Stellwerk 8: what the 200 to 800 scale really means

After a Stellwerk test the same sentence is said in every classroom: “I got about half of them right.” In Stellwerk that sentence is meaningless. It describes nothing the test measures, and it says nothing about whether the result is good or bad. Almost everybody who got about half right got about half right, regardless of how strong they are.

That sounds like a contradiction at first. It is the direct consequence of how the test is built. Anyone who has understood this one mechanism reads their Stellwerk profile correctly afterwards, and stops falling for the same misreading in other aptitude tests.

What Stellwerk is, and who is behind it

If you are new to Switzerland, Stellwerk is probably something your child mentioned after the event rather than something you booked. It is an assessment of where a pupil currently stands, used in lower secondary school (Sekundarstufe I) and published by the Lehrmittelverlag St. Gallen, the teaching materials publisher of the canton of St. Gallen, which developed the task pool together with the Pädagogische Hochschule St. Gallen, the cantonal university of teacher education.

Five subjects are examined: mathematics, German, French and English, and since the 2023/24 school year also Natur und Technik, which covers physics, chemistry and biology. The basis for the test is the Lehrplan 21, the shared curriculum of the German-speaking cantons: the task pool was built on its testable competence levels and then normed.

There are three runs, Stellwerk 7, 8 and 9, roughly at ages 13, 14 and 15. Pupils take the assessment once a year. One test per subject costs CHF 10.00 per pupil, and the Stellwerk plus variant costs CHF 7.00. The school organises the whole thing, not the family: there is no test centre, no separate booking and no appointment to arrange.

That price is worth pausing on, because it usually arrives in the same conversation as the Multicheck, the commercial aptitude test taken at a test centre for CHF 60 to CHF 100 and paid for by the family. Ten francs against a hundred is not a quality difference. It is the difference between an instrument a canton buys in bulk for its schools and a product a family buys for one child.

Whether your child sits Stellwerk at all depends on where you live. Education in Switzerland is a cantonal matter: there are 26 cantons, no national school system and no national school test. Stellwerk is compulsory in some cantons, merely in use in others, and entirely absent in most. A family moving from Zurich, where Stellwerk 8 is compulsory, to neighbouring Aargau will meet a different system, with different school types and different assessments. Which cantons run it is set out in the canton overview for the Stellwerk test.

One last piece of orientation before the mechanics. The 8. Klasse, the year of Stellwerk 8, is the second-to-last year of compulsory lower secondary school, at roughly age 14. It is the year in which apprenticeship applications are written, which is why a measurement taken in it attracts far more attention than a mid-year test normally would.

And the most important framing of all: Stellwerk is not an exam with a pass mark and it is not a mark. It is an instrument for planning support, designed, in the provider’s words, to sit in the young person’s own responsibility. Read it as an exam and you will read it wrongly.

Adaptive means the test changes while you take it

The decisive sentence is in the provider’s interpretation guide: unlike traditional tests, the pupil’s answering behaviour influences how the test proceeds. After every task answered, the algorithm estimates that pupil’s ability and then presents fitting tasks, neither clearly too easy nor clearly too hard.

The high jump is a good picture for it. The bar starts at a middling height. Clear it and the next jump is higher. Knock it off and the bar comes down. At the end, what counts is not how many jumps succeeded but the height at which the limit lay. That height is exactly what the points value reports.

From which something follows that surprises people at first: a strong result and a weak result come out of roughly the same share of correct answers. Both pupils were corrected upwards or downwards until the tasks fitted them. Fitting here means roughly: just about solvable.

Why “half of them right” says nothing

A constructed example makes it tangible. Two pupils in the same class, same subject, same feeling afterwards.

The table is an illustration, not a measurement. But it shows why the playground comparison, “how many did you get”, leads nowhere: the two did not take the same test. They used the same task pool and were presented with different tasks out of it.

This is the point where the scale itself becomes readable. The 200 to 800 value does not report an amount of work done. It reports the difficulty of the tasks the pupil mastered. Two pupils with an identical share of correct answers can be hundreds of points apart, and nothing has gone wrong.

Multidimensional means the areas help each other

The second part of the procedure is explained even more rarely. In multidimensional testing, tasks from the individual competence areas are presented alternately. That produces one value per competence area and one overall value for the subject, and this overall value is explicitly not the average of the competence areas.

The reason is remarkable. The algorithm can draw conclusions about ability in the other competence areas from how a task in one area was handled. A pupil who calculates confidently in “Zahl und Variable” is initially treated as fairly strong in “Form und Raum” too, until tasks there show something different. The provider therefore describes the measurement as highly efficient and precise.

In practice: you cannot recalculate the overall value from the area values, and a mistake in one area works through differently from the way marking logic would suggest.

Why the test is not the same length for everyone

The length follows from the same principle. The measurement uncertainty shrinks with every task answered. Once a defined level of test precision is reached, the test ends.

A pupil who answers consistently reaches that precision sooner. A pupil who fluctuates, solving hard tasks and missing easy ones, needs more tasks before the algorithm is sure enough. A longer session is therefore not a bad sign and a shorter one is not a good sign. It says something about how unambiguous the answering behaviour was, and nothing else.

Two practical consequences. There is no point in trying to finish quickly, because the test does not end after a fixed number of tasks but when the measurement is precise enough. And there is no point in guessing at a hard task to move on: one lucky correct answer between two wrong ones makes the picture less clear, and the test will then tend to present more tasks rather than fewer.

The 200 to 800 scale, and where it comes from

Results are mapped onto a scale from 200 to 800 points, and that scale covers the whole third cycle, which is the Lehrplan 21 name for lower secondary school. The frame of reference is a broad norming among pupils in the 8. Klasse of the cantons of St. Gallen and Zurich. The distribution of that reference population’s results follows a normal distribution.

Two things follow. First, a value without a subject and without a class level cannot be interpreted, because the scale covers the whole of lower secondary school while the norming was done on the 8. Klasse. Second, the midpoint is not at 500. It lies where the interpretation guide puts it for the subject in question, and for mathematics in the 8. Klasse that is around 511. Those guides are published on lernpassplus.ch and stellwerk.ch.

It is worth saying the counter-intuitive part once more in plain words, because it is the reason this article exists. The number is not a percentage. It is not a mark out of 800. It does not go up because you answered more tasks. It goes up because the tasks you answered correctly were harder. A pupil who solves twelve hard tasks scores higher than one who solves twenty easy ones, and the test is designed so that neither of them will ever notice the difference from the inside.

Can you practise for Stellwerk, and what does it achieve

Sample tasks are available on stellwerk.ch for mathematics, German, French, English and Natur und Technik. The canton of Lucerne states the expectation soberly: in the second class of lower secondary school, sample tasks are enough by way of preparation, while Lernpass plus serves mainly in the third class, where it also supports development.

Kognita does not sell preparation courses and therefore has no reason to make practising sound bigger than it is. The honest version: knowing the format takes away nervousness, building competence raises the value, and there is very little in between.

Three sentences after the test, and what they are worth

Conversations after a Stellwerk test turn on almost the same three statements every time. All three come from the habit of reading a test as if it were an exam.

Take those three apart once and the playground comparison stops being interesting.

From the profile to a target

The actual purpose of the result lies behind the points value. The values in the competence areas can be mapped onto the competence levels of the orientation tests, and every level comes with a description of what somebody at that stage can do. A number turns into a sentence about an ability.

The provider illustrates this with an example pupil who reaches 542 points in mathematics, selects his chosen occupation in Jobskills and sees there that a higher level is expected in two competence areas. He writes his target onto a printout of his profile and works towards it with orientation tests and task sets. That is exactly what Stellwerk was built for: a starting point for setting a goal, not a ranking.

What the profile does not say

The provider is clearer on this point than most guidance sites. The Stellwerk profiles represent a snapshot, allow a cautious placement of the school performance achieved so far, but make no statements about the pupil’s future performance and success.

The reasoning is in the same document and is unusually open: young people are in dynamic development and are influenced by an environment that can act positively or negatively. School, home and social surroundings are part of that. A points value from the spring of the 8. Klasse is therefore a position report, not a prediction.

Nor does the profile say anything about willingness to make an effort, reliability or motivation, which are precisely the qualities training companies watch for during a Schnupperlehre, the short taster placement in a company.

The same logic sits in every properly built test

The procedure behind Stellwerk has a name: item response theory. The basic idea can be stated without formulas. Every task carries a known difficulty. From the combination of tasks solved and tasks not solved, it is possible to estimate where a person’s ability lies. The number of correct answers is only an intermediate step in that calculation, never the result.

The free test from Kognita rests on the same thing: every task has a stored difficulty, and the result is a position in a reference distribution rather than a percentage. It is a self-assessment and replaces neither Stellwerk nor an assessment by a qualified professional. Anyone who wants to see what such a task looks like from the inside will find the rules behind the figural series in understanding matrix reasoning. And anyone who wants to put a Stellwerk value into the wider context of the search for an apprenticeship can read the overview of aptitude tests for an apprenticeship.