The Pre-Employment Assessment Decoder, Test by Test

You were told to complete an assessment before anyone will interview you, and the invite does not say what it measures, whether you can prepare, or what counts as passing. This guide is a lookup keyed to the test in front of you: one row per assessment family, stating what it actually measures, whether practice moves the score, how the result is gated, and the specific way each type misreads a real candidate. Read the section that matches your invite, then leave.

Most pages about pre-employment testing are written for the employer deciding whether to buy the software. They explain business return on investment. This one is written for the person holding the link.

What the major pre-employment assessment test types measure

There are seven families, and the first thing to do is name yours. Each measures a different thing, and the label on the invite tells you which prep is worth your hours and which is wasted.

• Cognitive ability tests measure problem-solving, reasoning, and learning speed. Example: the Criteria Cognitive Aptitude Test (CCAT).
• Personality and behavioral assessments measure traits like conscientiousness and how you work with others. Examples: the SHL Occupational Personality Questionnaire, the HEXACO Personality Inventory, and Hogan's tools.
• Hard-skills tests measure job-specific competence such as data analysis, financial reasoning, and computer literacy.
• Coding assessments measure whether your code passes test cases. Examples: HackerRank and Codility.
• Situational judgment tests (SJTs) measure your judgment in realistic workplace scenarios.
• Integrity tests measure counterproductive tendencies, and they come in two very different flavors covered below.
• AI-fluency assessments measure how effectively you work with AI tools in a simulation of real tasks.

A note on one instrument you may see: the Myers-Briggs Type Indicator. Its own publisher states it should not be used for selection. If an employer is gating a hire on an MBTI type, that is a signal about the employer, not about you.

The vendor names cluster in predictable ways. Codility and HackerRank do coding. Hogan does personality, values, and judgment. Predictive Index does behavioral and cognitive. Criteria Corp does cognitive, personality, and skills. Knowing the vendor narrows the family before you read a single question.

Where the assessment sits in the funnel, and how much time you get

Assessments front-load the hiring funnel. They are the filter that runs before a human looks at your resume, which is exactly why they feel impersonal: at this stage, they are.

Time limits vary sharply by family, and the limit is itself a signal about what is being tested. A test with 15 minutes for 50 questions is testing throughput. A test with two hours for two problems is testing correctness and efficiency.

A HackerRank test typically runs 60 to 120 minutes depending on the employer and the number of questions, and some shorter skill screens run as little as 30 minutes. The CCAT is the tightest documented case at 50 questions in 15 minutes. When the clock is that tight, familiarity with the question format converts directly into answered items, which is why practice helps on a test whose underlying construct you cannot change in a week.

The decoder table: time, scoring, and whether practice moves the score

This is the row to jump to. Find your test family and read across.

Read "coachable" precisely. On cognitive tests, practice improves your speed and your familiarity with question types, but not the underlying ability the test targets. The CCAT is designed to measure your current cognitive ability, and there is no way to improve that in days; genuine cognitive improvement happens over months. So the honest claim is narrow: full-length timed simulations raise your score by making you faster and more familiar, not smarter.

On coding tests, the pattern is coachable because the question types recur. On personality tests, the profile is faking-resistant by design, and the section below explains why gaming it usually backfires.

Practice does not make you smarter in a week. On a speed-gated test it makes you faster, and that is enough.

How each test is scored and gated

The scoring model decides what result advances you, and it differs by family. Get this wrong and you will chase a number that does not exist.

Cognitive tests are scored two ways at once. The employer receives your raw score, the number you got right, and your percentile, how you did against everyone who has taken the test. A relative score places you on a bell curve alongside other candidates. For reference, the CCAT average is 24 out of 50, which is the 50th percentile.

Coding tests are threshold-gated. Your score is per problem, built from the test cases you pass, both the visible ones and the hidden ones, and reported as a percentage. Passing every visible case does not guarantee advancement, because the employer sets a pass threshold that stays private. The final bar is the company's call, not a fixed public number.

SJTs are usually customized for the employer and profiled against a norm. Your results are compared to others to place you in a percentile group, a good score is typically the 70th to 80th percentile, and the result is combined with your other assessments. A strong score helps you progress; it rarely decides alone.

Personality tests produce a fit profile against norms, not a pass or fail. Hogan's HPI, for example, has seven primary scales (Adjustment, Ambition, Sociability, Interpersonal Sensitivity, Prudence, Inquisitive, Learning Approach) mapped onto the Big Five and tuned for occupational prediction. There is no passing number to hit, only a match to what the role wants.

Coachable or faking-resistant: what your prep hours can actually buy

The single most useful classification is whether your test can be moved by practice. Spend hours where they pay off, and answer honestly where they do not.

Cognitive tests are coachable on speed and familiarity but not on the underlying ability. Skills and coding tests are coachable on patterns. Personality and integrity tests are largely faking-resistant, and the field data explains why gaming them is a bad bet.

Faking on personality tests is real but muted in practice. In a real applicant sample, 5.2% or fewer improved their scores on any scale on a second occasion, and scores were as likely to move down as up. That is the strongest argument for answering honestly: the expected gain from gaming is near zero, and the downside includes tripping a social-desirability flag.

Integrity tests split sharply, and the label matters more here than anywhere else.

An overt integrity test asks you directly about attitudes toward honesty, theft, and rule-following. A personality-based one infers the same tendencies from indirect trait statements. The overt type is roughly 3.7x more coachable. Forced-choice formats, where you pick between statements matched for perceived desirability, are a documented mitigation: they aim for more honest responses by removing the obvious "good" answer. If your items force you to choose between two equally flattering options, you are looking at a design built to defeat gaming.

The eight-step procedure for any assessment invite

Run these in order the moment an assessment lands. Two steps have a strict order note: classify before you prep, and request any accommodation before you start a timed test.

For the coachable prep step, the specific method that works on cognitive tests is full-length timed simulation. The full-length practice gives the real sense of what it takes to answer under the clock, which untimed drills never do. For the coding step, learning the submission mechanics in the official practice environment removes the interface surprises that cost real points on test day.

How the assessment misreads a real candidate

This is the section that saves scores. Each item below is a documented way the test lies about you, plus the check that catches it.

1. Wrong-platform prep. Some coding invites arrive under one brand when candidates expect another. A few Microsoft coding invites, for instance, arrive under the Codility brand, which is why candidates mix up Codility and HackerRank and prep for the wrong environment. Check: read the invite label, not the job board.

2. Passing visible tests, failing hidden ones. A function that returns the right answer on small examples can still lose points to overflow, an off-by-one error, an empty input, or an inefficient nested loop. The false positive is the row of green checkmarks on sample cases. Check: test edge cases and large inputs yourself before submitting.

3. A clean, fast solve triggers a cheating flag. Code replay can flag a flawless solution finished in fifteen minutes with no mistakes. Clean code alone will not clear a flagged session. Check: do not paste pre-memorized solutions, and let your working history show.

4. Tab-switching disqualification. Proctor Mode watches the webcam, extra screens, and screen sharing, flags tab switches, disables copy-paste, and uses screenshot analysis to catch external AI tools. Glancing at a second monitor can read as a violation. Check: single screen, no other apps open.

5. Over-managing a personality test. Trying to fake-good backfires because responses are as likely to move down as up, and social-desirability scales can flag the attempt. Check: answer consistently, and trust that forced-choice items are built to make gaming pointless.

6. Misreading a low SJT score as "wrong personality." A low SJT score is diagnostically ambiguous. It may mean you lack cognitive ability, or have a different personality profile, or lack relevant experience, or simply hold different values about handling interpersonal situations. Check: SJT keys are firm-specific, so align to the employer's published values rather than a generic notion of the right answer.

7. Assuming the percentile bar is public. Employers set private thresholds on coding and cognitive tests, and a high raw percentage does not guarantee advancement. Check: treat any published passing score as an estimate.

8. Skipping the accommodation request. Waiting until after a timed test to raise an ADA need forfeits the remedy, because the obligation attaches before the screen-out. Check: request in writing before starting.

What the science says these tests can and cannot predict

Assessments are gates, but they are imperfect predictors, and knowing the ceiling on their accuracy helps you keep your own result in perspective. The validity numbers below are the best available meta-analytic estimates, and they have been revised.

General mental ability was long the strongest single predictor, correlating with job performance at about r = .51 in Schmidt and Hunter's meta-analysis, well above experience (0.18) and education (0.10), coefficients described as unlikely to be useful. Sackett and colleagues later revised GMA down to r = .31 under more conservative corrections. Either way, no single test is decisive: a structured interview combined with GMA reaches .63, higher than either alone.

SJTs sit lowest of the tested methods at about r = .26, and they add only ΔR = .03 to .08 over cognitive ability and personality already measured. That small increment is why an SJT score is the least portable signal you will produce, and why studying one employer's values beats hunting for universal best answers.

There is a structural reason the same off-the-shelf tests keep appearing.

The people qualified to build and validate a bespoke test are scarce, so most employers license standardized instruments rather than construct their own. That is good news for you: practice materials for named tests like the CCAT generalize well, because the test you practice on is very likely the test you will sit.

Your rights and the fairness limits on these tests

In the US, adverse impact is governed by the four-fifths rule: a selection rate for any race, sex, or ethnic group that is less than four-fifths, or 80 percent, of the rate for the highest-scoring group is generally treated by federal enforcement agencies as evidence of adverse impact (29 CFR 1607.4). For scale, the EEOC received more than 88,000 discrimination charges in fiscal year 2024.

On disability, an employer's use of an algorithmic decision tool may violate the ADA if it screens out someone who could perform the essential job functions with or without reasonable accommodation, asks questions meant to reveal a disability before a conditional offer, or fails to offer reasonable accommodation, for example to an applicant who cannot interview on video. The practical consequence for you is timing: request accommodation before the screen begins.

The enforcement landscape shifted. An executive order directed federal agencies to step back from disparate-impact enforcement, and the EEOC moved to stop investigating charges based on disparate impact alone. Disparate impact remains part of Title VII, and private individuals can still bring these claims in court. A precise, published step-by-step candidate accommodation procedure is not established in the public sources beyond this employer-obligation framing, so if you need an accommodation, put the request in writing and keep the confirmation.

If you want to pressure-test what a specific screen actually asks, it helps to talk to people who administer or study it. Refolk turns that into a plain search over public professional profiles.

Refolk also writes your resume from your own history and scores how well you fit a posting before you spend an evening on its assessment, which tells you whether the screen is worth your hours at all. You can start from Refolk.

The pre-submission checklist

Run this before you click submit on any assessment.

Keeping this current

Two things in this guide move over time, and both have a mechanism you can re-check rather than a value to memorize. Validity estimates get revised as new meta-analyses land, which is exactly why the table shows both the 1998 and 2022 GMA figures; when you see a single confident number quoted anywhere, ask which correction it used. The legal enforcement posture also shifts, so before you rely on a fairness claim, confirm whether disparate-impact enforcement is active and whether your state has added its own rules on automated hiring tools. When a new invite arrives, start again at step one: name the test, find its scoring model, and decide whether your hours can move it.