For speech-language pathologists
For speech therapists: see the practice between sessions
Updated
In short
You would set the target sounds and practice words. The child wears Saylo between sessions. You would see daily counts per sound (correct, error and unsure) so practice between sessions becomes visible instead of guessed. Saylo is in development and hasn’t been validated with children; therapists get early access from spring 2027.
The gap Saylo is aimed at
Research protocols for speech sound disorders often use two to four sessions a week, while many services run once a week, and US school SLPs report a median caseload of 50 students (ASHA 2024 Schools Survey). Practice between sessions fills that gap, but it’s hard to know how much happened or how it went. The research behind Saylo.
How you would select targets
- Pick the target sound and practice words, with each word’s expected error form (“rabbit” and “wabbit”).
- Decide when home practice starts. Saylo is meant for after the child can produce the sound with you, not for establishing it.
- Set practice goals and criteria yourself; Saylo won’t hard-code them.
How uncertainty and data quality would be shown
| Score | What it means |
|---|---|
| Correct | The attempt sounded like the target, for example “rabbit”. |
| Error | The attempt sounded like the expected speech sound error, for example “wabbit”. |
| Unsure | Saylo couldn’t tell. It gives no buzz, and unsure attempts are counted separately, never as errors. |
- A clear “not enough data” state instead of a trend line built on a handful of attempts.
- Counts optionally by word (to pick words to drill) and by part of the day (school vs home).
Wear time, parent notes and performance are kept apart
How long Saylo was worn, how many practice words it heard, how confident it was, and how accurate the child was are different things. Saylo is designed to show them separately rather than collapse them into one score, and parent notes are labelled as parent notes, not as measurements.
What Saylo won’t do
- Diagnose, screen, or change a child’s targets or treatment on its own. Clinical decisions stay with you.
- Keep recordings. It scores on the device and keeps counts only, so there’s no audio to review.
- Claim accuracy or outcomes it hasn’t shown.
Current limitations
- Saylo hasn’t been tested with children. Its scoring has only been tried on synthetic voices, which isn’t clinical validation.
- The best published automated R classifiers agree with clinicians about as well as clinicians agree with each other, but in older children; younger children are untested.
- Automated scoring of lisps is much weaker, which is why S comes second and only if a pilot supports it.
- Noise, dialect and conversational speech all make automated scoring harder.
What a pilot would evaluate
- Agreement between Saylo’s scores and several SLPs’ ratings, by sound, age, word position and dialect.
- Whether families keep practising past the novelty period, and whether reviewing the data takes an SLP a minute or two per child.
- Whether children’s gains hold after the necklace comes off.
- Whether the buzz changes how much children want to talk.
Get involved
Therapists get priority: early access from spring 2027, before families. When you apply, you can also tell us if you’d like to advise on Saylo or take part in a pilot. Speech therapists get early access from spring 2027, families join a founding-family beta in summer 2027, and Saylo is planned to launch in early 2028.
Questions
Which sounds will Saylo support?
R is planned first. S will be added only if a pilot with SLPs shows Saylo judges it accurately enough.
How accurate is it?
We don’t have accuracy numbers with children yet. Before launch we plan to compare Saylo’s scores with several SLPs’ ratings and publish the results.
Can I listen to the attempts?
No. Saylo keeps no audio by design, which is why it shows uncertainty openly instead of hiding it.