What the score is measuring
A pronunciation score (typically 0–100) estimates how close your spoken sentence is to a native model, broken down into accuracy (individual sounds), fluency (pauses and pace), completeness (dropped words and swallowed endings) and prosody (stress and intonation). What it measures is "how clearly was this sentence said" — not "how good is your English": vocabulary and grammar errors are simply out of scope.
Rule one: the trend, not the single score
Single-sentence variance is normal: say it a bit faster, hold the mic a bit further, add background noise — the number moves. The correct read is the trend across days on the same content: the same shadowing passage averaging 62 on day one and 78 on day five is real progress. Staring at a one-day dip produces anxiety and zero practice value.
Rule two: in a low sentence, find the one sound
A sentence scoring 55 is not wrong in every sound — usually one or two drag the whole thing down: ship said as sheep (vowel), a lost final consonant (workin' for working), misplaced stress (DE-vel-opment for de-VEl-opment). The right move on a low sentence is to replay the recording against the transcript and find the one sound that hurts comprehension — not to re-drill the whole sentence ten times.
| Common problem | Typical symptom | Fix |
|---|---|---|
| Vowel confusion | ship/sheep, bad/bed merged | Minimal-pair contrast shadowing |
| Lost final consonant | workin', go din' | Slow down and land the last sound, then speed back up |
| Misplaced stress | Wrong syllable in long words | Check the dictionary stress, over-articulate the right syllable |
| Flat intonation | Whole sentence monotone, robotic | Shadow emotion and stress contours |
Rule three: one fix per round
Practising twenty problems at once means practising none — under load, the brain consolidates one change at a time. Each round (one conversation or one shadowing pass), choose one target sound or one prosody feature; when it's done, put it in the review queue. Next round, review that item first, then open a new target. This is also why FluentDojo files corrections into a review queue: so the last problem isn't washed away by the next one.
FAQ
Does a low pronunciation score mean my English is bad?
No. The score measures only how clearly that sentence was spoken; vocabulary, grammar and content are out of scope. A fluent but accented speaker may score mid-range; a clean-sounding speaker reciting memorised lines may score high. Treat it as pronunciation feedback, not a proficiency test.
How fast will my score improve?
It varies, and nobody can promise a date. Watch process signals instead: the same material's multi-day average rising, fewer low sentences of the same type, and the sound that used to drag you down now landing consistently.
Should I chase 100?
No need. The bar for communication is clear, not perfect — natural native speech is full of blends and swallowed sounds. For most learners the sensible zone is: everyday sentences consistently clear, with extra polish for professional or exam settings.
Does the score relate to IELTS Speaking?
No direct relationship. IELTS is a human, holistic rating across four dimensions; AI pronunciation scoring covers only part of the pronunciation dimension. It cannot predict any exam score — for exam prep, follow the official descriptors.
Which matters more, shadowing or scoring?
The combination. Shadowing supplies correct input; scoring supplies corrective feedback. Shadow without scoring and you can't hear your own errors; chase scores without shadowing and you have no model to copy. FluentDojo's shadowing module carries per-sentence scoring, so both live in one flow.