What Is a Good Predictive Index Cognitive Assessment Score?

There is no universal good score on the Predictive Index (PI) Cognitive Assessment, and PI says it is not a test you pass or fail. Scores run from 100 to 450 with an average of 250, which PI's own blog says corresponds to a raw score of 20 out of 50. A candidate's score is compared with a Cognitive Target that the hiring manager sets for the role, and PI recommends against cutoff scores. In its current software PI no longer shows a percentile. This guide sets out what PI's support pages, administrator's guide and sample report say, works out what its published percentile examples imply, and explains why PI says the test does not measure IQ.

What is the PI Cognitive Assessment, and how long does it take?

PI's administrator's guide describes a 12-minute timed assessment of 50 multiple choice questions in three categories: verbal, numeric and abstract reasoning. PI's scoring page says a respondent can attempt at most 16 verbal, 18 numerical and 16 abstract reasoning items; the 2018 guide gave the split as 14, 18 and 18. PI says it developed the test in 2010, after clients asked for a short cognitive ability measure, and that the work followed the SIOP Principles for the Validation and Use of Personnel Selection Procedures and the Standards for Educational and Psychological Testing.

The time limit is part of what is measured. The guide says most respondents will not answer all 50 questions in 12 minutes and calls that okay and an intended facet of the design, because the limit brings in information processing speed, which PI treats as a key component of general cognitive ability. A score therefore reflects pace as well as accuracy. For a timed test that reports on the standard IQ scale, see the timed IQ test.

• Length: 50 multiple choice items in 12 minutes, ending automatically at 12 minutes
• Content: verbal, numerical and abstract reasoning items, five on screen at a time
• Not adaptive: PI says a linear-on-the-fly engine builds each 50-item set to equivalent content and difficulty
• Extended time: 18-minute and 24-minute formats, reported on the same 100 to 450 scale
• Retakes: the employer decides; PI's guide says no more than three attempts, with the highest score used
• Delivery: online on a computer, laptop or larger tablet, proctored or unproctored

How is the PI Cognitive Assessment scored?

PI reports one number for decisions: the scaled score at the top of the report. Scores run from 100 to 450, the global mean is 250, and results come in 10-point steps. The guide says the score is based on the number of correct answers: incorrect answers do not count against the respondent, and there is no other penalty or weighting. The report also lists raw scores, meaning items answered correctly and items attempted, and a subscore for each content area. PI says to disregard those. Raw scores are only descriptive statistics about the administration, subscores are not valid or reliable enough to support hiring decisions, and raw scores are not always comparable between test revisions or when a respondent had extra time.

There is one published anchor between the two scales. PI's blog on understanding scores says the average scaled score of 250 translates to a raw score of 20, set during assessment development with a reference group of more than 288,000 assessments that scored 19.8 out of 50 on average. So the popular claim that the average is about 20 correct answers is right. Beyond that pair PI publishes no raw-to-scaled table, and because raw counts do not carry over between revisions, a prep-site chart mapping every raw score to a scaled one has no PI source.

What do PI's percentile examples show?

Percentiles are what most candidates search for, and PI's own documents give only examples. The 2018 administrator's guide shows a score of 340 at the 86th percentile of the global adult working population, a sample of over 900,000 working-age adults, meaning the respondent scored as well as or higher than 86% of the global workforce. PI's sample report, generated from a January 2018 assessment, shows a score of 200 and says the respondent scored better than 26% of the global population of adults across all industries. PI's later blog posts describe the reference group as more than 288,000 respondents, and the pages do not explain the difference in size.

If scores follow a normal curve, the two examples imply a standard deviation of about 78 points (from 200 and the 26th percentile) to about 83 points (from 340 and the 86th). PI states no standard deviation in the pages we read, so treat what follows as our estimate, not a PI table, and do not extend it far beyond the examples.

• 200: the 26th percentile (PI's sample report)
• 250: the global mean, the middle of the curve
• 300: about the 73rd percentile (our estimate)
• 340: the 86th percentile (PI's guide)

On the IQ scale most clinical tests use, mean 100 and standard deviation 15, the 86th percentile is about IQ 116 and the 26th about IQ 90. That is a translation of percentiles, not a claim that the two tests measure the same thing. Our percentile calculator and score converter do that arithmetic, and our guide to IQ z-scores shows how a percentile becomes a standard unit.

Why has PI removed the percentile?

This is the change most prep material has not caught up with. In a blog post first published on 17 December 2025 and updated on 10 March 2026, PI answers the question of why its assessment does not give the familiar percentile. It describes the test as a high jump, not a race: a criterion-referenced tool whose scores are read against a Job Target set by the hiring manager, not against a bell curve. It says the reference group is the wrong group to compare a candidate against, and that the decision was made to remove the percentile comparison in its software to align the platform with that design. It also argues, citing the 1971 US Supreme Court case Griggs v. Duke Power Co., that a generic percentile from a broad population is difficult to prove as job-related.

PI's reliability page goes further. It says the assessment is criterion-referenced rather than norm-referenced, so scores rest on the number of questions a person answers correctly, not on a comparison with other people. That sits awkwardly beside a mean of 250 and a percentile against a reference group, and the pages we read do not reconcile the two. The Technical Manual PI refers to is available only on request through a PI Certified Partner, so we could not read it. A current report may therefore show no percentile at all, and the examples above come from older documents.

Target scores, match scores and pass marks

PI tells employers to compare a candidate with the job target, not to rank candidates against one another. Its blog says plainly: we don't recommend the use of cutoff scores, and it prefers match scores instead. In the 2018 guide a Cognitive Match Score on a scale of 0 to 10 is 10 whenever the score is the same as or higher than the target, and PI says there is no research to support the worry that a candidate above the target will be bored. Below the target the match score falls on a sliding scale. PI says candidates below a target should not be disqualified on the score alone, and that the score is one data point next to personality, experience and education. With no job match to go on, the guide's baseline target is 250.

PI does not publish current targets by occupation. An older Target Scoring Guide dated 2017, hosted by a PI partner and written for the test's earlier name, the Learning Indicator, suggested starting targets from 230 (operations, unskilled workmanship) to 310 (natural sciences) across 33 job functions, with customer service at 250 and sales at 270, adjusted by up to 30 points and raised by 10 for managers and 20 for executives. We could not confirm those numbers are still used. Criteria's sample CCAT report, by contrast, lists suggested minimum scores for 42 job families; see our guide to what a good CCAT score is. The practical answer to what a good PI score is: at or above your employer's target, which you may never be told.

Is the PI Cognitive Assessment an IQ test?

PI says no, in so many words. Its scoring page says the test measures general cognitive ability and the ability to adapt, grasp and handle complexity, and that it does not measure IQ or acquired knowledge but indicates how fast a person can be expected to acquire new knowledge. PI's documentation adds that IQ tests measure intelligence, which is closely related to g, but not the same. The 2018 guide, though, describes the score as a measure of general cognitive ability, also called g. General cognitive ability is what the g factor names, and g is what full IQ batteries are built to estimate. What differs is the instrument: a 12-minute speeded test of 50 items reported on a 100 to 450 scale, against a battery that takes 60 to 90 minutes with a psychologist and reports on a mean-100 scale, as our guide to WAIS index scores shows.

What evidence does PI publish?

PI reports correlations between the assessment and job performance of .29 to .44, without giving sample sizes, jobs or the correction method on the page we read. A correlation of .29 means the score accounts for about 8% of the variation in performance, and .44 about 19%; the rest is everything else. How large these figures should be is contested, as our guide to what a cognitive ability test predicts at work explains. On reliability the page gives no figure, saying only that studies show at least acceptable levels; PI's general research page sets 0.70 as the bar for acceptable internal consistency.

PI announced on 18 August 2020 that the assessment had received a certificate of conformity under the European Federation of Psychologists' Associations Test Review Model, reviewed by DNV-GL auditors, and a PI page says it was recertified in 2023. That is the vendor's own account: we found no published review ratings, no Buros or BPS review and no journal paper on this test. PI says it checks every question for measurement bias between groups, and its administrator's guide tells employers to validate the test for each job and to check adverse impact at least annually, using in the US the 80% rule and the standard deviation method. The 80% rule is the four-fifths rule covered in our guide to adverse impact in hiring tests.

What this means if you are about to take it

You may never see your number. PI recommends keeping cognitive scores confidential and telling candidates about fit or ranking rather than the score, and its guide says employers should state in advance whether respondents can see their scores. Because wrong answers cost nothing on PI's stated rule, an unanswered question earns nothing while a guess cannot lower the score. PI also says the assessment is for pre-hire decisions only, and concedes that familiarity with its question types can inflate scores; our guide to the practice effect on cognitive ability tests covers how much a second attempt changes a result. An unproctored remote test raises the questions in our guide to AI agents and proctored hiring tests.

PI is one of several employer tests with no public pass mark. SHL reports a percentile against a named comparison group, as our guide to SHL test scores explains; the Watson-Glaser publisher says a score cannot be called good without a norm; and Wonderlic says outright that there is no universal good score, as our Wonderlic guide shows. To see where a result of your own lands, take the IQ Metrics test and read it against the IQ bell curve.

Common questions