Sessions & data collection
A session is one therapy meeting. Each student gets a panel with their own goal — mixed groups are the norm, not an edge case — and every attempt you score is a trial: outcome, cue level, teach/probe mode, the stimulus word, and a timestamp, all in one tap.
The scoring pad
- ✓ / ≈ / ✗ — correct, approximation, incorrect. Approximations count as incorrect in the percent but report separately.
- Cue chips (Ind/Vis/Vrb/Mod/Tac) — the prompting hierarchy. The selected chip sticks, so every following trial is tagged. Independent and cued accuracy chart separately, because fading cues IS progress even when the percent doesn’t move.
- Teach vs. Probe — probes are short uncued checks of true ability; mastery is judged on probe data when it exists, so practice never inflates a progress report.
- Undo — mis-taps happen; nothing locks until the session ends.
- 🎤 Voice scoring (Pro) — hands in the play activity? Tap the mic and say “correct”, “close”, or “no” — one phrase per tap, never an open microphone (a therapy room is wall-to-wall PHI). The app echoes exactly what it heard so a misheard word can’t silently score, “no” words are checked before “yes” words so “not correct” can never land as a ✓, and the command words are yours to edit in Settings → AI & voice — “yep” and “nope” are valid clinical vocabulary. Uses the platform recognizer: offline-capable, free, no AI quota.
Six ways to measure
Each goal picks its data type, and every report downstream speaks that unit:
| Mode | For |
|---|---|
| Trial accuracy | articulation, most language goals |
| Frequency count | fluency (%SS), behavior tallies |
| Duration | sustained phonation, on-task time |
| Interval recording | attention and behavior sampling (with a beep/vibrate cue) |
| Rating scale | voice quality, pragmatics rubrics |
| Task analysis | multi-step checklists |
Reduction goals (where a LOWER number is success) are handled correctly — the app never marks a child “mastered” for getting worse.
Beat the Monster
The first data-generating game: from a student’s ⋮ menu, the drill becomes a comic monster fight. The word comes from a goal-matched deck; the student says it; you judge — a correct word lands a BAM! (the monster shakes and loses power), an approximation glances, a miss gets BLOCKED! — and every judged turn records a real trial with the same outcome, cue level, and stimulus a plain drill would. Ten clean hits win, with a fanfare. There’s deliberately no error sound: a game must never make a child dread the tone. After a victory, the game asks how it was — real reactions and ideas reorder what gets built next.
Ending a session
End session stamps the duration, drafts each student’s note from the data, and shows the summary. Signing a note locks it — corrections from then on are addenda, which is what an audit-safe record looks like.