If the data sheet is blank by October, fix the sheet

The trap

We are about six weeks into the school year. The BIP has a data sheet attached. It asks the para to count every instance of the behavior, all day, with the antecedent and consequence for each one. Or it asks for partial-interval recording in 15-minute blocks from bell to bell. The first two weeks, the sheet came back full. Now it comes back half empty, filled in at 2:45 from memory, or not at all.

Then the team meets, and the decision gets made from how the week felt to whoever is in the room.

The usual fix is a data collection training or a reminder email. Sometimes that helps. More often the sheet was never something one adult could run while also teaching, prompting, and keeping a room full of other students safe.

The better question

Not "how do I get staff to take better data," but "what is the cheapest measure that would still lead us to the same decision, and can the adult in the room actually run it during instruction?"

That is a design question, and there is research on it.

Sampling problem behavior usually gave the same answer about treatment

Meany-Daboul, Roscoe, Bourret, and Ahearn took five treatment evaluations that had already been recorded second by second, for four people with autism, ages 7 to 21. Three had stereotypy and one had self-injury. They re-scored the same sessions two cheaper ways. Momentary time sampling means you look up at the end of each interval and mark whether the behavior is happening right then. Partial-interval recording means you mark the interval if the behavior happened at any point in it. Then expert reviewers, and a structured visual inspection method, judged from each graph whether the treatment worked.

Most of the time, the cheaper graphs led to the same answer as the full record. Momentary time sampling matched the duration record a little better, which fits behavior that lasts, like stereotypy or being out of area. Partial interval matched the counted behavior a little better, which fits brief behavior like hitting. The errors were not random, though. When the cheaper methods got it wrong, they usually missed a treatment effect that was really there.

For a school BCBA, that last finding matters more than the method names. A sampled measure can tell you whether a plan is working, but when it errs, it leans toward "not working." Before the team drops a plan because the sampled data look flat, ask whether a small real change could be hiding in that method. One limit changes how far you can push this: the sessions were 5 to 10 minutes in a small room with 10-second intervals. The study does not show that 15-minute intervals across a school day would give you the same answer, so shorten the interval or the observation window before you trust it.

Meany-Daboul, M. G., Roscoe, E. M., Bourret, J. C., & Ahearn, W. H. (2007). A comparison of momentary time sampling and partial-interval recording for evaluating functional relations. Journal of Applied Behavior Analysis, 40(3), 501-514.

Read the full Meany-Daboul et al. article on PubMed Central

Recording three teaching trials was close enough, with one rule attached

Lerman, Dittlinger, Fentress, and Lanagan went back to teaching data for 11 children and teens with autism or developmental disabilities, ages 4 to 15, across 24 new skills taught at three day programs. Therapists had recorded every trial and the prompt level needed. The researchers re-analyzed the same records as if staff had written down only the first trial, or only the first three trials, of each session.

First-trial-only data gave a rough read on how the session went, but it called most skills mastered too early when the rule was two good sessions in a row. It was also slow to show progress. On average, more than twice as many sessions went by before first-trial data showed any change at all. Data from the first three trials tracked the full record closely for mastery decisions. Adding the prompt level to those few trials made the shortcut data show early progress much sooner.

For a school BCBA, that is a data sheet you can actually hand a para: record the first three trials, write down the prompt level, and require three sessions in a row before calling anything mastered. The limit that matters here: these were one-to-one sessions with trials back to back. The authors note that when trials are spread across the day's routines, a small sample may track less well. If the skill is taught in the flow of a general education class, probe it there before you trust the shortcut.

Lerman, D. C., Dittlinger, L. H., Fentress, G., & Lanagan, T. (2011). A comparison of methods for collecting data on performance during discrete trial teaching. Behavior Analysis in Practice, 4(1), 53-62.

Read the full Lerman et al. article on PubMed Central

What to do this week

1. Pull one data sheet that comes back blank, half done, or filled in after the fact, and write the one decision the team needs it to make.

2. Watch the adult who is supposed to run it for ten minutes during instruction, and note when they could realistically make a mark.

3. Swap in the cheapest method that still answers that decision: a look at the end of each short interval for behavior that lasts, a did-it-happen mark per block for brief behavior, or the first three trials plus prompt level for teaching.

4. Write the rule that keeps the shortcut honest: three sessions in a row before mastery, and no plan dropped on flat sampled data without a closer look.

Don't guess, assess

Behavior School now has "Don't guess, assess" tees, mugs, totes, and crewnecks, including a Data Path design with a line graph on the front. If your team could use a reminder in the copy room, [see the Behavior School shop](https://shop.behaviorschool.com/).

Don't Guess, Assess Tee, Data Path

Don't Guess, Assess Crewneck Sweatshirt, Data Path

Where this goes next

I say this in the Transformation cohort a lot: the hard part of school data is not the graph. It is designing a system people will actually run on a busy Tuesday. When the measure fits the routine, the data come back, the graph shows up at the meeting, and the decision comes from the picture instead of the loudest memory of the week.

That is the work inside the School BCBA Transformation Program: goals that generate a data sheet, measures that fit the room, and a review routine that turns data into decisions. The next live cohort meets six Thursday evenings, January 14 to February 25, 2027, 6 to 8 PM Pacific Time. [See the School BCBA Transformation Program](https://behaviorschool.com/transformation-program?utm_source=newsletter&utm_medium=email&utm_campaign=2026-w42)

See the School BCBA Transformation Program

Reply if you want: which data sheet on your caseload is most likely to come back blank this week, and why?

Until next week,

Rob Spain, BCBA, IBA
Founder, Behavior School

Unsubscribe: https://behaviorschool.com/unsubscribe
Behavior School LLC

How this issue was made: Written and reviewed by Rob Spain, BCBA, IBA. Research and drafting may be AI-assisted; source interpretation and final editorial decisions are Rob's. Read the editorial standards.