Almost every school tracks attendance. Almost none of them use it. The check-in kiosk fills a database that nobody reads until a parent calls to cancel. By then the decision is made. The point of attendance data is not the record. It is the warning that fires while you can still change the outcome.
Logging is the easy 80%. The alert is the 20% that matters
A student who is about to quit almost never announces it. They come less often. Two classes a week becomes one. One becomes one every other week. Then a month goes by and the membership is a line item the parent finally notices on a statement. The drift is visible in the attendance data weeks before the cancellation. The question is whether anyone is looking.
Nobody is looking, because looking means someone remembering to pull a report, sort it, and cross-reference it against who is paying. That work never survives a busy week. So the data sits there, complete and useless.
Make check-in cost one tap
The early-warning system only works if the data is complete, and the data is only complete if check-in is effortless. If logging attendance takes an instructor more than a second, it gets skipped on the busy nights, which are exactly the nights the data matters most.
- Self check-in kiosk. Tablet at the door, tap your name or scan a code. Students do it themselves.
- Roster tap-through. Instructor opens the class roster on a phone, taps present. Ten seconds for a full mat.
- No paper. A clipboard that gets transcribed later is a clipboard that doesn't get transcribed.
The patterns that predict churn
Not every absence means anything. A kid with the flu misses a week and comes right back. The signal is in the shape of the attendance, not a single gap. Three patterns are worth an automated flag:
1. The frequency drop
A student whose weekly attendance falls and stays down. Three times a week for two months, then consistently one. That is not a sick week. That is disengagement, and it is the single most reliable predictor of a cancellation.
2. The hard stop
A previously consistent student who simply stops. Two weeks of zero after months of regular attendance. The window to win them back is short and closing.
3. The pre-test fade
A student who drifts right before a belt test they were on track for. Often a confidence problem, sometimes a schedule problem, almost always solvable with one conversation if you catch it.
A single absence is noise. A change in the pattern is signal. Build the alert on the pattern, not the gap.
What the alert should do
When a student trips one of those patterns, the system should put them in front of a human with context, not bury them in a report. The right output is one line in the daily staff inbox: the student, the pattern that fired, and a suggested action. "Maya hasn't been in for 12 days. Was on track for her green belt. Call the parent." That is a task someone can actually do on a Tuesday morning.
The intervention is the whole point
An alert that nobody acts on is just a slower version of doing nothing. The reason to catch the drift early is that early interventions are cheap and effective. A text from the instructor. A quick "we missed you" from the front desk. A nudge about the belt test they were close to. None of it works after the cancellation. All of it works in the two-week window the data hands you.
Retention is not a campaign you run once a quarter. It is a steady habit of noticing who is drifting and saying something before they are gone. The attendance data already knows. The job is to make it speak up.