How to calculate percent accuracy in speech therapy
Percent accuracy is the number of correct responses divided by the number of trials, times 100. The formula is the easy part. What makes the number trustworthy is deciding what counts as a trial, how cued responses are scored, and how sessions are combined for a report.
The formula
percent accuracy = correct responses ÷ total trials × 100
A student who produces /s/ correctly in 14 of 20 words scored 14 ÷ 20 = 0.70, or 70%. Round to a whole number. A progress report gains nothing from 70.4%, and the extra decimal suggests a precision that 20 trials can't deliver.
Decide what counts as a trial before the session
- One trial is one opportunity to produce the target. Define it in the same terms as the goal: one word, one sentence, one answer to a question, one conversational turn.
- Keep the definition fixed for the life of the goal. If September's trials are single words and November's are sentences, the graph shows a drop that isn't regression. When you move up a level of difficulty, note it, or start a new goal step.
- Count every attempt you planned to count. Leaving out the attempts that "didn't really count" quietly inflates the score.
- Keep teaching trials and data trials apart. Many SLPs take goal data only on probe trials (no model or feedback just before the response) and treat the rest of the session as practice. Others score every trial. Either works; mixing the two within one goal doesn't.
- Decide once how to score self-corrections. Scoring the first attempt is the more conservative choice. Whatever you pick, use it every session.
How many trials?
Fewer trials make each one count for more. With 5 trials, a single error is the difference between meeting an 80% criterion and missing it.
| Trials | One error costs | Scores you can get near 80% |
|---|---|---|
| 5 | 20 points | 60%, 80%, 100% |
| 10 | 10 points | 70%, 80%, 90% |
| 20 | 5 points | 75%, 80%, 85% |
| 50 | 2 points | 78%, 80%, 82% |
Articulation drill can easily reach 20 to 50 trials or more per goal in a session, and more practice is better for the student anyway. Language goals where each response takes longer, such as retelling a story or answering inferential questions, may only get 5 to 10. That's fine, but look at several sessions together before drawing conclusions from them.
Scoring cued responses
Most IEP goals name a support level: "with 80% accuracy given minimal cues." The data has to show whether that is what happened. There are two common ways to handle it.
- Record the cue level with every trial. Score each response correct or incorrect and note the level of help it took: independent, minimal, moderate or maximal. You can then report accuracy at each level, or the share of responses produced independently. This is the more informative option, because it shows a student needing less help even before accuracy changes. The cue levels guide covers how to define and record the levels.
- Count only responses at or below the goal's cue level as correct. If the goal says minimal cues, a correct response that needed a full model is scored incorrect. This is simpler on paper, but you lose the record of how much support the student needed.
Whichever you choose, write the rule at the top of your data sheet so a substitute or co-treating SLP scores the same way.
Combining sessions for a progress report
This is where most reported percentages go wrong. Averaging session percentages gives a 5-trial session the same weight as a 20-trial session.
| Session | Correct | Trials | Accuracy |
|---|---|---|---|
| Oct 6 | 18 | 20 | 90% |
| Oct 13 (fire drill, 10 minutes) | 2 | 5 | 40% |
| Oct 20 | 15 | 20 | 75% |
| Pooled | 35 | 45 | 78% |
Add up the correct responses and the trials across the sessions, then divide: 35 ÷ 45 = 78%. Averaging the three percentages instead gives (90 + 40 + 75) ÷ 3 = 68%, because the short session counts as much as the full ones. Report the pooled figure with the trial count beside it: "78% accuracy (35 of 45 trials)." The count tells the reader how much data is behind the number.
To describe where a student is now rather than across the whole period, use the most recent sessions and say so: "averaging 85% over the last three sessions."
Mastery is a pattern, not a session
One good day isn't mastery. Goals usually define it as a percentage held across several sessions, such as 80% across three consecutive sessions. Read the goal's wording closely: "three consecutive sessions" and "three of four sessions" are different criteria, and a data sheet should make it easy to check whichever one the goal uses.
Common mistakes
- Making the stimuli harder or easier mid-goal without noting it.
- Rounding each session before combining them. Combine the raw counts, then round once.
- Reporting a percentage without the number of trials or sessions behind it.
- Totting up the score at the end of the session from memory. Mark each response when it happens.