Research
Hiring failure and high performance: what the research says.
How often hires go wrong, what it costs, what predicts a good hire, and how high performance is measured, in the UK and US. Every finding names its source, year, sample and what the number measures.
- More than one in five hires are bad or regretted decisions.
- A structured interview (same questions, scored the same way) is now the best single predictor of how someone will do in the job. It scores .42, against .19 for an unstructured chat, roughly double.
- In professional jobs, the top 1% produce roughly double the average. The gap between people is about 48% of average output in professional jobs, against 19% in routine jobs, and bigger still in sales.
- Over two thirds of UK employers never check whether the people they hired went on to perform. Only 7 in 100 check whether their interviews and tests predict who does well.
How many hires fail?
There's no single agreed figure. Published numbers come from different surveys that count different things, so read what each one measures.
Across countries and in the UK
More than one in five hires are bad or regretted decisions.
Source: CEB (now Gartner), 2013, a survey of thousands of hiring managers around the world, as cited in the REC's Perfect Match report, 2017.
- What it measures:
- Share of hires
- Where:
- Global
There's no official UK figure for how many hires fail or how many people fail probation.
Source: Department for Business and Trade, Oct 2024, in its own assessment for the Employment Rights Bill: the department has no robust data on the incidence of dismissal for those under two years of employment. From 1 Jan 2027 the qualifying period for unfair dismissal drops from two years to six months, so getting the hire right matters more.
- What it measures:
- Official
85% of HR decision-makers say a business they have worked for took on someone who proved a bad fit for the job.
Source: REC, Perfect Match, June 2017. YouGov survey of 501 senior HR decision-makers in GB private firms, April 2017. This counts anyone who has ever seen one bad hire anywhere in their career. The REC itself says it doesn't show how often bad hires happen.
- What it measures:
- Share of HR people, ever
Early quits are falling as the job market cools. In 2026, 29% of employers said new recruits at least sometimes quit within 12 weeks (41% in 2024), and 34% said they at least sometimes quit between 12 weeks and six months.
Source: CIPD, Resourcing and Talent Planning, Sept 2026. 1,014 UK HR professionals, YouGov, April to May 2026. Only 5% said it happens always or mostly.
- What it measures:
- Share of employers
Most employers don't check whether their hires work out. Three in ten measure the performance or retention of new hires in their first year, and only 7% check whether their hiring methods actually predict success.
Source: CIPD, 2026, same survey as above.
- What it measures:
- Share of employers
One in 10 new recruits is regarded as a "poor hiring decision".
Source: Robert Half UK, March 2015. More than 200 senior HR executives; research firm not named, question not published. A recruiter's own UK figure, far below "two in five".
- What it measures:
- Share of hires (HR estimate)
46% of senior UK leaders believe they made a bad hire in the past 12 months. 61% blamed settling for someone whose skills didn't match, 56% rushing.
Source: Robert Half UK, April 2022. 300 C-suite leaders at UK firms with 50+ staff, surveyed by Rigour Research. The press release headline "regret almost half of new hires" reads it as a share of hires; the survey figure is a share of leaders.
- What it measures:
- Share of leaders
16% of new hires at UK small firms left within 12 months.
Source: Project Brains, April 2024, reported by HR Review. Project Brains sells freelance talent. No sample size was published by the firm (one write-up says 100 UK executives). Its "51%" headline adds the 35% who "needed support", which isn't failure.
- What it measures:
- Share of hires (estimate)
Almost one in five new employees either fail probation or have it extended.
Source: Spring Personnel (a recruitment agency), May 2014. 403 managers and about 1,500 employees. Original report no longer online. Counts extensions, so it isn't a failure rate.
- What it measures:
- Share of hires
16.5% of UK workers had been with their employer under 12 months in 2023.
Source: ONS, 2024 (Annual Population Survey, 32.7 million in work). Speak to Kit's own sum from the ONS table. This is a snapshot of how long people have been in their jobs, not a failure rate.
- What it measures:
- Share of workers
Being new to the role is the top cause of staff not being fully up to the job: a factor in 71% of skills gaps, and 92% in financial services.
Source: Department for Education, Employer Skills Survey 2022. 72,918 employer interviews. Gaps from being new are expected to close, so this isn't a failure rate.
- What it measures:
- Share of skills gaps
Only 56% of UK graduates are still with the same firm after five years; 47% in law and in retail, FMCG and tourism.
Source: Institute of Student Employers, June 2023. Large graduate employers reporting on their own intakes. Many leave for better pay, so leaving isn't the same as a failed hire.
- What it measures:
- Share of graduate hires
41% of UK employers who hired in the past year said new recruits at least sometimes quit within 12 weeks.
Source: CIPD, 2024, 1,016 UK HR professionals, surveyed by YouGov. This is a share of employers, not a quit rate: it doesn't mean 41% of hires quit. The newer 2026 edition puts the same figure at 29%, as the job market cooled.
- What it measures:
- Share of employers
- Where:
- UK
In the US
46% of new hires fail within 18 months.
Source: Leadership IQ (a US training firm), first published 2005. 5,247 hiring managers at 312 organisations, who between them hired more than 20,000 people. "Fail" means fired, pushed out, disciplined or given a very poor review, as judged by their own manager.
- What it measures:
- Share of hires
- Where:
- US
Managers said they wouldn't rehire 39% of the people they hired in the past 12 months.
Source: IBM Smarter Workforce Institute, 2015, from IBM's own slides. Sample size not stated. IBM sold hiring software.
- What it measures:
- Share of hires
74% of employers say they have hired the wrong person for a job.
Source: CareerBuilder / Harris Poll, Dec 2017. 2,257 US hiring managers and HR professionals. "Ever" in their career, so not a share of hires. CareerBuilder sells job ads.
- What it measures:
- Share of employers, ever
40% of the people who left their job in 2017 had been there less than a year. About half of those left in the first 90 days.
Source: Work Institute, 2018 Retention Report. 34,378 exit interviews at 167 companies. This is a share of leavers. It doesn't mean 40% of new hires quit.
- What it measures:
- Share of leavers
20.6% of US workers had been with their employer a year or less in January 2026. Median time in a job: 4.1 years.
Source: US Bureau of Labor Statistics, Sept 2026. About 60,000 households. A snapshot, not a failure rate.
- What it measures:
- Share of workers
Up to 1 in 4 new hires at tech companies leave in their first year (17% to 25%, higher at smaller firms).
Source: Pave, Oct 2025. Pay data on 200,000+ employees who started in 2023 and 2024, mostly at venture-backed tech firms. Pave sells pay data. Does not split quits from firings.
- What it measures:
- Share of hires
Most executives (52%) guess that 10% to 25% of their new hires leave within six months.
Source: Korn Ferry, March 2017. 1,817 executives, not US only. A guess, not a measurement.
- What it measures:
- Executives' estimate
41.7% of jobs that people started aged 25 to 34 ended within a year.
Source: US Bureau of Labor Statistics, 2025 (National Longitudinal Survey of Youth). 9,964 people followed since 1979. Counts every reason a job ended, not only failed hires. For jobs started aged 35 to 44 it's 25.3%.
- What it measures:
- Share of jobs
- Where:
- US official
What does a bad hire cost in the UK?
A bad hire of a middle manager on £42,000 could cost £132,015, over three times salary.
Source: REC, Perfect Match, 2017. One worked example the authors built, not an average. Over 60% of it assumes a whole team slows down and a colleague quits. Without the colleague quitting it comes to £78,015 (1.86 times salary).
- What it measures:
- Modelled example
Replacing a member of staff on £25,000+ costs £30,614 on average, mostly lost output while the new person gets up to speed. Law £39,887, accountancy £39,230, retail £20,113. About 28 weeks to full productivity.
Source: Oxford Economics for Unum, Feb 2014. 500+ firms in five sectors, plus ONS data. Cost of replacing any leaver, not only bad hires. 2014 money. Unum sells income protection.
- What it measures:
- Cost of any leaver
27% of UK employers who had a bad hire said that single hire cost them more than £50,000.
Source: CareerBuilder / Harris, May 2013. 6,000+ employers across 10 countries, Nov 2012. Their own guess of the cost.
- What it measures:
- Share of employers
The visible cost of filling a role is small: a median £3,500 for senior managers and £2,000 for other staff (ads, agency fees, in-house time).
Source: CIPD, 2026. Small groups of respondents (263 and 223); CIPD warns to take care. Leaves out training, getting up to speed and the cost of a bad hire.
- What it measures:
- Cost to fill a role
The wrong person can cost a typical small business £13,799 (£17,900 in London).
Source: Sandler Training, Oct 2014. 1,000 small business owners, their own guess. Sandler sells sales training.
- What it measures:
- Owners' estimate
What does a bad hire cost in the US?
A bad hire costs nearly $17,000 on average: $11,000 at firms of 500 or fewer staff, $24,000 at firms over 1,000. The 2017 survey said $14,900.
Source: CareerBuilder / Harris Poll, Nov 2016. 2,379 hiring managers and HR professionals. Only employers whose bad hire hurt the business; their own guess.
- What it measures:
- Employers' estimate
Replacing a worker typically costs about a fifth of their salary (median 21%). For senior and executive jobs it can reach 213%.
Source: Boushey & Glynn, Center for American Progress, Nov 2012. 30 case studies from 11 papers, 1992 to 2007. Mostly lower-paid jobs and direct costs, so it understates professional roles. A 2020 follow-up (Washington Center for Equitable Growth, 31 case studies) found a median of 23.5% and an average of 39.6%.
- What it measures:
- Cost of any leaver
The median cost to fill a role is $1,200 for non-executive jobs and $10,625 for executive jobs.
Source: SHRM, 2025. 2,371 HR members. Direct spend to fill a role only.
- What it measures:
- Cost to fill a role
Professionals take about 20 weeks to reach full productivity, executives more than 26.
Source: Rollag, Parise & Cross, MIT Sloan Management Review, 2005. A figure the authors borrowed from an earlier source.
- What it measures:
- Weeks to full speed
Managers spend 17% of their time, nearly a day a week, dealing with poor performers.
Source: Robert Half, Nov 2012. 1,400+ CFOs, their own estimate. Poor performers in general, not only bad hires.
- What it measures:
- CFOs' estimate
Gallup estimates replacing an employee costs half to twice their annual salary.
Source: Gallup, March 2019. No method or source given.
- What it measures:
- Estimate
Why do hires fail, and what predicts success?
When managers overruled a job test and hired against it, their hires left sooner and did no better at the job. Bringing in the test made hires stay just over 25% longer.
Source: Hoffman, Kahn & Li, 2018, Quarterly Journal of Economics. 15 firms, nearly 266,000 hires, mostly low-skill service jobs.
- What it measures:
- Share of hires staying
- Where:
- US
Given the same information about candidates, a fixed scoring rule predicted job performance more than 50% better than experts weighing it up in their heads.
Source: Kuncel, Klieger, Connelly & Ones, 2013, Journal of Applied Psychology. 25 samples, 9 of them on job performance. Score .44 for the rule against .28 for the experts.
- What it measures:
- How well a method predicts
A structured interview (same questions, scored the same way) is now the best single predictor of how someone will do in the job. It scores .42, against .19 for an unstructured chat, roughly double.
Source: Sackett, Zhang, Berry & Lievens, 2022, Journal of Applied Psychology. A re-check of decades of studies. The score runs from 0 (no better than guessing) to 1 (perfect).
- What it measures:
- How well a method predicts
Years of experience tell you almost nothing about how well someone will do (.06) and nothing about whether they will stay (.00).
Source: Van Iddekinge, Arnold, Frieder & Roth, 2019, Personnel Psychology. 81 samples. Sackett 2022 puts experience at .07.
- What it measures:
- How well a method predicts
University grades are a weak sign of performance (.21) and say nothing about staying (-.02). Years of education: .10.
Source: Van Iddekinge et al., 2024, Journal of Applied Psychology; Schmidt & Hunter, 1998, Psychological Bulletin.
- What it measures:
- How well a method predicts
The more structured the interview, the better it predicts: from .20 for the least structured up to .57 for the most.
Source: Huffcutt & Arthur, 1994, Journal of Applied Psychology. 114 results, entry-level jobs.
- What it measures:
- How well a method predicts
A free-form interview made predictions worse, and interviewers given random answers felt just as sure of their judgement.
Source: Dana, Dawes & Peterson, 2013, Judgment and Decision Making. Lab study with students predicting grades. Shows how the effect happens; it isn't a hiring statistic.
- What it measures:
- Lab study
The first few minutes of small talk shaped the interviewer's final rating (.42) and the job offer (.22).
Source: Barrick, Swider & Stewart, 2010, Journal of Applied Psychology. Internship interviews.
- What it measures:
- Study
Applicants a computer picked from CVs, but human screeners had turned down, were 14% more likely to get an offer and more productive once hired.
Source: Cowgill, 2020, Columbia Business School working paper. One large tech employer. Gains came mostly from candidates without a referral, an elite school or a typical CV.
- What it measures:
- Single firm
When an AI voice agent ran the first interview and people still made the decision, job offers rose 12%, starts 18%, and people still in the job after 30 days 17%. Given the choice, 78% of applicants picked the AI.
Source: Jabarian & Henkel, 2025, University of Chicago working paper, pre-registered trial. About 70,000 applicants for entry-level customer service jobs in the Philippines.
- What it measures:
- Randomised trial
Ability barely predicts who quits (.01). Whether the job matched what they expected does.
Source: Griffeth, Hom & Gaertner, 2000, Journal of Management. 42 studies.
- What it measures:
- Predictors of quitting
Many early leavers can be spotted before hire: why they want this job, what they know about it, how long they stayed in past jobs, and whether they know people who work there. The same people also tended to perform better.
Source: Barrick & Zimmerman, 2005 and 2009, Journal of Applied Psychology and Human Resource Management.
- What it measures:
- Predictors of quitting
Being honest about the job, the hard parts too, keeps more people, mainly because they trust the employer. People referred by staff quit less.
Source: Earnest, Allen & Landis, 2011, Personnel Psychology, 52 studies; Burks, Cowgill, Hoffman & Housman, 2015, Quarterly Journal of Economics, 9 large firms.
- What it measures:
- What keeps people
A great hire is worth most in complex jobs. In professional jobs the gap between an average and a good performer is about 48% of average output, against 19% in routine work.
Source: Hunter, Schmidt & Judiesch, 1990, Journal of Applied Psychology.
- What it measures:
- Output gaps
Estimates of how many managers fail range from 30% to 75%, about half on average. The failures are mostly about how they treat people, not technical skill.
Source: Hogan & Kaiser, 2005, Review of General Psychology. The "half" is an average of earlier estimates, not a measured rate. The derailment themes come from Center for Creative Leadership interview studies.
- What it measures:
- Estimates
89% of new hires who fail do so on attitude (taking feedback, managing emotions, motivation, temperament), only 11% on skills.
Source: Leadership IQ, 2005. Managers' own view of why their hires failed. Not peer-reviewed.
- What it measures:
- Share of failed hires
What happens with senior hires and in professional sectors?
Senior and executive
People hired from outside were let go at a 61% higher rate than people promoted into the same jobs, and were 21% more likely to quit. They were paid around 18% more, and rated worse for their first two years.
Source: Bidwell, 2011, Administrative Science Quarterly. 5,260 workers at one US investment bank, 2003 to 2009.
- What it measures:
- Outside hires vs promotions
- Where:
- US
Star analysts who moved to a new firm saw their performance drop straight away, and it stayed lower for at least five years. Those who moved with their team didn't drop.
Source: Groysberg, Lee & Nanda, 2008, Management Science. 1,052 top-ranked US equity analysts, 1988 to 1996.
- What it measures:
- Performance after a move
30% of outside executive hires fail to meet expectations within two years, against up to 21% of internal moves. 68% of failures were put down to people and leadership skills, 15% to technical skills.
Source: Institute of Executive Development and Alexcel Group, 2008. About 150 executives and talent staff, opinion survey. Sells coaching.
- What it measures:
- Respondents' view
Two years after an executive moves into a new role, 27% to 46% of moves are seen as failures or disappointments.
Source: McKinsey, May 2018. A summary of two earlier surveys: 27% from the IED/Alexcel study above, 46% from CEB.
- What it measures:
- Share of executive moves
CEOs brought in from outside are forced out more often: 29% of outside CEOs who left were forced out, against 19% of inside CEOs.
Source: Strategy& (PwC), 2015 study of the world's 2,500 largest listed companies. Seen only second-hand.
- What it measures:
- Share of CEO leavers
234 CEOs of major listed companies left their roles in 2025, up 16% on the year. Average CEO time in post fell to 7.1 years.
Source: Russell Reynolds Associates, CEO Turnover Index, early 2026.
- What it measures:
- Count of CEO exits
Professional sectors
US and Canadian law firms lost 19% of their associates in 2025. 83% of the associates who left had been at the firm less than five years.
Source: NALP Foundation, April 2026. 141 firms, 6,335 associate hires. The 83% is a share of leavers, not of hires. Much of it is people choosing to move.
- What it measures:
- Share of leavers
US sales development reps stay 1.9 years on average. 40% leave each year, 13% of them let go, and only 60% hit quota.
Source: The Bridge Group, Feb 2025. 351 B2B companies. A sales consultancy.
- What it measures:
- Share of reps
Fewer than half of US graduate hires who had no internship are still with the employer after a year. More than three-quarters of those who interned there stay.
Source: NACE, May 2026. 284 organisations. Older NACE figures were much higher, so check the definition before using.
- What it measures:
- Share of graduate hires
UK law: about 80% of newly qualified solicitors were kept on by the firm that trained them. Only 28% of graduate intake in law was still there after three years.
Source: Chambers Student Guide via Law Society Gazette (2011 cohort, large firms); ISE via Personnel Today, 2022 (small group of law firms). Neither is a failure rate. No SRA or Law Society figure exists.
- What it measures:
- Share of trainees / graduates
How many people apply for each job?
Applications per job more than doubled, from 116 in 2022 to 244 in 2025, while recruiters per company fell from 10.43 to 4.62. Time to fill went from 43.6 to 59.7 (almost certainly days).
Source: Greenhouse, The Hire Standard, March 2026. 640 million applications at 6,000+ companies, mostly North America. Greenhouse sells hiring software.
- What it measures:
- Per job
Applications per hire tripled from 2021 to 2024 and stayed above 300 through 2025. Candidates are about half as likely to reach interview as five years ago.
Source: Ashby, April 2026. 109 million applications, 247,000 jobs, tech-heavy. Per hire, so not comparable with the Greenhouse per-job number.
- What it measures:
- Per hire
UK graduate jobs got a record 140 applications each in 2024-25 (38 in 2002-03), and 290 in retail, FMCG and tourism.
Source: Institute of Student Employers, Oct 2025. 155 large employers, 1.8 million+ applications.
- What it measures:
- Per vacancy
About 11,000 applications a minute went through LinkedIn in June 2025, up more than 45% on a year earlier.
Source: LinkedIn data via the New York Times and Fortune, 2025. LinkedIn's own number, no method.
- What it measures:
- Platform count
58% of UK employers report a rise in unsuitable candidates applying.
Source: CIPD press release, Sept 2026.
- What it measures:
- Share of employers
42% of UK HR professionals think fewer than half the applications they get meet all the criteria. Nearly a quarter spend three to five hours a day reviewing applications.
Source: LinkedIn UK, Jan 2025, via HR Review. Sample not stated.
- What it measures:
- Share of HR people
Recruiters glance at a CV for an average of 7.4 seconds before deciding to keep or reject it.
Source: Ladders, 2018. A jobs site; sample never published. Measures the first glance only.
- What it measures:
- Vendor study
Is avoiding a bad hire the same as finding a high performer?
No. Avoiding a bad hire is one thing, and finding a great one is another. Most of the shape of performance is a long tail: a small group of people who add output faster and keep doing it. Those people show up most where the job gives freedom, is complex, and has no cap on what one person can produce.
In most jobs, performance isn't a bell curve. In three quarters of the jobs studied it's lopsided, with a long tail of top people. A bell curve fitted best only 4.8% of the time.
Source: Joo, Aguinis & Bradley, 2017, Journal of Applied Psychology. 229 samples of real output, from bank tellers and call centre staff to writers and athletes. A bell curve fitted 11 of 229; a lopsided shape with a long tail fitted 75%.
- What it measures:
- Share of samples
The extreme "winner takes all" shape fits only about 2% of jobs. Top people are mostly those who steadily build output faster than others, not lottery winners.
Source: Joo, Aguinis & Bradley, 2017, Journal of Applied Psychology. Of 229 samples: pure "winner takes all" 5 (2.18%), long tail with a limit 113 (49.34%), steady long tail 58 (25.33%), bell curve 11 (4.80%). The authors suggest an employer could heavily select for individuals with the highest levels of output accumulation rates.
- What it measures:
- Share of samples
Stars appear where the job has freedom, complexity and no ceiling on output. Where output is capped, like a checkout or an assembly line, people are closer together.
Source: Aguinis, O'Boyle, Gonzalez-Mulé & Joo, 2016, Personnel Psychology. 229 datasets, about 625,000 people, mostly US, in research, entertainment, politics, sports, sales and manufacturing. Countable output, not ratings.
- What it measures:
- Where stars appear
The other side of the argument: when performance is measured within one job, by ratings or capped output, it looks closer to a bell curve.
Source: Beck, Beatty & Sackett, 2014, Personnel Psychology. Both sides agree performance is lopsided in free, uncapped, complex work. They disagree on how much for ordinary jobs.
- What it measures:
- Counterpoint
The idea that a few stars produce a big share of output comes from changes in work: more knowledge work, more technology, more freedom.
Source: Aguinis & O'Boyle, 2014, Personnel Psychology. A theory paper with no data.
- What it measures:
- Theory only
How is job performance measured, and how well?
Two managers rating the same person agree only about half the time (.52 on a scale from 0 to 1).
Source: Viswesvaran, Ones & Schmidt, 1996, Journal of Applied Psychology. A review of studies, mostly US. The overall figure rests on 40 studies and 14,650 people, as reported by others. Some experts argue raters may fairly see different things (Murphy & DeShon, 2000).
- What it measures:
- How well two raters agree
More than half of a 360 rating reflects who is doing the rating (62% and 53% in two groups). Only 21% to 25% reflects the person being rated. The rest (11% to 18%) is random.
Source: Scullen, Mount & Goff, 2000, Journal of Applied Psychology. 4,492 US managers, each rated by 7 people. Development ratings, not annual pay reviews.
- What it measures:
- Share of a rating score
Manager ratings and hard results (sales, output) only partly agree: .39.
Source: Bommer et al., 1995, Personnel Psychology. A review of studies measuring the same people both ways.
- What it measures:
- How well two measures agree
How someone does when they know they're being watched isn't how they do on an ordinary day. The two barely matched.
Source: Sackett, Zedeck & Fogli, 1988, Journal of Applied Psychology. Supermarket cashiers at 12 chains, two samples (635 and 735). Speed and accuracy in a short, watched test against everyday till data. So interviews, tests and work samples show someone's best effort, not their usual effort. Shop-floor data; the step to professional jobs is reasoned, not measured.
- What it measures:
- Best effort vs usual effort
- Where:
- US
Deloitte stopped asking managers to score skills and started asking what they would do about each person. One of its four questions: would they always want this person on their team.
Source: Buckingham & Goodall, Harvard Business Review, April 2015. Deloitte found its old forms, meetings and ratings took close to 2 million hours a year. The new questions are asked at the end of every project, or every quarter. A company account, no results published.
- What it measures:
- One firm's method
The CIPD says avoid forced ranking, because it makes people feel the system is unfair.
Source: CIPD, Could do better?, Dec 2016, with the Center for Evidence-Based Management. UK evidence review. Its words: Avoid using forced or guided distribution to rate performance. On the shape of performance it says the argument seems far from settled.
- What it measures:
- Evidence review
- Where:
- UK
When firms removed ratings, performance fell by nearly 10% at most companies, a CEB survey suggested.
Source: CEB (now Gartner), 2016, nearly 10,000 employees in 18 countries, reported second-hand. Method not published. A USC research paper says it's the only study it found showing harm, and it was a one-off survey.
- What it measures:
- Employees' survey
What managers mean by a top performer
When a manager calls someone a top performer, they mostly mean "does the core job very well and causes no trouble". That explained 51% to 64% of the overall rating. Helping others counted for only 4% to 20%. And managers didn't agree on the recipe.
Source: Rotundo & Sackett, 2002, Journal of Applied Psychology. 504 managers in the US and Canada rating made-up employee profiles in five jobs, including accountant and nurse. Real ratings may differ.
- What it measures:
- Share of a rating score
Performance has two parts: doing the core job, and the extras (helping others, sticking at it, following the rules when no one is checking). Both count in a manager's overall view. Ability predicts the first more, personality the second.
Source: Borman & Motowidlo, 1993 and 1997; Motowidlo & Van Scotter, 1994, Journal of Applied Psychology, 421 US Air Force mechanics.
- What it measures:
- What performance is made of
Adapting is its own kind of performance: handling emergencies and stress, solving problems in new ways, learning new tasks and technology.
Source: Pulakos, Arad, Donovan & Plamondon, 2000, Journal of Applied Psychology. Eight kinds of adapting, built from over 1,000 real incidents across 21 jobs. Relevant to jobs that AI is changing.
- What it measures:
- What performance is made of
How do employers measure quality of hire?
The usual checks, in order: did they stay, is the hiring manager happy, what was their first review, how fast did they get up to speed. All four mostly show the person didn't fail. Only the review touches high performance, and it carries all the rating problems above. No large survey was found that measures quality of hire as output against colleagues.
Few employers check whether their hires turn out well. In the UK only 7% check whether their hiring methods actually predict who does well. In the US only 20% of organisations use any quality-of-hire measure, and that's falling.
Source: CIPD, Resourcing and Talent Planning, Sept 2026, 1,014 UK organisations. SHRM, 2025, 2,371 US members, random sample; 23% in 2017, 27% in 2022, 20% in 2025.
- What it measures:
- Share of employers / Share of organisations
- Where:
- UK / US
Three in ten UK organisations measure how new hires perform or whether they stay in their first year. Only 7% check whether their hiring methods predict who does well. 19% measure the return on their recruitment.
Source: CIPD, Resourcing and Talent Planning, Sept 2026. 1,014 UK organisations. Measuring return is mostly large private firms (37%). Of the 141 that measure it, 44% look at how new hires perform, 70% at cost per hire. In 2024, 31% tracked new hires' first-year performance; in 2022, 24%.
- What it measures:
- Share of employers
- Where:
- UK
Only 20% of US organisations use any quality-of-hire measure, down from 27% in 2022 and 23% in 2017. The most used are interviews about performance, review scores and whether people stayed.
Source: SHRM, 2025 Recruiting Executives Benchmarking. 2,371 members, random sample, Jan to Mar 2025, not weighted.
- What it measures:
- Share of organisations
- Where:
- US
89% of recruiting leaders say measuring quality of hire will matter more. Only 25% feel highly confident they can do it.
Source: LinkedIn, The Future of Recruiting 2025. 1,271 recruiting professionals at manager level or above, 23 countries, Sept 2024. 61% think AI can help measure it. LinkedIn sells recruiting tools.
- What it measures:
- Share of recruiters
- Where:
- Global
Recruiters call quality of hire the most valuable measure, but most check it by whether people stayed (57%), then manager satisfaction (42%), then new-hire reviews (40%). Only 33% feel they measure it well; 5% "best in class".
Source: LinkedIn, US Recruiting Trends 2016. 3,894 hiring decision makers in HR who opted in. Quality of hire was picked as the single most valuable measure by 32% in 2015 (39% in 2014). Old and from a vendor.
- What it measures:
- Share of recruiters
What predicts top performance?
The usual hiring scores above give one number per method across everyone. A method can separate bad from average without separating good from great. These sources look at the top end.
For thinking ability, more is better right to the top, often more so at the top than the bottom.
Source: Arneson, Sackett & Beatty, 2011, Psychological Science. 170,286 people in four datasets, including SAT scores and grades at 110 colleges and US Army job data. Earlier: Coward & Sackett, 1990, found the same. Job data are military.
- What it measures:
- How well a method predicts
Ability matters more in complex jobs. Its link to performance was roughly .58 in professional and management jobs against .23 in unskilled work.
Source: Hunter & Hunter, 1984, Psychological Bulletin; Schmidt & Hunter, 1998. Since cut back: Sackett et al. 2022 put ability at about .31 overall. Complex jobs also have the widest gaps between people, so picking well matters most there.
- What it measures:
- How well a method predicts
Being very conscientious and calm helps up to a point, then stops helping. In complex jobs that point comes later, so more is still worth more.
Source: Le, Oh, Robbins, Ilies, Holland & Westrick, 2011, Journal of Applied Psychology. Two US samples, one of 956 employees in 25 organisations, rated by their bosses. Carter et al. 2014 found mixed support using other methods.
- What it measures:
- Personality and performance
- Where:
- US
This year's performance is a good guide to next year's in the same job, but less so in complex jobs and on hard numbers.
Source: Sturman, Cheramie & Cashen, 2005, Journal of Applied Psychology. A review of repeated measures. Year-to-year match ranged from .85 down to .67; repeat-measure match from .83 (ratings, simple jobs) down to .50 (hard numbers, complex jobs). Past performance at another firm travels even less well: see the star analysts and outside hires above.
- What it measures:
- Performance over time
A test of the work itself scores about .33, but it shows someone at their best, not what they do every day.
Source: Sackett et al., 2022; Roth, Bobko & McFarland, 2005, Personnel Psychology, which cut the older .54 to about .33.
- What it measures:
- How well a method predicts
Only about 15% of people doing well in a role today have what it takes to succeed in a more senior one.
Source: CEB (now Gartner), 2013, reported by HR Review. Claimed sample 6.6 million people worldwide; method not published, original not found.
- What it measures:
- Share of high performers
For senior hires, look for motivation, curiosity, insight, engagement and determination, rather than a list of skills.
Source: Fernández-Aráoz, Harvard Business Review, June 2014. From an executive search firm (Egon Zehnder). Based on its own experience, no published test of it.
- What it measures:
- Practitioner view
What is a top performer worth?
In professional jobs, the top 1% produce roughly double the average. The gap between people is about 48% of average output in professional jobs, against 19% in routine jobs, and bigger still in sales.
Source: Hunter, Schmidt & Judiesch, 1990, Journal of Applied Psychology. A review of output studies. "Roughly double" is Speak to Kit's sum using the paper's own method. In life insurance sales the gap is 96.6% of average output.
- What it measures:
- Output gaps
Stars are real in sales. In one team of 219 sales reps, 5 earned over $500 an hour against an average of $124. That's 20 times more often than a bell curve allows.
Source: Aguinis, O'Boyle, Gonzalez-Mulé & Joo, 2016, Personnel Psychology. A worked example inside a study of 229 datasets and about 625,000 people. Revenue per hour, not manager ratings.
- What it measures:
- Sales revenue per hour
- Where:
- US
The gap between people in the same job is about 48% of average output in professional jobs, 32% in mid-level skilled jobs and 19% in routine jobs. In sales it's bigger: 96.6% in life insurance sales, 76.3% for sales account managers.
Source: Hunter, Schmidt & Judiesch, 1990, Journal of Applied Psychology. Professional figures come from doctors, dentists, lawyers and map technicians. Short sales periods overstate the gap: over 4 weeks, sales figures weren't very steady (.23 life insurance, .39 other sales).
- What it measures:
- Output gaps
As a floor, a better-than-average person is worth at least 40% of salary a year more. In a job paying £60,000, someone in about the top sixth is worth at least £24,000 a year more than an average one.
Source: Schmidt & Hunter, 1983 and 1998, Journal of Applied Psychology and Psychological Bulletin. The spread is at minimum 40% of the mean salary. A rule of thumb, not a measured figure for any one job.
- What it measures:
- Value per person, minimum
One year of hiring US government programmers with an aptitude test was worth between $5.6 million and $97.2 million to the government, depending on assumptions.
Source: Schmidt, Hunter, McKenzie & Muldrow, 1979, Journal of Applied Psychology. 105 of 147 supervisors in 10 agencies estimated what programmers at different levels were worth. A top-sixth programmer was worth $10,871 a year more than an average one. Supervisors' estimates, 1979 dollars.
- What it measures:
- Method / Supervisors' estimate
- Where:
- US
The real gap between programmers is far smaller than the "10x" story. Only half the difference in effort came from the person; half the group sat within a factor of 2 of each other.
Source: Nichols, Carnegie Mellon Software Engineering Institute, Jan 2020 (and IEEE Software, 2019). 494 students on a training course, mostly working professionals, 10 set exercises, 2000 to 2006. His verdict: the gap has been greatly exaggerated.
- What it measures:
- Programmer effort
Senior executives estimated a high performer adds 40% more productivity in operations, 49% more profit in general management and 67% more revenue in sales.
Source: McKinsey, War for Talent, 2001. 410 corporate officers at 35 large US companies. What executives believe, not a measurement, though close to the measured 48% above.
- What it measures:
- Executives' estimate
What does the UK evidence say about management and performance?
Better-managed GB manufacturers are more productive: a 0.1 rise in the management score goes with 8.6% higher output per worker, or 6.7% after allowing for size and industry.
Source: ONS, Management Practices Survey, 2015 data. The score covers targets, tracking results, and hiring, promoting and dealing with poor performance on merit. The ONS calls the sample relatively small and says it can't prove cause.
- What it measures:
- A link, not proof
- Where:
- UK official
Across all sectors, the average GB management score was 0.58 in 2019 and 0.60 in 2020. Better management still went with higher productivity after allowing for many other factors.
Source: ONS, Management practices in Great Britain: 2016 to 2020, 2021. The rise since 2016 is mainly because many poorly run very small firms closed. The ONS points to a trial in India where textile firms given free management help raised productivity 17% in the first year (Bloom et al. 2013).
- What it measures:
- A link, not proof
- Where:
- UK official
The UK's professional body for HR says drop forced ranking, and that the shape of performance isn't settled.
Source: CIPD, Could do better?, 2016. Evidence review.
- What it measures:
- Evidence review
Over two thirds of UK employers never check whether the people they hired went on to perform. Only 7 in 100 check whether their interviews and tests predict who does well.
Source: CIPD, 2026, 1,014 UK organisations.
- What it measures:
- Share of employers
What nobody has measured yet
These are the gaps we found while compiling this page.
- No official UK figure for failed hires, probation failure or dismissal in the first two years. The government said so in Oct 2024.
- No peer-reviewed general rate of new hires failing in the US either. The best official figures (Bureau of Labor Statistics) count every job ending, for any reason.
- No first-year leaver rate with a published method for UK or US Big Four, consulting, MBA hires or software engineers.
- No figure from the bodies that run UK law and accountancy (SRA, Law Society, ICAEW, ACCA).
- No real study of what an empty role costs per day. Every number out there is a vendor calculator.
- Most employers don't measure it themselves: only three in ten UK employers track how new hires do in their first year (CIPD 2026).
- No large study measures quality of hire as output against colleagues. Every big survey measures staying, manager satisfaction or first review.
- No UK study of how performance is spread in professional jobs. Nearly all of this data is from the US.
- Lawyers: no study read on the spread of billing or client value.
Common questions
How many hires fail?
No official UK figure exists: the Department for Business and Trade said in October 2024 that it doesn't have robust data on dismissal in the first two years. Published surveys count different things. CEB (now Gartner), 2013, as cited by the REC in 2017, found more than one in five hires are bad or regretted decisions (a share of hires, worldwide). Leadership IQ, 2005, a US training firm surveying 5,247 hiring managers, found 46% of new hires failed within 18 months, as judged by their own managers. CIPD, 2024, found 41% of UK employers who hired in the past year said new recruits at least sometimes quit within 12 weeks. That's a share of employers, not a quit rate, and the 2026 edition puts it at 29%.
What does a bad hire cost in the UK?
There's no single measured figure. The REC, in 2017, built one worked example: a bad hire of a middle manager on £42,000 could cost £132,015, over three times salary, or £78,015 without a colleague quitting. It's an example, not an average. Oxford Economics for Unum, 2014, found replacing a member of staff on £25,000 or more costs £30,614 on average, across any leaver and in 2014 money. CIPD, 2026, found the visible cost of filling a role is a median £3,500 for senior managers and £2,000 for other staff, which leaves out training, getting up to speed and the cost of a bad hire.
What predicts top performance?
A structured interview (same questions, scored the same way) scores .42 for predicting job performance, against .19 for an unstructured chat, on a scale from 0 to 1 (Sackett, Zhang, Berry and Lievens, 2022). Years of experience scores .06 (Van Iddekinge and others, 2019). For thinking ability, more is better right to the top, with no point where it stops paying (Arneson, Sackett and Beatty, 2011). Past performance at another firm travels less well: star analysts who moved to a new firm saw their performance drop and stay lower for at least five years (Groysberg, Lee and Nanda, 2008).
How do employers measure quality of hire?
The usual checks are whether the person stayed, whether the hiring manager is happy, their first review, and how fast they got up to speed. Three in ten UK organisations measure how new hires perform or whether they stay in their first year, and only 7% check whether their hiring methods predict who does well (CIPD, 2026). In the US, 20% of organisations use any quality-of-hire measure, down from 27% in 2022 (SHRM, 2025). 89% of recruiting leaders say measuring it will matter more, and only 25% feel highly confident they can do it (LinkedIn, 2025). No large survey was found that measures quality of hire as output against colleagues.
How much better is a top performer than an average one?
The gap between people in the same job is about 48% of average output in professional jobs, 32% in mid-level skilled jobs and 19% in routine jobs, and bigger in sales (Hunter, Schmidt and Judiesch, 1990). In most jobs performance isn't a bell curve: in three quarters of 229 samples of real output it's lopsided with a long tail of top people, and a bell curve fitted 4.8% of the time (Joo, Aguinis and Bradley, 2017). In the UK, a 0.1 rise in the official management score goes with 8.6% higher output per worker in manufacturing (ONS, 2015 data). The ONS says this is a link, not proof of cause.