How your readiness is calculated

Step by step

1 · Each session produces one percentage

A timed paper, a short set or a practice run gives you a number of correct answers out of the total. Writing and speaking are self-marked against the published criteria, so those percentages are your own honest judgement of your work — which is a real limitation, and we say so below.

2 · Your last three sessions in a skill are averaged

Not your best. Not your most recent. The mean of the last three, so one lucky paper cannot flatter you and one bad evening cannot sink you.

3 · The average is compared to 70%

70% accuracy in this app is what we treat as the NCLC 7 line. It is a working threshold chosen to be demanding rather than reassuring — it is not published by any exam board, and it is the single most arguable number on this page.

4 · Each skill is given one of four labels

Label	Means
On target	Averaging 70% or more
Close	Between 50% and 70%
Focus here	Under 50%
Needs more data	Fewer than two sessions — deliberately not counted as a pass

5 · The verdict is set by your weakest skill

Never by the average. The exam requires NCLC 7 in listening, reading, writing and speaking separately, so three sevens and a five is a five. If any skill is unmeasured, the verdict says partly measured rather than giving you a green light it cannot justify.

What this number cannot tell you

This matters more than the method itself.

• It is not an official score, and it does not predict one. The real exams are marked by trained examiners on their own scales — 300 points for TEF reading, 699 for TCF, and so on. We do not convert to those scales, because a conversion would imply a precision we do not have.
• Writing and speaking are marked by an AI examiner, not a human one. It reads your work against that task's published criteria and scores it. It is consistent and it is strict, which is more than most self-marking manages — but it is not an accredited examiner, and on a borderline piece a real one may disagree with it.
• Typing speed is measured, but it is not one of the four skills. Since June 2026 the TCF Canada writing paper is taken on a computer with the spell checker disabled, so producing accented characters at speed is a real exam skill — and the course drills it. Your words per minute is shown in Progress and deliberately does not move the NCLC readiness verdict, because it is not French ability.
• Conversation practice is marked the same way, and with the same limit. You speak, a French speaker answers, and at the end it scores what you said against the exam criteria — but from the words, not the sound.
• A spoken answer is marked from words, not from sound. After recording, you play your answer back and type what you said, and it is that text which gets marked. Some browsers can write it down for you automatically while you record — mostly phones — and you can switch that on, but it is off by default because on many computers it stops the recording capturing anything at all. Either way the score covers your content, vocabulary, structures and grammar, and says nothing about your pronunciation, accent or hesitation, which carry real marks in the exam. That part still needs your own ear or a French speaker.
• You can still mark yourself. The option sits beside the AI marker on every task, and it is what you get when you are offline or have used up the month's tutor requests. If you mark yourself generously, this app will happily agree with you.
• It measures accuracy, not exam-day performance. Nerves, tiredness, an unfamiliar test centre and a bad night's sleep are all real and none of them are here.
• The content has been written for this course. It is modelled on the published formats and question types; it is not retired exam material, and no preparation provider has that.
• A percentage is not a level. It tells you where to spend the next hour. That is what it is for.

Why we show it anyway

Because the alternative is worse. Most people preparing alone have no measurement at all: they practise what they enjoy, feel vaguely better, and book on a hunch. A rough number that names your weakest skill and is honest about its own error beats a confident feeling every time.

The number is there to direct your effort, not to predict your result. If it ever reads as a promise, we have written it badly — tell us.

How the practice material is built

• Formats follow published specifications. TEF Canada listening is 40 questions in 40 minutes and reading 40 in 60; TCF Canada is 39 and 39. Papers in the app are exactly those lengths, or they are not offered as papers.
• Mock papers are assembled, not recycled. A numbered paper must be at least 60% material you have not already seen, or the app will not number it. Sitting a paper you half-remember measures your memory, not your French.
• Every answer has an explanation, including why the tempting wrong option is wrong.
• Automated checks run on every change — answer keys, paper lengths, duplicate questions, and whether the correct answer is spread evenly across A, B, C and D rather than sitting on one letter a guesser could exploit.
• Native-speaker review. The French here is written for this course, not reproduced from a real exam. All of it has been reviewed by French teachers — the listening, reading, writing and speaking material, the vocabulary and grammar drills, and all six stages of the beginner course. That review is what a bank of automated checks cannot give you: whether the French sounds like French, and whether an answer marked correct really is. If you still find an error, send it to us and we will fix it and credit you.

See the number on your own results

Start with 3 days free. Card required to begin; cancel before it ends and you pay nothing.

Start your 3-day trial