Test Partnership Score Report Explained: Percentiles, Stens and Norm Groups

Test Partnership is unusual in publishing a full candidate report, game by game, for a sample candidate called Simone. It is the clearest public picture of how one UK publisher turns a test into a number, and it answers most of the questions candidates ask about their own results.

What each result shows

Every test gets its own page with three readings of the same performance. The first is a percentile, which the report defines as “the percentage of people's scores that Simone's score exceeds”. The second is a sten from 1 to 10, which Test Partnership's technical manual defines as a standardised scale with a mean of 5.5 and a standard deviation of 2. The third is a verbal band. If percentiles and stens are new to you, our guide to what a good psychometric score is covers both scales in general; this post sticks to what Test Partnership's report does with them.

Here is the sample candidate's full set of MindmetriQ game results, compared against the graduates, professionals, managers and executives norm group, with the percentiles the report gives for the two alternative groups.

Why the score is not the number you got right

MindmetriQ games are adaptive. In the report's own words, “A correct answer causes the next question to be more difficult, and an incorrect answer causes the next question to be slightly easier”, and “difficult questions are worth more than easier questions”. Two candidates who each answer eight questions correctly can therefore get quite different percentiles. The technical manual adds that MindmetriQ picks questions you have about a 70% chance of getting right, so a strong candidate and a weaker one can both finish feeling they got most of them. How hard the questions became is the signal, not how many you got right. Our Test Partnership test tips cover what that means for how you should pace yourself.

The norm group moves the number, not always the way you expect

Every result names the group it was compared with. In the sample, each game's graduate norm group holds 3,416 people. The report then shows the percentile the same performance would earn against the other two MindmetriQ groups: administrative, operational and non-graduate staff, and apprentices, school leavers and entry-level staff.

You might expect the graduate group to be the toughest comparison every time. The sample says otherwise. Word Logic moves the most, from the 68th percentile against graduates to the 81st against non-graduate staff, but Net the Numbers goes the other way, from the 90th against graduates to the 88th. The lesson is narrower than the usual advice: your percentile belongs to one comparison group, and the report tells you which one, so read that line before comparing your result with anyone else's.

The overall combined score

When an employer uses several games, the report opens with an overall combined score, again as a percentile, sten and band. Test Partnership says this figure is “a more valid predictor of job performance than the result of any individual assessment on its own”, because several tasks together measure general ability better than one.

In the sample, the six game percentiles of 90, 58, 71, 68, 97 and 78 produce an overall 82, sten 7, Above average. The plain average of those six percentiles is 77, so the combined score is not simply an average of percentiles. Test Partnership does not publish the formula. What the sample does show is that one weaker game, Number Racer at the 58th, did not drag the overall below the Above average band.

The traditional tests report the same way, with more norm groups

Test Partnership's traditional tests follow the same model. The Insights numerical, verbal and inductive tests are also adaptive with what the publisher calls “Intelligent IRT scoring”, and are normed against three standard groups, with the option of employer-defined ones. Concepts Critical Thinking offers twelve norm groups. The TPAQ personality questionnaire is different in kind: it reports a profile across many traits against one of fourteen norm groups, and gives no overall score unless the employer has set up an optional JobFitScore, which combines selected traits into a single role-fit figure.

So what is a pass?

Test Partnership does not publish a pass mark for any of its tests. The employer decides which norm group to use and where to set the bar, and that decision is not usually shared with candidates. Nor is there any negative marking: the live test instructions tell candidates to use their best guess rather than leave a question unanswered. Practical advice follows from both. Answer everything, and judge yourself against the band rather than a rumoured cut-off.

You can try the games and the Insights tests on our Test Partnership practice page. For where Test Partnership sits among other UK publishers, see the psychometric tests guide.