They practise what they are already good at.
Almost nobody arrives here for the love of a language. They arrive because a document requires a number. A Syrian doctor who needs B2 plus the Fachsprachprüfung before the Approbation. A woman who needs B1 for Einbürgerung and has failed Schreiben twice. An Indian software engineer who needs IELTS 7.0 for a visa route and has 6.5 with a 6.0 in writing. A nurse whose employer is paying for the third attempt. Two things are true of nearly all of them: the deadline is external and immovable, and they have already failed at least once.
Someone with Lesen 82, Hören 78, Schreiben 51 will spend four evenings a week doing reading exercises, because reading feels like progress and writing feels like failure. They buy another book of practice tests, do a whole test, score it, get a number, feel bad, and do another one. Nothing in that loop tells them which of their sentences lost marks or why. Twelve weeks later they score within two points of last time, and this is genuinely bewildering to them, because they worked very hard.
- Scoring a mock against the published criteria, per criterion, not as one number
- Classifying every error in writing or a recorded answer by type, and counting them across attempts
- Allocating practice hours in proportion to where marks are actually lost
- Generating targeted drills on those specific error classes, in the exam’s real format
- Spaced repetition seeded from their own errors rather than a generic word list
- Measuring speaking mechanics: speech rate, pause length and placement, filler frequency, lexical range
- Building the countdown schedule against real available hours
- Registration admin: deadlines, centre availability, ID requirements, fees
- Whether they are ready to book
- What an examiner will actually reward in the room
- The anxiety, which for a third-attempt candidate is often the largest single factor
- Whether this is even the right exam at the right level
- The diagnostic
- One full mock under real conditions, scored against the published descriptors per criterion. For IELTS Writing that means four separate results; for ÖSD or Goethe the module- and criterion-level breakdown, not a pass or fail. The output is deliberately not a band: here is each criterion, here is where the marks went, and here are your own sentences quoted next to the descriptor they fell short of. Most candidates have never once seen this. They have seen numbers.
- The error ledger
- Every error in every piece of writing and speaking, classified and counted. After three pieces the ledger says something like: sixty-three per cent of your grammar deductions come from one structure, or you have lost marks for task response in four consecutive essays and never for grammar. That sentence is worth more than a hundred practice tests, and no book can produce it because no book knows what this specific person does.
- The prescription
- Available hours allocated in proportion to where the marks are lost, not evenly and not by preference. Someone losing everything in Schreiben gets a plan that is roughly seventy per cent writing, and they hate it for about ten days. It also has a rule about the strong skill: enough maintenance to hold it, and not one hour more.
- The speaking workbench
- They record the real task under real timing. It comes back with a transcript, the mechanical measures — speech rate, where the pauses fall and how long they are, filler frequency, lexical variety, structures repeated — and the recurring errors mapped to the ledger. It measures mechanics. It does not score their personality, and it says so.
- The countdown and the retake check
- From today to the exam date against the hours they genuinely have, with a fresh mock every three weeks. Plus the modular retake question: several of these exams allow a single failed module to be retaken alone, which can mean a fraction of the fee. The system raises the question and names the office to ask — it does not assert the answer. I have watched someone re-sit an entire exam because nobody told them to ask.
It never promises a score, and never writes their essay.
It produces criterion-level estimates against published descriptors and labels them as estimates, every time. The failure mode here is specific and expensive: a candidate who believes a predicted 7.0 books the exam, pays the fee, books the visa appointment, and scores 6.5. So the readiness signal is deliberately conservative — three consecutive mocks above target on the weakest criterion, not on the average — and even then it is a signal to a person, not a decision.
It never writes their essay and never speaks for them, not even as a "starting point". In the exam room there will be nothing, and a candidate who has been submitting improved text has been measuring the tool rather than themselves. Nothing touches a live exam — no tooling during a test, no assistance in a remote-proctored session. No official past papers are reproduced. And no immigration, visa or eligibility advice: whether a certificate is recognised for a particular Approbation route is a legal question with real consequences, and the system points at the Behörde, the Ärztekammer or a lawyer.
Not the overall band — bands move slowly and hide everything. Movement on the weakest criterion, which is the only number that matters and the one twelve weeks of self-study never moved. Error frequency per hundred words by class over time. The distribution of study hours, which typically starts at seventy per cent on the strong skill and should invert within a month. And whether the ledger is still being updated eight weeks in, which predicts the outcome better than any single mock.
When the timeline cannot work — someone at a solid A2 wanting C1 in eight weeks is not going to get there, and I say the honest number of months in the first hour. When the target level is wrong, which twice meant a lower certificate or a different exam the institution also accepts. When the barrier is anxiety rather than language, because drills make that worse. And when someone wants the score without the language, which is a short conversation.
One full mock every three weeks under real conditions, which is uncomfortable and non-negotiable. Ten minutes after each writing task. And about ten days of doing the practice they have been avoiding.
B1 for the Einbürgerung, failed Schreiben twice, and both times I had studied reading because reading felt better. The first session he showed me my own sentences next to the criteria I had missed, and nobody had ever done that — I had only ever seen numbers. The plan he built was seventy per cent writing and I hated it for about ten days. He also made me ask the centre about retaking only the one module, which nobody had told me was possible, and it saved me most of the fee.
I needed 7.0 for the visa route and I had 6.5 with a 6.0 in writing, and I had done maybe thirty practice tests. The error ledger told me that sixty per cent of my deductions came from task response and almost none from grammar, which is the opposite of what I had assumed and the opposite of what I had been practising. He refused to predict my band as a promise and only said I was ready after three consecutive mocks above target on my weakest criterion, not the average. I got 7.5 in writing.
Tell me what your week looks like.
I'll tell you honestly which parts you can hand over — and which you can't.