The numerical reasoning test: what the score compares you with

0 The short answer

An employer's numerical reasoning test is a timed ability test of reading and calculating with figures, tables and charts, and it reports a percentile against a comparison group of earlier test takers, not a number that says how well you reason in general. SHL describes its Verify Interactive version as at most 10 questions in 18 minutes, and Korn Ferry says its cognitive tests adapt to your answers; the research on retesting says a second attempt adds about a quarter of a standard deviation on average, and more when coaching is involved. ACIS has no table and chart interpretation task and is not preparation for, or a predictor of, any employer test, but it can show your own quantitative profile on adult norms: the Full Scale form is 50 dollars for about 175 minutes and reports a Quantitative Reasoning Index, and a Custom Test of Arithmetic plus Mathematical Achievement is 10 dollars (prices read October 29, 2026, and they can change). The decisive limit is simple: only the employer's own test, scored against the employer's comparison group, answers the employer's question, and the page on quantitative reasoning as an ability explains what that ability is.

1 What Is an Employer's Numerical Reasoning Test?

A numerical reasoning test is an ability test that an employer buys from a publisher and gives to applicants to see how well they reason with numerical data, and it is one narrow sample of ability, not an IQ test. The publishers describe it in similar words. SHL's fact sheet for its Verify Interactive version says the test measures "a candidate's ability to make correct decisions or inferences from numerical or statistical data," in a realistic workplace context, for mid-level positions in fields such as finance, IT, professional services, production and sales (fact sheet with a 2022 copyright line, read October 29, 2026). Korn Ferry's candidate guide (2021, read the same day) lists numerical reasoning beside verbal reasoning, logical reasoning and checking skills, and defines the numerical test as a measure of the ability to analyze and draw inferences from numerical information and data, important for roles where working with data is key. Saville Assessment's aptitude page, read on October 29, 2026, says numerical reasoning shows how well someone interprets figures and spots what the numbers are saying.

Three words in those descriptions carry the weight. The first is "employer": the employer chooses the test, the time limit, the calculator rule and the cut score, so two applicants who both say they took "a numerical reasoning test" may have taken different instruments. The second is "data": the material is tables, charts and short passages of figures, and the publishers describe the answer as a decision or an inference drawn from them. The third is "reasoning": the publishers speak of inferences and decisions, not of computation alone, and the CHC distinction between reasoning with quantities and knowing mathematics is why the two can diverge in one person.

This page is the candidate's guide, written for an adult who will sit, or has sat, such a test. It is not an employer's guide, and it does not teach any test. The wider question of why employers screen with cognitive tests at all, and how the main screening tests differ, belongs to the page on the pre-employment cognitive test. Four neighbors in this group cover single publishers or sibling tests of the numerical test: the page on the SHL test, the page on the CCAT test, the page on the verbal reasoning test, and the page on the error checking test, which is what Korn Ferry calls checking skills, a measure of the ability to detect errors in data quickly and accurately.

2 What Does the Format Look Like in the Publishers' Own Words?

The format is a short, timed set of multiple choice or interactive questions about figures, and the three publishers read for this page differ on length, on adaptivity and on how much they say. The table gives what each page states and says "not stated" where it is silent, because a guessed time would be worse than a blank. The employer's invitation, not this table, sets the rules for any real attempt.

Two details in the table matter to a candidate. First, adaptive means the questions you see depend on your earlier answers. Korn Ferry's guide explains that you begin with a question of average difficulty, a correct answer within the time limit leads to a harder question, and an incorrect or unanswered one leads to an easier question, so the guide tells candidates not to worry if the questions feel difficult. SHL's sample report says the same thing in a different way: because each candidate receives a different set of items, there is no direct correlation between the share of questions answered correctly and the percentile achieved, and a person with a lower hit rate may reach a higher percentile. A count of how many you got right therefore does not tell you your result.

Second, versions differ. SHL's fact sheet describes the newer interactive version, and its practice tests page offers both an interactive and a standard multiple choice practice test, so an applicant may meet either style. Other pages on the web give other lengths, from other versions and publishers, and this page does not use them. Read the instructions you were sent.

3 Which Abilities Does a Numerical Reasoning Test Sample?

A numerical reasoning test samples more than arithmetic: in CHC terms it draws on quantitative knowledge (Gq), quantitative reasoning within fluid reasoning (Gf), the ability to hold intermediate values in mind (Gwm), the language needed to read the question (Gc) and the speed the clock rewards (Gs). That mapping is our reading, not a claim by any publisher, and the CHC model page explains the framework behind the names. The publishers describe the task, not the factor structure, and a single test never isolates one ability cleanly.

Start with what the sources say. SHL's fact sheet gives analysis and interpretation of numerical data, for example budgets and costs, trends and margins, as sample tasks. Our reading is that this is mostly applied knowledge of percentages, ratios and trends, which is the territory of quantitative knowledge. The same fact sheet and Korn Ferry's guide stress inference and decisions, which means seeing which quantity relates to which and which operation answers the question. That is quantitative reasoning, a narrow ability that CHC files under fluid reasoning, and the ACIS page on quantitative reasoning tests, linked in the short answer, covers how it differs from acquired mathematical knowledge.

Hausknecht and colleagues (2007), whose retest meta-analysis appears below, describe tests of quantitative ability as measuring mathematics skills and typically including items that involve arithmetic computation, mathematical reasoning or other numerical problem solving. That is a useful plain definition of the category: some of the score is what you learned, some is what you can work out on the spot.

The other abilities enter through the conditions. Holding a figure in mind while you find the next one loads working memory, which is why the ACIS page on mental math argues that part of a calculation test is memory. Reading the labels, units and wording of a question draws on verbal comprehension, which is part of why the page on verbal versus quantitative ability matters to anyone whose two profiles differ. And a time limit adds processing speed, the subject of the next section. How strongly mathematics and reasoning are linked in the research is the subject of the page on math ability and IQ, and the page on a math IQ test separates the two for someone choosing a test of their own.

A careful reader should hold one more point. Because the tests are short, the score is an estimate built from few items. The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) say that, in general, if an assessment is shortened the reliability is likely to decrease, and in an adaptive test the item selection rules may be designed partly to increase the precision of the score estimate. The page on reliability and validity explains in general terms what a reliability coefficient does and does not say about one person's score.

4 Why Does the Clock Matter Alongside the Arithmetic?

The clock matters because a timed numerical test scores how much correct work you finish in the time allowed, so speed and accuracy are measured together, and the publishers tell candidates to work both quickly and accurately. Korn Ferry's candidate guide (2021, read October 29, 2026) says exactly that: read each question and the response options carefully, keep a good pace, and remember that procrastinating "can waste time." The same guide notes that its tests have a time limit for each question and that the best strategy is to use the full time given and not rush into an unconsidered answer, so the instruction is to be neither hurried nor slow. SHL's fact sheet gives 18 minutes for at most 10 questions, which works out to an average ceiling of 1.8 minutes a question (our arithmetic), and SHL's candidate support page warns that if you step away from a timed test the timer keeps running.

SHL's sample report shows how the two ideas are separated in a result. It defines work rate as the number of questions attempted divided by the total number in the test, a measure of how far the candidate got, and hit rate as the number answered correctly divided by the number attempted, a measure of accuracy. A report that lists both is saying that finishing and being right are different things, and the percentile blends them in a way the report does not break apart. The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) define speededness in their glossary as the extent to which scores depend on the rate at which work is performed as well as on the correctness of the responses. A numerical reasoning test is speeded in that sense, and a score from it is not a pure measure of reasoning.

Calculators change what the clock rewards. Korn Ferry advises candidates to have a calculator to hand, and SHL's support page says a calculator may be used for a numerical or calculation test unless you have been instructed otherwise; both leave the final rule to the employer's instructions. Where a calculator is allowed, our reading is that the work left is choosing the right operation and reading the display correctly under time pressure, which is more reasoning and less arithmetic. Where it is not allowed, mental calculation takes a larger share, and the page on a mental math test explains how much of that task is working memory.

Speed is also an ability of its own in the CHC model. ACIS reports Processing Speed (Gs) as an index of its own, built from Symbol Search and Coding, which are simple timed tasks, and a candidate who is slow on a data interpretation test is not necessarily weak at numerical reasoning, because the clock folds several things into one number. The page on why IQ tests are timed covers what a time limit adds to a score, and the page on a processing speed test shows what a purely speeded task looks like. No source read for this page says how an employer test weights speed against accuracy, so any claim of a fixed ratio would be invented.

5 What Does a Percentile Against a Comparison Group Mean?

A percentile on an employer test tells you what share of a named comparison group you scored above, so the same performance can earn different percentiles against different groups, and the first question to ask is who is in the group. Korn Ferry's guide says your responses for each ability test are compared with the scores of a comparison group, people who have previously completed the tests, and that the result is given as a percentile; a score in the 75th percentile means you did better than 75 percent of the sample. SHL's sample report, dated September 27, 2018 and read on October 29, 2026, says the same about its scale: a percentile of 50 means the individual performed better than 50 percent of the people in the comparison group.

The sample report is useful because it shows two separate labels on one result. The numerical reasoning result carries a test level, "Graduate/University," and a comparison group, "General Population" with a year attached. The level names the version of the test; the comparison group names whom the percentile is measured against. The sample does not describe how either group was built, so a candidate who wants to read a percentile properly needs the employer or the publisher to say which group applies. The Standards (Standard 5.8) ask that norms refer to clearly described populations that include the people with whom test users will ordinarily wish to compare their examinees, and the APA Guidelines for Psychological Assessment and Evaluation (2020, Guideline 11) note that an individual may be poorly represented in a norm sample, which limits what a comparison tells you about them. The page on how IQ scores are normed explains how norm groups work in general.

The sample report also lists scales other than the percentile, which it says are provided for users trained in their interpretation. The table sets them beside the ACIS scales so that a number from one report is not read as if it were another. The last column is our arithmetic from the normal curve, valid only if scores are roughly normally distributed, and it describes position on a scale, not any particular test.

Percentiles and standard scores stretch differently, so a few percentile points at the top of a distribution cover more standard deviation units than the same points near the middle. The page on IQ score versus percentile explains the stretching, and the page on why the standard deviation is 15 explains the IQ scale. For a numerical test the practical consequence is that the 50th and 60th percentiles are neighbors, while the 90th and 99th are not, so a cut score placed near the top asks more of a candidate than its distance in percentile points suggests.

Who sets that cut score? The employer. The Standards describe an employer interviewing everyone above a given level on an employment test and say that the particular value of the cut score may be largely determined by the number of persons to be interviewed or further screened. They add that the likelihood of misclassification is generally relatively high for people whose scores are close to a cut score. A percentile just under a cut score is therefore a thin reason to conclude anything about ability: it records where one attempt landed against one group on one day.

6 How Well Does a Numerical Test Predict Job Performance?

Ability tests of this kind predict job performance on average, but the size of that prediction was revised downward in 2022, and an average across many people never tells an employer or a candidate what one person will do. Korn Ferry's guide, written for candidates in 2021, says ability tests "have been found to be the strongest predictor of future job performance." That is the publisher's statement. The research literature gives a more qualified picture, and the two are worth reading as a pair.

Bertua, Anderson and Salgado (2005) published a UK meta-analysis of tests of general mental ability and of specific abilities (verbal, numerical, perceptual and spatial) in the Journal of Occupational and Organizational Psychology. They drew on 283 independent samples for job performance (13,262 people) and 223 for training success (75,311 people), and reported operational validities in the magnitude of .5 to .6, higher in occupations with greater job complexity. The abstract does not give a separate figure for numerical tests and we did not open the full text, so we quote none.

Sackett, Zhang, Berry and Lievens (2022), in the Journal of Applied Psychology, argued that corrections for restriction of range had often overcorrected, so that the validity of many selection procedures had been substantially overestimated, and reported that mean validity estimates fell by .10 to .20 points while most procedures kept their rank. Their follow up in Industrial and Organizational Psychology (Sackett and colleagues, 2023) tabulates the change for general mental ability tests: the 1998 estimate of Schmidt and Hunter, .51, became .31, and cognitive ability, the paper says, is no longer the stand out predictor, with structured interviews now highest. The same paper reports that a 21st century meta-analysis of 113 studies found a mean observed validity of .16 and a corrected value of .23, and it stresses that the 2022 findings reflected what was known at the time and are subject to updating.

Two cautions keep the numbers in proportion. First, these are validities for general mental ability tests as a class, not for one numerical test or publisher, and a validity coefficient describes a group. Squaring .31 gives about .10, which is our arithmetic: on that estimate about a tenth of the variation in performance is shared with the test score. Second, a prediction is not a verdict. The Standards (Standard 11.4) warn, in the context of employment testing, that viewing a high score as indicating overall job suitability would be an inappropriate inference from a test of a single narrow domain, and that the inferences drawn from a score should be limited to those for which validity evidence is available. A numerical percentile is one input an employer may combine with interviews and other evidence.

For a candidate the useful reading is modest. Numerical tests are used because, across many people, scores relate to later performance in jobs that involve data; they do not show that a given person will or will not succeed. The ACIS page on IQ and job performance takes up the broader evidence.

7 What Does Retest Research Show About Practice?

Retest research shows that scores on cognitive ability tests rise by roughly a quarter to a third of a standard deviation on a second attempt, that the rise is larger with coaching or an identical form, and that it levels off after a few attempts. Two meta-analyses are the main evidence, and the table sets out what each reports.

Hausknecht and colleagues describe the .26 as scores improving about one quarter of a standard deviation across the first and second administrations, and report an adjusted .20 from the second to the third attempt and .56 across non consecutive first to third attempts. They also compared content: practice effects were .19 for verbal, .30 for quantitative and .32 for analytical ability, but the confidence intervals overlapped and the differences were not statistically significant, so the study does not show that numerical tests are more coachable than verbal ones. The coaching figure rests on 2,323 participants, a small base beside the 81,374 in the studies without coaching, and the authors say additional research is needed on how retesting affects the validity of the inferences drawn from scores.

Scharfen and colleagues add that numerical tasks showed smaller first retest gains than verbal tasks (0.18 against 0.33) and a larger gain from the second to the third attempt (0.35), which they discuss as possibly reflecting how familiar and how memorable the content is. They found no study of numerical tests with more than three administrations, so these data say nothing about a fourth attempt. The retest gain also shrank slightly with the time between attempts, by 0.0008 standard deviations per week, which the authors call a weak effect, and the spread of true effects was wide: for the overall first retest, their own prediction interval ran from about minus 0.30 to 0.96. An average gain is not a promise to any one person.

To see what a quarter of a standard deviation means for a percentile, assume a normal distribution (our arithmetic). A gain of 0.26 moves a person at the 50th percentile to about the 60th, and a person at the 84th to about the 90th. A gain of 0.33 moves the person at the 50th to about the 63rd. These are group averages from studies that included coaching, research settings and operational settings, and they describe test takers who retook the same kind of test, not a particular candidate who studies for a week.

What is practice for? The meta-analyses cannot say whether a gain reflects a better measure of ability or only familiarity with a format. Korn Ferry's guide takes the practical view that practicing the kinds of questions you may face is the best way to maximize performance and boost confidence, and it points candidates to free practice assessments while cautioning that third party practice sites that charge for test content do not provide recognized content and may not store data securely. SHL and Saville also publish practice material of their own (read October 29, 2026). The Standards note that some programs give all test takers study guides or sample questions, and that practice materials should have the same appearance and format as the actual test. The same chapter lists among the responsibilities of test takers representing themselves honestly and protecting the security of the test materials.

Retaking an employer test is a separate matter from retaking a test of your own. SHL's candidate support pages tell candidates to ask their recruiter about a reattempt and describe a cooling period in which recruiters can restrict multiple attempts, and the Standards say retesting policies should be set by the test developer or user and communicated in advance. The page on whether to retake an IQ test deals with the self-scored version of the question.

A last word on the searches that bring people here. This page does not reproduce practice questions, answers or a pdf of test items. Items from a live employer test are the employer's secure material, the publishers offer free official practice in their own formats, and the research above is about what practice does, which no list of answers can show. If you want examples, the publisher pages named in the Sources section are the legitimate place to start.

8 What Should You Do Before and During an Employer's Test?

Before an employer's test, read your invitation for the time limit, calculator rule and retake policy, use the publisher's own practice page, and set up a quiet room with a working device, because the employer decides the rules and the timer does not wait. Korn Ferry's guide (2021) asks candidates to be rested and unhurried, to sit in a quiet place free of interruptions for the whole assessment, to check their computer and connection, to have paper, a pen or pencil and a calculator to hand, and to read the online instructions carefully. It adds that the tests work independently of connection speed, so a slow connection will not affect the test itself.

SHL's candidate support page (read October 29, 2026) adds practical rules. If a test freezes, refresh the page and do not click the Exit button, which can submit your responses. If technical issues interrupt a test, you need to speak to the company running it, because SHL support teams cannot reset a completed test without authorization. Tests have a restart window, and an expired test is submitted automatically. None of this is a trick: it is the difference between a score that reflects the numerical task and one that reflects a lost connection.

One more fact from SHL's sample report matters for anyone who took a test at home. The report says that if the tests were unsupervised, there is a small possibility that the scores do not represent the actual level of ability, and it recommends a verification test. The Standards (Standard 3.4, comment) make a similar point about unproctored administrations, where standardization cannot be ensured: results should be interpreted with caution. A candidate who did well unsupervised may be asked to confirm that score under supervision, so the honest strategy is to take the first test the way you would take the second.

The Standards also say that test takers should be given, in advance, as much information about the test, its intended use, its scoring and its accommodation policies as is consistent with valid results, including any time limits. If something is unclear, ask the recruiter before the attempt, not after. Nothing in this section is preparation in the sense of rehearsing a particular test, and nothing on this page is.

9 What Are Your Real Options for Practice and for Seeing Your Own Profile?

There are five real routes, and each answers a different question: the employer's test answers the employer's question, official practice builds familiarity with a format, and only a psychologist or a normed battery describes your own abilities. The table uses only what was readable on the pages named, read on October 29, 2026.

The first row is the only one that answers the question the employer is asking. A psychologist is the route when documentation or an examiner's observations matter, and Pearson's WAIS-5 product page (read October 29, 2026) lists it for ages 16:0 to 90:11, published in 2024, with the administration times above. A free quiz can satisfy curiosity if its page names whom it compares you with, and the pages on free IQ tests and on where to take an IQ test compare the routes in general. The page on IQ tests for adults describes the adult category. Whichever you choose, none of the other rows is a substitute for the first.

10 How Does ACIS Cover the Quantitative Abilities?

ACIS has no table and chart interpretation task, so it cannot show how you would do on an employer's data interpretation test; what it offers is adult normed measures of quantitative knowledge and quantitative reasoning, the abilities such a test samples. The page does not describe any ACIS subtest beyond what its own page and the technical manual say, and the timing of individual subtests is not stated there.

Arithmetic and Mathematical Achievement together form the Quantitative Reasoning Index (QRI), and the page on the Quantitative Reasoning Index explains why it is an ancillary index. Figure Weights and Visual Number Series sample quantitative reasoning inside the Fluid Reasoning Index. The technical manual's composite CQRI combines "Mathematical Achievement, Figure Weights, Arithmetic, and Visual Number Series." A Custom Test with Arithmetic and Mathematical Achievement costs 10 dollars at 5 dollars a subtest, and the Custom Test returns a full index score for any domain you complete and a scaled score for every subtest. A Full Scale IQ needs the Full Scale battery.

What a reader can do with such a profile is read it against itself. Whether the two knowledge based subtests sit above or below the two reasoning based ones is the question the page on IQ tests that leave math out helps frame. None of this predicts a result on an employer's test, prepares for one, or converts to one.

11 What Can a Numerical Reasoning Score Not Tell You?

A numerical reasoning score cannot tell you your IQ, your general ability, your fitness for a job or whether you are good at mathematics, because it is a short, timed sample of one narrow ability, compared with one group, on one day. Several limits follow from the sources above.

It is a group comparison. The same performance can earn different percentiles against different comparison groups, and the sample report separates the test level from the comparison group. It is an estimate from few items: the Standards say shortened assessments are likely to be less reliable, and an adaptive test gives each candidate a different set of questions. It is speeded, so it blends accuracy and pace, and it reflects conditions: the APA Guidelines for Psychological Assessment and Evaluation (2020, Guideline 8) say examinees may underperform for many reasons and that standardized administration supports valid interpretation of norms and other score based inferences. Unsupervised conditions add a caution, as shown above.

It is also a prediction at the level of groups. A mean validity of .31 in the revised estimates does not say what one person will do, and the research on retesting says candidates tested again score higher on average. A modest percentile on a short timed test says little about how someone works with numbers when they can take time, check a source or use a tool, which the test does not sample. It is also not a clinical result. Group statistics are never individual diagnoses, and a person worried about difficulties with mathematics should speak to a qualified professional, not infer anything from a hiring screen.

12 Where ACIS Sits, and What a Buyer Should Do

ACIS is a normed online battery for adults curious about their own cognitive profile, with real strengths and plain limits, and it is not preparation for, a predictor of or a substitute for any employer's numerical reasoning test. The strengths are concrete: a Full Scale IQ and six primary indices on adult norms for ages 16 to 90, scaled subtest scores, percentiles and a 95 percent confidence interval, with breaks allowed, progress saved between subtests and 30 days to finish. Quick is 15 dollars for 6 subtests (about 45 minutes), Optimized is 30 dollars for 13 subtests (about 110 minutes) and Full Scale is 50 dollars for all 20 subtests (about 175 minutes), with a Custom Test at 5 dollars a subtest (prices read October 29, 2026; they can change). One free trial, no card, shows the format. The technical manual documents how the battery is built and normed, and this page states nothing about that beyond what the report shows.

The limits are as concrete. ACIS is online and unsupervised, not a clinical or diagnostic instrument, English only, and its research page says it "is not intended for clinical diagnosis, hiring decisions, accommodation determinations or admission to high IQ societies" (read October 29, 2026). The technical manual says verbal and knowledge tasks are affected by English proficiency and educational exposure, and a knowledge based subtest such as Mathematical Achievement is partly a measure of schooling. The refund policy, quoted literally, is: "Full refund within 5 days if the report does not deliver the features described at checkout or a technical issue prevents you from accessing your results." That is a conditional guarantee, not an unconditional one.

The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) are the yardstick this page tries to meet, not an endorsement of ACIS. Standard 1.1 asks that the intended interpretation, population and construct be clearly set out and says each recommended interpretation for a given use requires validation; Standard 1.4 puts the burden on a user who applies a score to an unvalidated use; Standard 5.8 asks that norms refer to clearly described populations; and the comment to Standard 3.4 urges caution with unproctored administrations. The APA Guidelines for Psychological Assessment and Evaluation (2020) ask psychologists to recognize when a test is used for a purpose different from the one intended (Guideline 4) and to attend to norms, effort and standardized administration (Guideline 8). Read that way, a numerical percentile from an employer answers the employer's question, and an ACIS profile answers a personal one.

What a buyer should do, in order:

1. Decide the question. If it is the employer's test, use the invitation and the publisher's free practice pages. If it is documentation, start with a licensed psychologist. If it is curiosity about your own quantitative profile, continue.
2. Pick the form. Full Scale for a Quantitative Reasoning Index and all six indices, or a 10 dollar Custom Test of Arithmetic and Mathematical Achievement if only that pair interests you.
3. Start the free trial. It needs no card and shows the format before anything is spent.
4. Set up the room. Quiet, working sound for Arithmetic, a rested start and a plan for breaks.
5. Read the report as one source. Keep the confidence interval in view, compare indices with one another, and do not use the number for a decision it cannot support.

13 Sources Behind This Page

All web pages and documents below were opened on October 29, 2026, and the journal articles were checked against Crossref records.

1. AERA, APA and NCME (2014). Standards for Educational and Psychological Testing. American Educational Research Association. Open access edition: testingstandards.net. Standards 1.1, 1.4, 2.14 (reliability and cut scores), 3.4 (comment), 5.8, 11.4, the chapters on administration and on test takers' rights, and the glossary entry for speededness.
2. American Psychological Association, APA Task Force on Psychological Assessment and Evaluation Guidelines (2020). APA Guidelines for Psychological Assessment and Evaluation, Guidelines 4, 8 and 11. apa.org.
3. Bertua, C., Anderson, N., and Salgado, J. F. (2005). The predictive validity of cognitive ability tests: A UK meta-analysis. Journal of Occupational and Organizational Psychology, 78(3), 387-409. doi.org/10.1348/096317905X26994.
4. Hausknecht, J. P., Halpert, J. A., Di Paolo, N. T., and Moriarty Gerrard, M. O. (2007). Retesting in selection: A meta-analysis of coaching and practice effects for tests of cognitive ability. Journal of Applied Psychology, 92(2), 373-385. doi.org/10.1037/0021-9010.92.2.373.
5. Korn Ferry (2021). Candidate guide: Psychometric assessments. kornferry.com.
6. Pearson Assessments. Wechsler Adult Intelligence Scale, Fifth Edition (WAIS-5), product page. pearsonassessments.com.
7. Sackett, P. R., Zhang, C., Berry, C. M., and Lievens, F. (2022). Revisiting meta-analytic estimates of validity in personnel selection: Addressing systematic overcorrection for restriction of range. Journal of Applied Psychology, 107(11), 2040-2068. doi.org/10.1037/apl0000994.
8. Sackett, P. R., Zhang, C., Berry, C. M., and Lievens, F. (2023). Revisiting the design of selection systems in light of new findings regarding the validity of widely used predictors. Industrial and Organizational Psychology, 16(3), 283-300. doi.org/10.1017/iop.2023.24.
9. Saville Assessment. Aptitude tests, product page and FAQ. savilleassessment.com.
10. Scharfen, J., Peters, J. M., and Holling, H. (2018). Retest effects in cognitive ability tests: A meta-analysis. Intelligence, 67, 44-66. doi.org/10.1016/j.intell.2018.01.003.
11. SHL. Verify Interactive Numerical Reasoning: assessment fact sheet (2022 copyright line). shl.com.
12. SHL. Verify Ability Test Report, sample report dated September 27, 2018. shl.com.
13. SHL. Practice tests and assessments, candidate page. shl.com.
14. SHL. Candidate support: Assessments, timing, submitting and general issues. support.shl.com.

14 Frequently Asked Questions