One score is a guess.
Four is a profile.

Yes. If you subscribed on the web, manage or cancel from your account. The iOS app is not for sale on the App Store yet.

No. Your practice history and results are always saved to your account.

Yes, €9.99 billed monthly until cancelled.

No. SLP Command is an independent educational platform, not affiliated with NATO, any Ministry of Defence, or any official examining body. AI-generated feedback is indicative guidance to help you prepare — not an official SLP / STANAG 6001 assessment.

SLP Command trains SLP Level 2 and SLP Level 3 across all four skills — Reading, Listening, Writing and Speaking. Every exercise and every AI evaluation adapts to whichever level you select.

Yes. Your account, progress, scores and Academy status sync automatically, so you can pick up training on any device without losing history.

From your own rated attempts, nothing else. Each answer, submission and spoken response updates an estimate of your ability along with how uncertain that estimate is; items close to your current level carry more weight than ones far from it, and older evidence counts for less than recent evidence. The level shown is bounded by what you have actually demonstrated — it is never an average across skills, because STANAG 6001 levels are not averages.

Because it has not measured enough to give you an honest one. When there is too little evidence about a sub-skill, the platform says so and tells you what would resolve it, rather than showing a plausible figure derived from nothing. A number you cannot rely on is worse than no number, particularly when you are deciding whether to book an exam.

Usually because recent results are below earlier ones, or because evidence ages. That is a measurement, not a verdict on you: language skills decay when they are not used. It is also the most actionable thing the platform can tell you, since recovering something you had is considerably faster than learning it for the first time.

There is one other reason, and it is not about you at all: we sometimes improve how the estimate is worked out. When that happens your number can move without anything about your English having changed. You are told when it does, and why — see My estimate changed and I did not do anything differently below.

Then the change is ours, not yours. An estimate is only as good as the method behind it, and when we find that a method credited progress too easily we correct it — which can move your number even though your English is exactly where it was.

The earlier method nudged your level up a little with every correct answer and down by less with every wrong one. Over enough questions that drifts upward on its own, so the figure could end up above what your answers actually supported. The current method asks a different question: how do you perform on questions at each level? That is the same question the real assessment asks.

We show you the new figure alongside the old one before anything changes, so it is never a surprise. Your history is kept as it was — we do not rewrite numbers you were previously shown — and results from before and after the change are drawn separately, because they are not directly comparable.

That we are no longer confident the estimate describes you today. It says nothing about how good you are — it is a statement about the evidence, which has simply aged.

Confidence answers one question: how far should we trust this estimate right now? So it has four settings, and each tells you something different:

• Reliable — enough recent work to be confident.
• Fairly reliable — supported, but a little more would make it firm.
• Limited evidence — not enough yet. Do more.
• Out of date — your work was solid, but it is old. One recent session is usually enough, and doing more is not what is needed.

Confidence starts to fade after about six weeks without practice in a skill, and each skill ages on its own — practising Reading every day does not keep your Speaking estimate current.

You can still open any skill you like at any time — nothing is locked. But the app opens on a plan rather than a menu, because choosing what to practise is a decision most people get wrong in the same direction: they practise what they are already good at, in long blocks of one skill.

The session is built the other way round. It interleaves rather than blocks, it puts recovery before new material when something has gone out of date, and it never places two productive skills — writing and speaking — back to back, because that is what makes a session unfinishable rather than difficult. Every block on the plan states the reason it is there, so you can overrule it knowing what you are overruling.

Because half of what an instructor is for is saying what can wait. If a skill is current and well evidenced, another session on it today buys you very little, and the time is worth more elsewhere. The app names those skills explicitly instead of quietly leaving them off the plan.

It will never tell you to skip something that is also on today's plan. If that ever happens, it is a bug — report it.

No, and this is deliberate. We can tell you what your evidence supports, how confident we are in it, and how far it is from your target level. Turning that into “72 % likely to pass” would require calibration against real STANAG outcomes that we do not have. People book exams and make career decisions on that number, so we do not produce one rather than produce one that looks credible.

Almost always because the response did not answer the task that was set. Writing assessments rate two things that are genuinely separate: how well you write, and whether you did what the prompt asked. A fluent, accurate, well-organised piece about something other than the question cannot be credited as though it answered the question — an examiner would not do that, and neither will we.

When this happens the feedback says so explicitly rather than leaving you to guess from the number: it tells you the language was not the problem, which part of the task was missed, and what to change. Usually nothing about your English needs to change at all.

The evaluation is produced by AI models against fixed rating criteria, and the criteria — not the model's opinion — are what the score is expressed in. Every rating comes back with the reasoning that produced it, so you can disagree with it on specifics. It remains indicative guidance for preparation: it is not, and does not claim to be, an official STANAG 6001 rating.

A course delivers a syllabus in a fixed order to everyone. This does not have a syllabus: it has a model of you, rebuilt after every session, and it uses that model to choose the next thing. It also does something a course structurally cannot — it tells you when it does not know, and it shows you the evidence behind every claim it makes about your English.