How to hire top talent using abstract reasoning test

An abstract reasoning test is worth running when the job asks people to work out something new, and worth skipping when it does not. That one line decides whether the score helps you or just adds a number to the file. Used on the right roles, at the right stage, with a threshold set in advance, it tells you how a candidate handles an unfamiliar problem before anyone books an interview.

TL;DR

• A pre-employment logic test measures pattern-finding and rule-inference, not knowledge. The abstract version uses shapes instead of words, so reading level matters less.
• Test the roles that actually solve novel problems. In BLS data, 53.3% of legal jobs and 49.0% of computer and mathematical jobs carry a real problem-solving requirement, against 0.8% of food preparation and serving jobs.
• Reasoning scores do predict performance, but the famous validity numbers were revised downward in 2022 and researchers are still arguing about them in print. Treat the score as one signal with a known error bar, not a verdict.
• Set the pass mark before you see the results, write down why, and resist re-scoring a candidate after a good interview. Combining evidence by rule beats re-weighing it by feel.
• Practice changes scores. Test-prep sellers advertise exactly that, so a polished score can mean preparation rather than ability. Variable question banks and timing limits blunt it.
• If you hire in New York City and the tool scores candidates automatically, Local Law 144 puts a bias audit and a 10-business-day notice on you, not on your vendor.

What is a logic test for pre employment?

A pre-employment logic test is a short, timed assessment that measures how well someone spots a rule and applies it to something they have not seen before. It does not test what a candidate knows. It tests how they work out what they do not know, which is why a strong score turns up fairly often in someone with a thin resume.

The abstract reasoning test is the nonverbal branch of that family. Candidates see shapes, grids and sequences, then pick the option that completes the pattern. No words, no arithmetic. Its verbal and deductive cousins, including the logical reasoning test, put the same reasoning behind sentences and statements instead of images.

That distinction matters more than it sounds. A shape-based test asks less of a candidate's English, so it travels better across a multilingual pipeline. It also tells you less about whether someone can read a messy requirements document and find the contradiction in it. Pick the format that matches the thinking the job actually requires, and read the broader cognitive ability testing picture before you commit to one.

How do abstract reasoning tests work?

The shape is consistent across formats: a timed set of items, one hidden rule to infer per item, and a score reported as a percentile against a comparison group rather than a raw mark. Testlify's logical reasoning test is a worked example of the usual proportions, at 10 minutes and 10 questions, intermediate difficulty, in English, German, Spanish, Dutch and French.

Three question shapes do most of the work:

1. Matrix completion. A grid of shapes with one cell blank. The best-known version is the Raven's Progressive Matrices style item, where two or three rules combine at once.
2. Sequence continuation. A row of figures that changes by a hidden rule. The candidate picks what comes next.
3. Odd one out. Several figures share a property and one breaks it. Finding the shared property is the test.

The settings around those questions matter as much as the questions. Testlify's test library carries abstract reasoning and psychometric categories at three difficulty levels (easy, medium and hard), with per-question time limits, shuffled questions and shuffled answer options. Question banks can be fixed, so every candidate sees the same items, or variable, so the system draws unused questions from a larger pool for each candidate. Real-time translation of questions and options into a candidate's preferred language is available, with the honest caveat that quality varies by language pair.

Variable banks are the setting most teams skip and then wish they had used. Fixed banks are easier to compare but easier to share, and the product tells you so directly: smaller question banks raise the chance that two candidates see overlapping variable questions. So the real tradeoff is pool size, not fixed-versus-variable. A small pool set to variable gives you the comparison problems of a shuffled test and the leak risk of a fixed one.

Which jobs need an abstract reasoning test?

Test the jobs where novel problems are part of the work. The Bureau of Labor Statistics measured this directly in its Occupational Requirements Survey, asking how often a worker is responsible for solving problems that take more than five minutes to think of a good solution. The answer varies enormously by occupation group.

Source: BLS Occupational Requirements Survey analysis, published March 2021.

Read that as a starting assumption, not a rule. Legal at 53.3% still means roughly half of those jobs do not carry the requirement, and sales at 7.8% can be the exception when the territory is new and nobody has a playbook yet. The question is narrower than "is this a smart-person job": does this person have to figure out something unfamiliar, often, without anyone to ask?

Three patterns where a reasoning test earns its place: a role with no clean work sample (early-career hires with nothing to show you), a role where the tools change faster than the training, and a high-volume pipeline that needs a defensible first cut. Two where it does not: a role defined by a specific verifiable craft, where a work-sample or skills test beats a reasoning score outright, and any role where you cannot say out loud what the score would change.

Pro Tip: before you add a reasoning test to a requisition, write one sentence naming the decision the score will inform and the threshold that triggers it. If you cannot finish the sentence, the test is decoration and it is costing your candidates 20 minutes each.

Where an abstract reasoning test fits your hiring process

Put it after the application and before the first human conversation. That is where it saves the most time, because it reorders the queue before anyone spends an hour on a call. Running it later turns it into a tie-breaker nobody trusts, and running it before the application costs you candidates who would have converted.

In practice the sequence that holds up is short: application, reasoning test plus one role-relevant skills assessment in the same sitting, structured interview for the people above your line, then a reference check. Score thresholds can gate the stage change itself, so the shortlist is already ordered when a recruiter opens it. On a 200-candidate pipeline with the line set at the top 40%, that is the difference between reading 200 applications in no particular order and working down 80 ranked profiles.

Do reasoning test scores predict job performance?

Yes, and less cleanly than the hiring industry claims. Cognitive and reasoning tests remain among the stronger predictors of job performance available before a hire. The specific numbers everyone quotes, though, come from meta-analyses whose corrections were challenged, and the challenge has largely held.

The 2022 reanalysis by Sackett and colleagues argued that decades of validity estimates had been systematically overcorrected for range restriction, and revised many of them downward. That paper is not the end of it. Oh, Le and Roth replied in 2023, Bobko and colleagues pushed back in 2025 on how predictors should be ranked at all, and Sackett's team answered again the same year. The ranking of methods mostly survives every reanalysis. Any single coefficient does not.

So the honest posture is neither "score decides" nor "tests do not work". This measure carries real signal, with an error bar wider than the brochure suggests, on one dimension of a job that has several. The 1998 Schmidt and Hunter summary that is still widely quoted is a landmark, not current consensus: a precise validity figure lifted from it without the 2022 revision is a number that has moved.

The Testlify Multi-Signal Talent Evaluation Model is built on exactly that: one signal is fragile, several independent ones create confidence. A candidate should advance because the reasoning score, a role-relevant skills assessment and a structured interview point the same way, not because one of the three was strong. It does not mean every process needs every signal, and it does not weight them equally.

How to set a passing score you can defend

Set the threshold before results arrive, and write down the reason. Percentile benchmarking makes this concrete: Testlify reports percentile, percentile rank and candidate rank, and compares scores against other candidates and across reviewers, so "top 40%" is a line you can actually draw rather than a feeling about a raw mark.

Four steps that survive a challenge:

1. Pilot on people already doing the job. Run the test on a group of current employees in the role. Their spread tells you what the score looks like in someone who is working out fine.
2. Pick a band, not a cliff. A single pass mark turns a two-point difference into a hire-or-reject. A band (advance, review, decline) matches the measurement error you actually have.
3. Write the rule down before the first candidate sits the test. A threshold invented after the fact is indistinguishable from a preference.
4. Combine signals by rule, not by instinct. This is the step teams skip.

That last one has the strongest evidence behind it of anything in this article. Kuncel and colleagues found that mechanically combining assessment information predicts outcomes as well as or better than experts re-combining the same information by judgment. Read it carefully, because it is narrower than "trust the algorithm": it is evidence against letting intuition rework the numbers at the end, after the evidence is already in. Humans choose what to measure and own the decision. What they should stop doing is quietly re-weighting a score because a candidate interviewed well.

In practice: set the weights in advance, let the system aggregate, and when you override it, log why. Testlify supports score-based weighting, where each test's impact follows its total score, and weights-based, where you set x0 to x5 per test. Writing the override down is what turns a gut call into something a reviewer can audit.

Keeping the test fair and legally defensible

The EEOC's position on employment tests is short and specific: employers "should ensure that employment tests and other selection procedures are properly validated for the positions and purposes for which they are used." If a procedure produces a disparate impact, determining that ordinarily requires a statistical analysis, and the employer then has to show the procedure is job-related and consistent with business necessity. Even then, a challenger can point to a less discriminatory alternative that would have done the job.

Read against the table earlier in this article, that is a practical instruction. A reasoning test on a food-service role, where 0.8% of jobs carry a problem-solving requirement, is harder to defend as job-related than the same test on an engineering role at 47.1%. The business-necessity argument is made at requisition time, by role, or it is made badly later.

Location adds a second duty. New York City's Local Law 144 requires an automated employment decision tool to be "subject to a bias audit within one year of the use of the tool", the audit information to be publicly available, and candidates to be notified 10 business days before use. Enforcement began on July 5, 2023. Whether a given reasoning test falls inside that definition depends on how it is scored and how much the output drives the decision, which is a question for your counsel and not for a vendor's marketing page. The compliance work lands on the employer either way.

Two fairness mechanics worth setting up properly. Accommodations work best as a structured request a candidate initiates, routed to an administrator who adjusts the session, rather than an informal email thread; Testlify's flow asks about English fluency and about conditions affecting memory or concentration, and it is human-reviewed, not automatic extra time. And demographic data for fairness monitoring should be collected anonymously, kept away from the people making hiring decisions, and optional, which is how Testlify's DEI collection is built.

What do hiring teams get wrong about reasoning tests?

Four mistakes account for most of the damage.

Treating a practiced score as a measured one. An entire industry sells reasoning-test practice to candidates, and it advertises score improvement as the product. That is not a conspiracy, it is a measurement problem: a candidate who has sat a few hundred matrix items is not showing you the same thing as one seeing their first. Variable question banks drawn from a large pool, firm per-question timing and a current item set are what keep the comparison honest. A test you have run unchanged for three years is measuring preparation.

Letting one score carry the decision. A reasoning test says nothing about whether someone finishes work, takes feedback or tells you when a deadline is slipping. Teams that over-rotate on the score hire people who interview as clever and perform as unreliable, then conclude the test does not work. The test worked. It was answering a different question.

Running it on the wrong roles. Adding a cognitive hurdle to a role that does not need one costs candidate goodwill, raises the adverse-impact exposure described above, and buys no information. The table earlier is the cheapest sanity check available.

Telling candidates nothing. A candidate who sits a timed abstract test and hears nothing back learns that your process takes and does not give. A single line naming the band they landed in costs almost nothing and is the part of a rejection people remember.

Hire on evidence, not one number

A reasoning test is a good first cut, a fair ordering mechanism for a crowded pipeline, and a weak basis for a final call. Pair it with one role-relevant skills assessment and a structured interview, set the threshold in writing beforehand, and the shortlist that reaches a hiring manager is already defensible. You can book a demo to see how the abstract reasoning test, scoring bands and reviewer workflow fit together, or work through the wider pre-employment testing guide first.

FAQs