A parent once told me her son's tantrums had not improved at all in two months. She had the tally marks to prove it: about four a day in July, about four a day in September.

Then she mentioned, almost in passing, that they used to last twenty minutes and now lasted three.

She had been counting the right behavior with the wrong measure. The tantrums had changed enormously. Her data sheet just could not see it.

Count when the behavior is countable

Frequency is a tally. Every time the behavior happens, you make a mark.

It works when the behavior has an obvious beginning and end and each instance looks roughly like the last: requests, hits, interruptions, times leaving the table.

It stops working when instances vary wildly in length, which is exactly what went wrong above. Four brief protests and four half-hour meltdowns produce the same number.

Time it when length is the problem

Duration recording means writing a start time and an end time.

Use it when your honest complaint is about how long something goes on. Tantrums, getting dressed, settling at bedtime, staying with a task.

Duration is also the measure most likely to show early progress, because length usually shrinks before frequency does. A child learning to recover faster is making real progress that a tally will completely miss.

Sample it when the behavior never stops

Some behaviors have no clean edges. Humming, rocking, staying on task, playing near a sibling. There is no moment where one instance ends and another begins.

For these, use interval recording. Split a routine into short blocks, say ten blocks of one minute, and mark whether the behavior happened at all in each block. You end up with something like "occurred in 7 of 10 intervals."

It is an estimate rather than a count, and that is fine. An estimate you can collect during dinner beats a precise measure you cannot.

Two others worth knowing

  • Latency. How long between the request and the start of the response. Useful when your real concern is "he heard me and did nothing for four minutes."
  • Task analysis. A checklist of the steps in a skill, marking which steps were independent. Useful for handwashing, dressing, or any routine you are teaching in sequence.

Choose by what would count as better

The most reliable way to pick is to describe what improvement would look like in a sentence.

"Fewer meltdowns" points to frequency. "Shorter meltdowns" points to duration. "He gets started when I ask" points to latency. "He can do more of it himself" points to a task analysis.

Whichever measure you choose, keep it for at least two weeks before switching. Changing measures mid-stream leaves you comparing two things that were never the same thing, and you lose the baseline that makes the comparison meaningful.

If you are not sure, take frequency and duration together for a week. It is slightly more work and it settles the question.

Measuring something and not sure it is the right something? Book a discovery call and we can look at what you are tracking.