A boy in class 8 stopped coming in the second week of August. Ask the school in October when it started and nobody can tell you the date. The class teacher knew he had been absent a lot. The accounts clerk knew the second-term instalment had not come. The maths teacher knew his marks had fallen off a cliff in the unit test. Each of those three people knew one third of a story, and none of them ever met to tell it.
The school finds out properly when someone in the office is preparing the outstanding-fees list and realises the family has stopped answering. By then the boy has been out for two months, has found something else to do with his days, and the conversation that would have worked in the third week of August is now a much harder one.
The information was there the whole time. It was recorded, daily, in ink, by a teacher who did her job. It was simply never read.
What a register is for, if not for the government file
Attendance gets marked in every school in the country, every working day. It gets marked for the file — for the inspector who might come, for the annual data return, for the meal count where that applies. It is a compliance artefact, and as a compliance artefact it works perfectly.
What almost no school does is read it as a signal about a particular child. The data is complete and the information extracted from it is close to zero. A register in a cupboard has the same practical value as no register at all, with more paperwork.
The gap between recording and reading is where every avoidable dropout lives. Closing it does not require a new philosophy of education. It requires the register to be a thing you can ask questions of.
Attendance percentage against the classes actually held
The first thing to fix is the denominator. When a teacher says a child is at sixty-two percent, the immediate question is: of what? Calendar days? Working days? The days that section actually had class?
Batches do not all meet the same number of times. A period gets cancelled for a sports day. One section loses a week to exam duty. A coaching batch runs on an irregular schedule that changes with board dates. Percentages calculated against a school-wide calendar quietly punish children in one batch and flatter children in another, and once teachers spot that, they stop believing the number entirely.
This is why attendance should be a per-batch session opened on a date, rather than a column in a monthly grid. The session is the record of a class that actually happened. Mark the whole batch present or absent in one action, correct the handful of exceptions, and the percentage per student is computed against the sessions that batch really held.
Two distinctions inside the marking are worth more than they look. Late is not absent — a child who is chronically late usually has a transport problem or a household that needs them in the morning, which calls for a completely different conversation than a child who is not coming at all. And excused is not unexcused: a fortnight lost to typhoid is not a dropout signal, and a risk list that cannot tell the difference will flag every sick child in the monsoon. That is exactly how teachers learn to ignore the list. When the medical certificate arrives two days later, the record should be amended, and the amendment should be visible rather than erased.
Three signals that arrive before a dropout: absence, marks, fees
A child almost never drops out on a Tuesday. It is a drift, and it announces itself for weeks in three places at once.
- Attendance changing shape — the child who ran at ninety percent all of last year and is now at seventy. The change matters more than the level. So does the pattern: consecutive days, or a run of Mondays and Saturdays, means something a raw percentage hides.
- Marks falling against the child's own past rather than against the class. A student who sat around twelfth and is suddenly thirtieth has told you something; a student who has always been thirtieth has not.
- Fee instalments slipping. This is the most honest of the three, because families protect school fees hard and stop paying them late only when something has genuinely broken at home.
Any one of these alone is ordinary noise. Two of them arriving in the same month is a different situation and deserves a phone call that week.
The reason schools miss the combination is organisational, not analytical. Attendance sits with the class teacher. Marks sit with the exam department. Fees sit with accounts, in a separate room, under someone who is not supposed to discuss a family's money with teachers. Three people each hold one signal, each doing their job correctly, and the child falls through the space between them.
Building a risk view teachers will trust, with no black box
There is one hard requirement for any risk score in a school: the teacher must be able to see why a child was flagged. A score that arrives as a number with no explanation gets ignored, and teachers are right to ignore it. They have thirty-five children and no time to investigate an oracle.
So the score should be built from the three visible components — attendance, fee delays and performance — and clicking a flagged child should show which of the three fired and by how much. When a teacher can look at a flag and say yes, that is the girl whose father lost work in March, the system has earned its place. When she cannot, it becomes another screen nobody opens.
- Every component visible, with the school setting its own thresholds rather than inheriting someone else's
- No behaviour, attitude or family-background inputs — unverifiable, unfair, and they turn a work-list into a prejudice
- No ranking children against each other; risk is measured against the child's own baseline
- Never shown to the child or written on anything the parent receives
And be clear about what the score is. It is not a prediction, and it should never be described as one inside the staffroom. It is a prioritised list of conversations to have this week — nothing more sophisticated than that, and nothing less useful.
The intervention record: what we did, and when
A flag with no record attached produces the worst outcome of all: the same child surfacing next month, the same teacher having the same conversation, with no memory that it already happened twice.
So every flag needs an outcome written against it — what was done, by whom, on what date, what the parent said, and what was agreed. Two lines is enough. The categories are few and every school already uses them: a phone call, a home visit, a meeting with the guardian, a fee rescheduling, extra coaching, or a referral to someone outside the school.
This pays off twice. Class teachers change every academic year, and without a record the new teacher inherits the child but not the history — she starts from zero on a case the school has already worked for two terms. And if a child does eventually leave, the school can say precisely what was tried and when, which matters both for its own honesty and for anybody who asks.
Talking to the parent before the pattern hardens
Timing decides this conversation more than skill does. In week three the child is still a student who has been absent. By week eight the household has reorganised around his absence — he is helping at the shop, she is minding a younger sibling while their mother works — and the school is now asking a family to give up something they have started to depend on.
Three things make the early conversation work. Open with the child, not the register: what you have noticed, not what the rules say. Do not lead with the fee — if money is the problem the parent knows it better than you do, and raising it first turns a conversation about a child into a demand. And meet the family in the format they can actually manage. A parent on daily wages cannot lose a Tuesday morning to be told their son is absent; a scheduled phone call or a video meeting at eight in the evening is not a lesser meeting, it is the one that will actually happen.
End with one concrete thing and a date on it. Not a promise to improve. He comes back Monday and the class teacher calls on Friday. She sits the retest on the fourteenth. The instalment moves to the fifth of next month. Write the agreed action down against the meeting, because an agreement nobody recorded is an agreement nobody will follow up.
Class-teacher routines that take five minutes a week
None of this survives if it adds an hour to a teacher's day. It has to fit inside the day she already has.
- Mark attendance in the first two minutes of the period, on the spot — bulk-mark the batch, then fix the exceptions. Never from memory at four o'clock.
- On Friday, look at the week for that batch: who missed two or more days, and whose pattern changed rather than whose percentage is lowest.
- Pick one name. Not twenty. One phone call made this week beats a list of twenty nobody rings.
- Escalate rather than carry. The class teacher is the sensor, not the intervention department.
That last line is the one schools get wrong. Teachers are given the responsibility for retention and none of the authority — they cannot waive a fee, cannot arrange transport, cannot commission remedial classes. Asking them to solve what they can only detect is how the whole routine quietly dies by November.
What the head of school should see every Monday
Twenty minutes, one screen, four questions. Today's attendance across batches, the students currently flagged, fees outstanding, and what is coming up in the test calendar.
- Which children were flagged last week, and what is recorded against each of them
- Which flagged children have no intervention recorded at all — the most important list in the school
- Which batches are running below the school's own attendance line, because that is a timetable or teaching problem, not a child problem
- Which families show both a fee delay and an attendance drop, since that pair is rarely a coincidence
The moment attendance becomes a teacher's performance metric, the register becomes fiction — and the early-warning system dies with it.
That warning is worth taking seriously. If class attendance percentage is used to grade teachers, absent children start getting marked present, and every downstream signal is poisoned. Measure the school on interventions recorded and closed, not on attendance achieved. One is a behaviour you want more of; the other is a number people will simply produce for you.
The mechanics are ordinary. Attendance sessions that take one tap per batch and compute a percentage against classes actually held. Test results that show a child against their own history rather than only a rank. Fee instalments and outstanding tracked in the same place as everything else. A transparent risk view built from those three, with the intervention written against the flag. A parent meeting scheduled with an agenda and an action item that someone owns.
The register was never the problem. Nobody reading it was.