Education & Childcare · Scenario 37

Cram School / Tutoring

Families enrol in a wave each spring and quietly drift away by winter, and the school never sees it coming. We predict which students are about to leave and step in while it still matters.

Method · Churn early-warning modelling

The situation

A tutoring school’s economics are retention: a student who stays two years is worth many times one who leaves after a term. But enrolments are celebrated while quiet withdrawals only show up when the revenue dips. The early signs — slipping attendance, missed homework, a dropped grade — are visible in the data weeks before a family cancels.

Nobody watches those signals systematically, so the school keeps refilling the top of the funnel at high cost while losing students it could have saved with a timely call.

Spring wave
enrol then drift
Unseen
until revenue dips
Weeks early
the warning signs
No model
of who will leave

Where we dig for the truth

We build an early-warning model from attendance, progress and engagement that flags students at risk of leaving — weeks before they go.

Attendance & punctualityTest scores & progressHomework completionPayment & renewal historyParent communicationCourse & cohort
Student retention by enrolment cohortShare still enrolled, by month0255075100M0M2M4M6M8M10M12Before early-warningAfter early-warning

Most attrition happens in a slow slide, not a single event — exactly the pattern an early-warning model catches in time to act.

Our approach — Student Churn Prediction & Early Warning

A churn model scores each student’s risk from attendance, progress and engagement, surfacing the slipping ones early. Teachers and the front desk get a weekly at-risk list with the reason, so the intervention is specific — a tutoring tweak, a parent call, a schedule change — not a generic retention email.

Acquisition spend is rebalanced toward the channels and cohorts that produce students who actually stay, so the school grows on retention, not just refilling.

From refilling to retaining1Gather the signalsCombine attendance,progress and engagementper student.2Score the riskModel who is mostlikely to withdraw, andwhy.3Intervene earlyGive staff a weeklyat-risk list with thereason to act on.4Buy who staysShift acquisitiontoward cohorts thatretain.
Withdrawal rate by risk factorShare who leave within a term0%16%31%47%63%28%Falling attendance24%Dropped grade19%Missed homework22%No parent contact53%3+ signals

A cluster of signals predicts withdrawal far better than any single one — and gives staff a clear, ranked list of who to reach first.

What changes

Same teaching, students kept instead of lost. Representative for a multi-class cram school.

Representative 90-day movementAnnual retention41%69%▲ +28 ptsStudent LTV$2.4k$4.1k▲ +71%At-risk saves / term~0~40▲ +40Annual revenue$880k$1.16M▲ +32%
Where the growth comes from+$0.28Mannual revenueHigher retention50%Timely interventions30%Better-retaining acquisition20%
Why this is not "social media management"
We didn't tell this school to enrol more students to replace the ones leaking out the back. We predicted who was about to leave and helped keep them. Churn modelling is the cheapest growth in education — the student is already in the building.

Frequently asked questions

How do you reduce student drop-off at a tutoring school?
We build an early-warning model from attendance, progress and engagement that flags at-risk students weeks before they leave, then give staff a weekly ranked list with the reason — so the intervention is a specific call or schedule change, not a generic email.
What signals predict a student leaving?
Slipping attendance, a dropped grade, missed homework and a lack of parent contact each raise risk; a cluster of them predicts withdrawal far better than any single signal. A churn model combines them into one score.
Isn't enrolling more students the answer?
Replacing students who leak out the back is the most expensive way to grow. Keeping the ones you have is far cheaper — they are already enrolled. Book a marketing audit.

Want this run on your numbers?

Send your attendance and progress data and we’ll flag the students about to drift away.

Book a marketing audit
Back to Education & Childcare