What a pass actually means
Five questions, written for one specific level of one specific skill, four of them correct to pass. Thirty seconds per question by default, which an administrator can change. That is the entire claim behind a verified level, and it is deliberately narrow: it does not mean a manager agreed, it does not mean years of experience, and it does not mean the person is good at the job.
Four out of five is chosen rather than three or five. One wrong answer is recoverable, so a single misread does not cost a level. Two are not, so guessing does not survive the set. The bar is identical for everyone, at every level, in every company using the product, which is what makes one person's Advanced comparable to another's.
What the four levels mean
Beginner, Intermediate, Advanced and Expert. Each is defined in writing for each skill, and each is tested with a different kind of question rather than a harder version of the same one — recall, application, judgement, trade-off. Asking an Expert something a Beginner could reason out measures nothing about either.
The level being tested is recorded on the attempt itself when it starts. If the level changes while someone is mid-quiz, the goalposts do not move: the sitting is scored against what it was written for.
You climb one rung at a time
If you claim a level you have not yet proven, the assessment tests that claim. If you have already proven what you claim, it tests one level above. It never skips a rung, and at Expert there is nothing left to offer.
Passing above your claim earns the level rather than merely confirming it. XP is owed once per level ever reached: climbing pays the difference between where you were and where you are, and re-verifying a level you already hold pays nothing. The default ladder is 25, 50, 100 and 200 XP for the four levels, and an administrator can change those values per company.
How the questions are written
The rules are explicit and they exist because each one describes a way a question can be answered without knowing the subject. No “all of the above”, “none of the above” or “both A and B”. No negative stems. No absolutes such as always or never in the wrong answers, because they signal which options are wrong. The proficiency level is never named in the question or the options. No trick wording or double negatives: the difficulty has to come from the thinking, not from the reading.
Stems are capped at 160 characters. A question that cannot be read inside a thirty-second timer is measuring reading speed, not skill.
The tell that gets caught automatically
The most common giveaway in a multiple-choice question is length: the correct option is the detailed, careful one and the wrong options are short. A test-taker who notices that can pass without reading the question at all.
So every generated quiz is measured for it. A question is flagged when its correct option is at least 30% longer in words than every wrong option. If two or more of the five carry that tell, the quiz is rewritten — and the correction quotes the count and the rule back rather than asking again more politely, because a bare retry re-rolls the same tendency. Whichever version has fewer tells is the one kept, since a retry that came out worse is not an improvement.
What this method does not do
It does not measure competencies. Ownership, judgement under pressure and the habit of surfacing a risk early are patterns of behaviour assessed from accumulated observation, and no five-question quiz reaches them. That distinction is set out in full on the skills and competencies page.
It does not replace a manager. A verified level says somebody could answer, under a clock, without notes, on one day. It is one input to a conversation about someone, deliberately a small and checkable one, and everything else on a profile remains context.