How we score your English level
Nest English gives you a CEFR level estimate from an adaptive test that runs entirely in your browser. This page explains exactly how that number is produced, what it can support, and where it stops. There is no black box here — if you are going to act on a score, you should be able to see how it was made.
1. Every item carries a level and a difficulty
Each question is tagged with a CEFR band from A1 to C2 and a difficulty value. The bank holds over a thousand items across grammar, vocabulary, reading, listening and speaking, so the test can keep finding an item close to wherever your level turns out to be.
2. Your ability is re-estimated after every answer
The engine uses Bayesian estimation (expected a posteriori) over a Rasch-style model. In plain terms: it holds a probability distribution over your possible levels and updates it with each answer, rather than counting correct answers. Two consequences matter. It never returns an impossible score. And it produces a genuine error margin — the ± you see in your report is real statistical uncertainty, not decoration.
3. The next question is chosen to be informative
After updating the estimate, the engine selects an item whose difficulty sits close to that estimate. An item you would almost certainly get right, or almost certainly get wrong, tells us very little. Choosing near your estimated level is what lets a short test say something useful.
4. The test stops when the level is clear
The test ends when confidence in a CEFR band reaches its threshold, or at the ceiling — typically between 24 and 40 questions. There is a hard floor as well, so a badly-behaved short session cannot produce a confident-looking result.
5. The score is mapped onto public reference scales
Your composite score is placed on CEFR bands and on approximate IELTS and TOEFL ranges using publicly referenced comparison tables. These are approximate public comparisons for guidance only. Nest English is independent and not affiliated with Cambridge, IELTS, ETS/TOEFL or any official provider.
6. Skills and topic areas are reported separately
Beyond the overall level, the report estimates each skill and each grammar or vocabulary area on its own, with its own reliability indicator. An area measured by three questions is shown with a wide uncertainty band; one measured by fifteen is shown narrower. We would rather show you a wide band honestly than a precise-looking number we cannot support.
What this result cannot do
- It is not a certified score and cannot replace an official exam.
- It is unproctored — nothing verifies who is answering.
- Item difficulties are reasoned and simulated, not yet empirically calibrated against a large sample. This sets a ceiling on accuracy, and we would rather say so than imply precision we have not earned.
- The speaking check is browser-based word matching, not certified pronunciation assessment. Grammar, vocabulary and reading are the most reliable indicators of your level.
What we deliberately do not do
There is no analytics, no tracking, no advertising and no visitor counter. Your answers, your score, your report and your certificate are generated on your device and are not sent anywhere. This is a real constraint we accept, not a marketing line: it means we cannot tell you how many people took the test, and we would rather be unable to answer that than track you.