How a Mirpur coaching centre found the 14% of fees it was quietly losing

Nasrin runs a coaching centre in Mirpur. Around 120 students across nine batches, four teachers including herself, subjects from class six maths to HSC physics. Monthly fees, mixed payment methods, plenty of discounts for siblings and long-standing families.

It ran on three registers — attendance, fees, exam marks — plus a diary for everything that didn't fit. By her own description, it worked. Which was true, in the sense that classes happened and most money arrived.

The question she couldn't answer

Her husband asked what the centre should have collected in the previous month. Not what it did collect — the bank told them that. What it should have collected.

Producing that number meant listing every enrolled student, applying each student's fee, subtracting each agreed discount, and accounting for mid-month joiners and leavers. Across nine batches and three registers it would have taken most of a day, and it would have been an estimate anyway.

Without that number, the collected figure means nothing. There is no gap to investigate, because there is no expectation to compare against.

Where the money actually was

When the first proper monthly expectation was produced, the gap was around fourteen percent. None of it was theft. It broke down into three depressingly ordinary categories.

Ghost enrolments — roughly half the gap

Students who had stopped coming but were never formally marked as left. Some had told a teacher verbally; one had told another student. They sat in the register as active, so they inflated the expected figure — and, worse, nobody had chased their genuinely-owed final month because nobody was certain they had left.

This cuts both ways: the expectation was too high and real money was uncollected.

Partial payments recorded as complete — about a third

A parent pays 1,500 of a 2,000 taka fee and says they'll bring the rest Thursday. The teacher writes "paid" because the parent is trustworthy and Thursday is soon. Thursday passes. The balance now exists only in one teacher's memory, and memory does not survive a month.

Undocumented discounts — the remainder

Sibling discounts and hardship reductions agreed verbally, sometimes years earlier, by whoever was at the desk. Nobody disputed that they were right to grant. But an undocumented discount means the expected total was wrong from the start, which makes every reconciliation after it meaningless.

She didn't have a collection problem. She had a problem knowing what collection should have been.

The rollout, and the thing that had to happen first

Batch structure before software

Her nine batches had grown organically and two of them overlapped almost entirely — the same students, different days, different fees. No system can represent that cleanly because it isn't clean. Two weeks were spent rationalising to seven well-defined batches, each with one schedule and one fee, before anything was entered.

This is the step people skip, and skipping it means encoding the existing mess into a database where it becomes harder to fix, not easier.

Enter students batch by batch, not all at once

One batch per evening, seven evenings. Each teacher entered their own batch, which meant each teacher had to confront who was actually still attending. Four ghost enrolments surfaced during data entry alone, before a single report existed.

Attendance first, fees second, exams last

Attendance is daily, immediate and obviously useful, so it built the habit. Fees followed once a month of attendance data made the roll trustworthy. Exams came last because they matter three times a year, and starting there would have meant three months before anyone saw value.

What changed

• A monthly expected figure exists. Produced in seconds, compared against collections, with the difference itemised by student rather than sitting as a lump.
• Leaving is now an event. A student marked inactive stops accruing expectation and triggers a final-dues check. This alone closed most of the ghost-enrolment loss.
• Partial payments show as balances. The Thursday promise is visible next month instead of evaporating.
• Discounts are attached to students. Whoever granted it, whenever, it is now attached to the record rather than to a memory.

What we'd do differently

1. Fix the batch structure first, always. Two weeks felt like a delay and was the highest-return part of the project.
2. Have teachers enter their own students. Delegating data entry to one person would have been faster and would have surfaced none of the ghost enrolments.
3. Don't start with the reports. Owners want the dashboard on day one; the dashboard is worthless until a month of honest daily data sits under it. Sequence for habit, not for enthusiasm.
4. Write down every discount, even the obviously fair ones. Especially those — they're the ones nobody thinks to record.

The transferable lesson

For any small business collecting recurring fees — coaching centres, gyms, tuition, societies, subscription services — the leak is almost never at the point of collection. It is in the absence of an expected figure to collect against. Build the expectation first, and the collection problem usually turns out to be four or five nameable, fixable gaps rather than a general suspicion.