The Churn Cliff: Why Students Leave at the End of a Term

Why do families leave in a cluster rather than one at a time?

Because leaving needs a moment, and a term boundary supplies one. Through the middle of a term the cost of stopping is high. The money is already spent, the routine is set, and telling the centre is an awkward conversation with no natural opening.

At the boundary all three of those costs fall to nothing. The family does not have to cancel anything or explain themselves. They simply do not re-book, and the decision costs them nothing at all.

That is why enrolment looks stable for weeks and then drops. Nothing changed about the teaching in the week the drop happened. What changed is that a decision made privately, sometimes a month earlier, finally became visible to you.

Where your own boundary sits is not a national date. The Indonesian school year runs in two semesters, and detikEdu, reviewing the published calendars, found each provincial Dinas Pendidikan issuing its own, so the assessment window and the break land on different dates in different provinces.

If your students come from more than one province, or from a mix of public and private schools, you do not have one cliff. You have several, a few weeks apart.

The fortnight either side

In the two weeks before a boundary the family is at its busiest and least receptive. Assessments are running, the child is tired, and sessions get missed for reasons that are entirely genuine. It is the worst window there is for asking a parent to commit to another term, and it is exactly when most centres ask.

In the two weeks after, the family is in a different mood and on a different budget. School fees have landed, the timetable has changed, and the household is deciding what the next few months look like. A centre that has not spoken to them by then is arguing against a decision that has already been made.

So the conversation has to happen earlier than feels natural, back when the term is still ordinary and a parent can actually think about it. Whether you get that timing right is a scheduling problem rather than a marketing one. Your enrolment and scheduling records already show which packages end in which week, and the work is putting a reminder against the ones that end together.

What does drift look like in the attendance record?

Drift is not absence. A family that has stopped coming has already gone. What you are looking for is the pattern that arrives first, and it shows up in the attendance record before it shows up anywhere else.

Attendance thins on one particular day. A child who came twice a week is now reliably there on one of those days and unreliably on the other. That is a timetable conflict the family has already settled in someone else's favour.

A missed session gets asked about and then never rebooked. The make-up class a parent requested and did not take is a stronger signal than the original absence, because it says the intention has gone rather than the diary.

A package runs down and sits unused. Sessions left over at the end of a term, with no move to book them, is the plainest signal on this list.

Happy Kamper keeps a timestamped attendance record and tracks the sessions remaining in a package, so the pattern is there to be read. It does not renew the package, and it does not raise anything on its own when the pattern appears.

One sibling stops and the other carries on. That is almost always a budget decision that has been half made and will finish itself at the boundary.

The principle here is not ours. The Dinas Pendidikan in Kabupaten Sleman runs an early warning system for children at risk of leaving school that works by watching attendance data and alerting when a student is recorded as frequently absent. People whose whole job is stopping departures treat the attendance record, not the enrolment list, as the thing that moves first.

A retention number you can work out on paper

Do this monthly rather than per term. A per-term figure arrives after the cliff, which is the one moment it cannot help you.

Take the number of students enrolled on the first day of the month, then subtract every student who was enrolled on that day and is no longer enrolled on the last day. Divide what is left by the number you started with. That is your monthly retention.

Two rules decide the count. A student finishing a fixed programme they signed up to complete is not a leaver, and counting them as one makes a healthy month look bad. A student still enrolled on paper but absent for a month is a leaver whatever the record says, and counting that student as retained is how a centre finds out about a problem a term too late.

The number on its own tells you nothing. Next to last month's, it tells you something. Next to the same month a year ago, it tells you the most, because that is the only comparison that holds the calendar constant and puts your boundary month against your boundary month.

The reason to work it out is not the number. Reichheld and Sasser, writing in Harvard Business Review, found that defection rates act as a leading indicator of profit swings, and that asking departing customers why they went is what surfaces the weaknesses worth fixing. The calculation only tells you which month to go and ask about.

Which three conversations are worth having?

An alert is not an intervention, and the Sleman system is explicit about the difference. The automatic warning there is the first of four stages, and the stage that follows it is mapping the cause, sorted into economic, family and health. A record can tell you a family is drifting, but only a person finds out why.

The first conversation is with the family showing drift, and it is not a sales call. Ask what changed about the timetable, because most drift is logistics, a new school run or a sibling's schedule or a parent's shift, and a different day solves more of it than any discount will. Offering money to a family with a scheduling problem tells them you were not listening.

The second is with the family who has already said no. Almost nobody has this one, because it feels like a conversation with no upside. It carries more information than anything else on this list.

Ask one question, ask it without trying to reverse the decision, and write the answer somewhere you will read it again. Ten of those answers will guide what you change better than any survey you could run.

The third is with the teacher, and it belongs before either of the others. The person in the room usually knows which child has gone quiet, which parent has stopped asking questions at pickup, and which class has lost its energy. None of that reaches an owner unless someone asks for it, and all of it is available long before it reaches a number.

Software will not save the family

Happy Kamper will show a centre its timestamped attendance, what each family has paid and still owes, and what enrolment looks like month by month. Those records are what turn the calculation above into a ten minute job instead of an afternoon, and they are what make drift visible while there is still time to act on it. Centres running this way can see the shape of it on the enrichment centre pages.

It will not tell you a family is about to leave. There is no churn prediction in it, no automatic renewal of a package, and nothing that raises a flag when a student's attendance thins. Those are patterns you go looking for, on a schedule you set yourself.

It also cannot make the call. The record is a prompt to have a conversation, and a centre that reads the record and never picks up the phone ends the term in the same place as a centre that had no record at all.

Software takes away the excuse that you did not know. It does not take away the work.