The progress report says "Jordan met 80 percent." Last month it said 60. The team celebrates. Then someone asks how many trials. The answer: five last month and five this month, so the change is one trial.
A percent is easy to say and easy to misread. This guide helps you pick the kind of number that fits the goal, and shows what each one can hide.
What the rule asks for
Section 300.320(a)(3) requires the IEP to describe how progress toward the annual goals will be measured. It does not say with which kind of number. That choice is made when the goal is written, and it should be made before the baseline is taken, because the baseline has to use the same kind of number.
The choices a team makes
NCII's guide on behavioral IEP goals says the team determines the measurement by deciding on a tool or approach (such as observation or Direct Behavior Rating), a scale (such as a 1 to 10 rating, a frequency count, or percentage of time), how often to collect data, the context, and decision rules. Utah's guide lists frequency by tally marks, duration by a timer, percentage of successful opportunities, and a rating scale.
| Kind of number | You record | Fits goals about | Example criterion |
|---|---|---|---|
| Percent of trials | Correct and total trials | Accuracy on set tasks | 8 of 10 trials correct |
| Count (frequency) | Events in a fixed period | How often something happens | 2 or fewer in a 30 minute observation |
| Rate | Events per unit of time | Fluency, or uneven observation lengths | 77 words correct per minute |
| Duration | Time per episode or total | How long something lasts | Stays with the task for 10 minutes |
| Rating | A score on a defined scale | Quality or degree | A 3 or higher on a 4 point rubric |
The academic example in NCII's guide is a rate: words correct in one minute.
Percent of trials: show the count
A percent compresses two numbers into one. With 10 trials, each trial is worth 10 points. With 5 trials, each is worth 20. So "80 percent" could mean a very different amount of evidence.
Use the calculator to see it. Enter the number correct and the number of trials. It shows the exact percent, how much one more trial changes it, and a rough range the true level might lie in. The range is a reminder, not a verdict.
Runs in your browser. Nothing you type is sent or saved. The range uses the Wilson score interval, a common way to show how uncertain a percent from few trials is. It is a rough guide, not a verdict.
Report both the count and the total in every note: "8 of 10." Anyone can calculate the percent from that, and nobody has to guess what was behind it.
Jordan's two months, in numbers
Last month Jordan got 3 of 5 trials right: 60 percent. This month 4 of 5: 80 percent. The report says the score rose 20 points. Put both pairs into the calculator and the rough ranges come out at about 23 to 88 percent for the first and about 38 to 96 percent for the second. They overlap almost completely. With five trials a month, one extra correct answer is the whole story, and it cannot tell you whether Jordan improved.
Nothing is wrong with the data. The mistake is in the claim. More trials per session, or more sessions per month, would give the percent something to stand on. NCII's list of proficiency levels includes "three consecutive probes," which shows the same instinct: one good day is not mastery.
Combine counts before you compute the percent
When you total several days, add all the correct answers and all the trials first. If Jordan got 2 of 2 on Monday and 5 of 10 on Tuesday, the average of the two percents is 75, but the combined result is 7 of 12, or 58 percent. The second number is the honest one, because Tuesday carries five times the evidence.
When a percent does not work
Utah's guide says percent-based measures should not be used if whether a behavior occurs is open-ended. Its example: a goal for a student to "raise his hand 80% of the time" is not measurable, because there is no way to count the chances to raise a hand. Fix it by defining the opportunities: "raises a hand in 8 of 10 teacher questions." Or switch to a count in a set period.
Counts need equal time
A count is the simplest record: a tally. But counts from observations of different lengths do not compare. Ten outbursts in 20 minutes and ten in an hour are different. Either keep the observation length the same, or convert each to a rate, such as events per hour. The conversion is division: 10 events in 20 minutes is 30 per hour; 10 in 60 minutes is 10 per hour.
Duration and rating
Duration suits goals where the point is how long: time on task, time to calm down. Record start and stop for each episode, then report the average or the total. A rating, such as Direct Behavior Rating in NCII's guide, merges a rating scale with direct observation. NCII says it is used repeatedly for behavior across a set period under similar conditions, and it prefers objective, valid, reliable measures over teacher anecdotal notes.
Match the measure to the verb in the goal
Look at the verb. "Complete" or "answer correctly" points to percent of trials. "Reduce" or "increase how often" points to counts or rates. "Stay" or "remain" points to duration. "Show quality" points to a rubric with defined levels.
| Goal wording | First choice | Note |
|---|---|---|
| "Answers with 90 percent accuracy" | Percent of trials | Set the number of trials |
| "Completes within 5 minutes" | Duration | Record each episode |
| "Initiates conversation more often" | Count per period | Same observation length |
Hold the measure steady
Once you pick a measure, keep it. Changing from counts to ratings midstream breaks the line on the graph. If you must change, note the date and start a new baseline. For more on timed and sampled methods, see interval recording and time sampling, and for behavior data in particular, behavior frequency and duration data.
A practical habit helps here: write the measure and its unit at the top of every data sheet, such as "correct out of 10 trials" or "outbursts per 30 minute observation." A substitute teacher or a new paraprofessional then collects the same kind of number without a conversation, and your graph stays clean.
Limits of this guide
The sources describe measures. They do not rank them or give one right answer. Your team and your district's procedures decide. If the goal says one thing and your tool measures another, change the tool, not the goal's meaning. The starting number comes from how to set an IEP baseline. Terms are in the glossary.