’re applying to do, and candidates rate them among the most acceptable selection methods (Anderson, Salgado & Hülsheger, 2010).

The evidence: Sackett et al. (2022) estimated validity at .33, the biggest downward revision of any method from the 1998 estimate of .54. The subgroup difference, at .67, is also larger than many hiring teams assume.

Where it falls short: work samples take significant time and cost to design for each role and are generally unsuitable for entry-level positions. Take-home exercises are highly exposed to AI assistance.

Tools that deliver it (as of October 2026): Vervoe and TestGorilla; for coding, HackerRank and Codility.

Verdict: Excellent where the work itself can be assessed fairly and practically.

6. Cognitive Ability Tests

Best for: complex roles where general reasoning ability is relevant, used with care.

Cognitive ability tests measure general mental ability through timed reasoning, numerical and verbal problems. They’re cheap, scalable and predictive across complex roles.

The evidence: Sackett et al. (2022) estimated validity at .31, well down from .51 in 1998, and later work found .22 in 21st-century data (Sackett et al., 2024). The size of that drop is actively debated (Oh, Le & Roth, 2023). The bigger concern is fairness: at .79, cognitive tests showed the largest subgroup difference of the common methods.

Where it falls short: alongside the adverse impact, timed testing raises accessibility concerns. Unproctored tests are exposed to reasoning models (Hickman et al., 2025), although the average real-world score inflation found so far has been small (Dabdoub & Pool, 2026).

Tools that deliver it (as of October 2026): SHL and Criteria.

Verdict: Predictive and scalable, but the fairness trade-off needs serious consideration.

7. Situational Judgement Tests

Best for: graduate and volume hiring.

Situational judgement tests present job-related scenarios and ask candidates to select or rank possible responses. They’re scalable and can work well when organisations need to assess judgement across large candidate pools.

The evidence: Sackett et al. (2022) estimated validity at .26, with subgroup differences in the .34–.39 range.

Where it falls short: they sit below several alternatives on predictive validity, and LLMs perform well on many SJT-style questions, which makes unproctored use increasingly exposed.

Verdict: Useful for volume screening, but not the strongest standalone predictor.

8. Integrity Tests

Best for: roles where counterproductive work behaviour is a significant concern.

Integrity tests use questionnaires to assess attitudes and past behaviour related to honesty, rule-following and reliability.

The evidence: Sackett et al. (2022) estimated validity at .31 – the same as cognitive ability – but studies not conducted by test publishers have found a much lower relationship with job performance, at .15 (Van Iddekinge et al., 2012). They do predict counterproductive work behaviour and show relatively small subgroup differences.

Where it falls short: that disagreement over validity is why they rank below SJTs here, despite the higher headline figure. Applicants also rate integrity tests among the least acceptable selection methods, and responses can be faked.

Verdict: Useful for a specific risk, but not a universal measure of performance.

Tier 3: Weak on Their Own

9. Contextualised Conscientiousness

Best for: supplementary assessment.

Contextualised personality questions frame conscientiousness around workplace behaviour rather than asking about personality traits in the abstract. Framed this way, Sackett et al. (2022) estimated validity at .25, compared with .19 for conscientiousness measured generally, with no meaningful subgroup difference (−.07).

The limitations are modest validity and fakeability: candidates can often infer the “right” answers, and LLMs can readily produce an ideal profile.

Verdict: A potentially useful supplementary signal, not a replacement for stronger methods.

10. Reference Checks

Best for: late-stage verification.

Reference checks can provide useful context from someone who has worked with the candidate. Schmidt and Hunter (1998) estimated validity at .26, but the method hasn’t been re-estimated in the same modern framework. Referees are also selected by candidates themselves and are rarely critical.

Verdict: Useful at the end of the recruitment process, but don’t expect references to rescue a weak screening process.

11. Unstructured Interviews

Best for: building rapport and selling the role.

Unstructured interviews are familiar because they feel conversational and natural. But they’re much less consistent: different interviewers ask different questions, place different weight on answers and can reach very different conclusions about the same candidate. Sackett et al. (2022) estimated their validity at .19, down from .38 in 1998 and less than half the .42 for structured interviews, with a subgroup difference of .32. That inconsistency also makes decisions harder to explain and defend.

Verdict: Keep the human conversation, but give it structure.

Tier 4: Counting Proxies 

12. Years of Education and Degree Requirements

Best for: roles where a specific qualification is genuinely required.

Degree requirements can be legitimate when a qualification is necessary for the work, but years spent in education aren’t a strong general predictor: Schmidt and Hunter (1998) estimated .10. And while many employers have dropped degree requirements on paper, research from Harvard Business School and the Burning Glass Institute (2024) found the change affected fewer than 1 in 700 hires in practice.

Verdict: Keep genuine qualification requirements. Question blanket credential filters.

13. Years of Experience

Best for: statutory eligibility gates only.

Years of experience requires no candidate assessment at all. The hiring team, or an applicant tracking system, simply reads a number from the CV and compares it with a threshold.

The evidence: as covered earlier, Sackett et al. (2022) estimated its relationship with performance at .07, down from .18 in 1998, and Van Iddekinge et al. (2019) found .06 for performance and .00 for turnover. It also carries age and career-break proxy risk, with a subgroup difference of .49. And because CVs are now routinely polished with AI, while AI screeners apply tenure filters at scale, the signal is getting noisier, not cleaner.

It’s cheap, fast and easy to explain, but it says nothing about what someone actually did.

Verdict: Keep experience requirements where they’re genuinely necessary as eligibility gates. Everywhere else, replace the number with a structured interview or a structured assessment of experience quality.

How to Assess Experience Properly, Without Counting Years

If you remove “five years’ experience required” from a job description, you still need an answer to an important question: what experience does this candidate actually have?

Here’s a practical way to answer it.

1. Start With the Job, Not the CV

Begin with job analysis. Identify the tasks, scope, responsibilities and outcomes that actually matter for the role, then decide whether previous experience is necessary at all. If it is, be specific about which experience matters and why.

Sapia’s Job Analysis Studio can help hiring teams structure this process.

2. Describe Experience by Type, Not Duration

Replace:

• “Seven years’ experience in financial services.”

With something more meaningful, such as:

• “Experience owning a P&L of meaningful scale in a regulated financial services environment.”
• “Experience leading an end-to-end technology migration across multiple business functions.”

The second versions tell candidates and hiring managers what the organisation actually needs, while giving the recruitment team something meaningful to assess. It’s also the approach Acas recommends.

3. Ask Every Candidate the Same Questions

For example:

• “What’s the most complex or challenging work you’ve been responsible for?”
• “What have you learned in your career that you’d bring to this role?”

The exact questions should be derived from the job analysis. What matters is that every applicant gets the same opportunity to demonstrate relevant experience, rather than being judged on how well their CV happens to summarise it.

4. Score the Answers Against Consistent Anchors

Assess three dimensions:

• Depth and progression – has the candidate’s scope and autonomy increased over time?
• Quality and scale of outcomes – what results did their decisions produce, and how much can they reasonably attribute to their own work?
• Relevance and transferability – how directly does the experience map to this role?

Anchors are short descriptions of what an answer looks like at each point on the scale, and they should be calibrated to the level of the job so that a particular score means the same thing for every applicant. For a mid-level role, an illustrative anchor for quality and scale of outcomes might look like this:

Without anchors, two assessors can read the same answer and score it very differently. With them, you’re scoring the evidence rather than the impression it makes.

5. Keep Experience Separate from Competencies

Experience and competency aren’t the same thing. Someone can have extensive experience without demonstrating the competencies required for the role. Conversely, someone can have less experience but demonstrate exceptional judgement, problem-solving and learning ability.

Score them separately, then read them together. One tells you how someone thinks and works; the other tells you what they’ve actually delivered.

6. Probe the Strongest Claims

Structured assessment doesn’t make self-reported experience independently verified. Candidates are still describing their own work, just as they do on a CV.

Use structured follow-up questions at the next stage to test the strongest claims and understand what the candidate personally owned. For example: “What was your specific role in that project?”, “What decision would you make differently now?” or “What would have happened if you hadn’t been involved?”

And remember, experience assessment isn’t a hard-skills test. If the role requires Python, Excel or another specific technical capability, pair it with an appropriate job knowledge test or work sample.

CV-Blind Experience Scoring: Where Sapia Fits

Sapia’s role in this process is deliberately specific. Its CV-blind, text-first structured interview gives every candidate the same job-related questions, with scoring based on their answers rather than their CV, education or previous employers.

For roles where experience matters, Sapia’s experience scoring adds structured experience questions to the same interview, assessed on the three dimensions above and scored separately from competencies. Job Analysis Studio recommends whether to include it role by role, but hiring teams make the final decision. Each experience score is presented alongside the competency scores with a plain-English explanation of how it was reached.

That creates a different kind of input for hiring managers. Instead of an applicant tracking system simply telling you “8 years’ experience”, you get structured evidence about what that experience actually involved. Sapia’s Talent Intelligence Assistant (Tia) then lets recruiters search and compare interviewed candidates on that evidence, including their experience scores.

Sapia’s experience scoring is grounded in established assessment approaches and is currently being validated in customer pilots. It doesn’t replace human judgement, and it isn’t a lie detector: answers are self-reported, although structured questions make thin evidence easier to spot

What to Look For in an AI Assessment Platform

Sapia has four characteristics worth considering when evaluating any AI assessment platform, and worth asking every vendor about, Sapia included.

• CV-blind scoring: The interview score is built from the candidate’s answers only.
• Independent bias audit: Sapia’s audit by BLDS, conducted in late 2022, covered 72 tests across its US and Canadian models and found no evidence of practically significant disparate impact within that scope. Several other vendors also publish audits of some kind, and because pooled results can mask differences at role level, buyers should ask any vendor for current, role-level analysis.
• Untimed, text-first format: Candidates complete the interview as a mobile chat, with no camera and no time pressure, so they can think before they answer. Asynchronous video is available as a later stage.
• AI-generated answer detection: Sapia reports a 98% detection rate with a 1% false-positive rate (vendor-reported figures). Research has found that AI-text detectors can over-flag non-native English writers (Liang et al., 2023), so it’s worth asking any text-based vendor for false-positive rates for that group specifically.

What Customers Have Seen

Costa Coffee replaced CV screening with structured chat interviews, and hires were 23% less likely to leave in their first 30 days. Qantas used the same approach for its graduate programme, differentiating more than 4,500 applicants without relying on CVs, with a 96.7% interview completion rate.

Where Sapia Isn’t the Right Fit

If validity documentation is your primary criterion: SHL publishes deeper validity documentation than Sapia, and buyers who place that above everything else should take it into account.

If you need to test hard skills: Sapia isn’t a hard-skills or coding assessment. If a role requires Python, Excel or similar, a dedicated job knowledge test or work sample remains appropriate.

If you’re looking for an ATS: Sapia isn’t an ATS or system of record; it sits alongside the recruitment technology an organisation already uses rather than replacing its applicant tracking system.

If you want voice or video interviews inside your ATS: Greenhouse and HireVue are better placed for interviews run on your own rubric in those formats.

One tool shouldn’t replace every part of your recruitment process. No single tool should replace your entire recruitment process. The point is that experience should become one more piece of evidence, not a number that decides who gets considered.

How to Replace Years-of-Experience Requirements: A 5-Step Process

The steps above cover how to assess experience for a single role. Changing how your organisation hires is a bigger shift, so it’s worth approaching your recruitment efforts systematically.

Ready to review your own job postings? Start here:

1. Audit Your Current Requirements

Pull your live job adverts, job requisitions and applicant tracking system screening rules. List every minimum and maximum years-of-experience requirement, including any knockout questions that filter candidates automatically.

Then ask which ones have a genuine business or statutory basis. Any requirement without one is a candidate for removal.

2. Run a Short Job Analysis

Start with what each role family needs to deliver for your wider business strategy. Then identify what the experience requirement is actually trying to achieve. Does “five years” really mean domain knowledge? Leadership experience? Experience managing a particular scale? Exposure to a specific type of problem?

Write down the underlying requirement rather than the proxy. This is the same process as steps 1 and 2 above, applied across a group of related roles rather than one job.

3. Rewrite Job Adverts Around Experience Type and Outcomes

Replace years with the specific experience candidates need, and remove maximum-years caps entirely: they’re the hardest requirement to justify and the most exposed to age-discrimination challenges.

This gives potential candidates a clearer picture of the job and gives hiring teams a more useful basis for assessment.

4. Replace the CV Sift with Structured Assessment

Use a structured interview for the core assessment, and add structured experience questions for roles that genuinely need them, scored as described above.

Your applicant tracking system can still manage candidate information, job vacancies, communications and workflow. It simply doesn’t need to make years of experience the deciding factor.

5. Pilot and Measure

Don’t redesign your entire recruitment strategy overnight. Choose one role family and test the new approach through one hiring cycle.

Track:

• Candidate completion rates
• Candidate satisfaction
• Pass rates by demographic group at role level, rather than pooled across all roles
• Interview and shortlist conversion
• Early performance
• Retention

Then compare the results with your previous cohort for the same roles. If pass rates differ noticeably between groups, look at which questions or anchors are driving the difference before scaling up. The focus shouldn’t be on simply removing a field from your ATS, but on finding out whether you’re getting better evidence about the people you’re hiring.

How to Assess Experience in Hiring Without Counting Years of Experience 

Years of experience is an easy hiring signal, but the evidence suggests it’s one of the weakest. At .07, it sits near the bottom of the modern research on what predicts job performance, and it can narrow candidate pools, create unnecessary age and career-break risks, and tell hiring managers very little about what someone has actually achieved.

Structured interviews and structured assessment of experience quality offer a better route. Both ask every candidate the same job-related questions and score the answers consistently, so experience is judged on what someone did rather than how long they did it for.

The simplest place to start is with your own job adverts. Pick one role family, audit its experience requirements, and pilot a structured alternative for one hiring cycle.

Sapia’s CV-blind structured interview, with optional experience scoring for professional roles, is designed for exactly that part of the recruitment process. It gives hiring teams a way to assess experience without letting the CV’s timeline make the decision. It won’t test hard skills or replace your ATS, and its experience scoring is still being validated in customer pilots. But when the question is whether someone has actually done the work, it gives you evidence rather than a number.

Experience still matters. You just don’t need to keep counting it.

See how Sapia’s experience scoring works or book a demo.

Frequently Asked Questions About Years of Experience as a Hiring Signal

About Author

Barb Hyman

CEO & Founder