Expected Credit Loss: Provisioning Before the Default Arrives

An expected credit loss is the slice of what customers still owe that a business does not expect to collect, recorded as a provision before any named customer has defaulted. Recording it early keeps the receivable figure close to what will really arrive instead of waiting for bad news. The estimate is usually built with a matrix that puts a loss rate on each ageing bucket, and judgement enters at every one of those rates.

Here is what sits underneath that. Nobody can tell in April which schools will fail to pay by March. But almost anybody who has run a book of credit customers for a few years can tell roughly what proportion of it never arrives, and can tell that the proportion is small for fresh invoices and large for old ones. The rough proportion and its rise with age are together enough. The name of the customer who will default is not needed in order to record, honestly and in advance, that some of them will. Recording the loss before the name is known is the whole idea. Everything else is machinery built on top of it.

Anjani Stationers Private Limited, an invented supplier of stationery to schools, raised its provision from Rs 3,00,000 to Rs 9,00,000 in a single year. The Rs 6,00,000 charge splits into a part that is arithmetic on facts and a part that is judgement, and a rising provision can be read two fair ways until the ageing report chooses between them.

What is an expected credit loss?

An expected credit loss is an estimate of money already earned and already invoiced that the business now thinks will never turn up. The estimate is recorded as a loss allowanceA figure held against an asset to bring its reported value down to what is expected to be recovered. The asset stays on the books at its full amount and the allowance is shown alongside it, so a reader can see both. sitting against the receivable, not as a change to the receivable itself, so a reader can always see both the full amount billed and the amount management thinks will arrive.

An expected credit loss is a statement about a population of invoices, not about any one invoice, and a statement about a population can be recorded before a single customer has missed a payment. Think about how a hospital plans its blood bank. The hospital cannot name the person who will need four units on Tuesday. From years of records it can still say roughly how many units a week of this size will consume, and it stocks accordingly. Being unable to name the patient is not a reason to keep the fridge empty. The same logic runs through a receivables book. Anjani Stationers cannot name the school that will fold, but it has three years of collection history, an ageing report, and a very clear sense that an invoice sitting unpaid for eight months behaves differently from one raised last week.

Two things it is not, and both matter. The balance is still there and the business is still chasing it, so a provision is not a write-off. And it is not a prediction of a specific default; it is a weighted view of many possible outcomes across many invoices. A provision of Rs 9,00,000 against Rs 95,00,000 of billing is not saying that Rs 9,00,000 of invoices are dead. The provision is saying that across the whole book, once good outcomes are weighed against bad ones, Rs 9,00,000 is the amount the business does not expect to see.

Why provide before anyone has actually defaulted?

Because the alternative is to report a receivable already known to be too high, and then take the whole correction in one lump on the day the bad news becomes undeniable. The older approach did roughly that: a loss was recorded once there was objective evidence that a particular balance had gone bad. The incurred approach was tidy and easy to audit. Its one large defect was that it reported the best news for as long as possible and the worst news all at once.

The expected approach and the incurred lossThe older way of recognising credit losses: nothing is recorded until there is objective evidence that a particular balance has gone bad, such as a customer entering insolvency or breaching a payment plan. approach usually disagree about when a loss is reported, not about how much loss there is in the end. Follow a single bad invoice through both. A school stops paying in June, negotiates through the winter, and finally shuts in the following year. Under the incurred approach almost nothing is recorded until the shutting, and then the full amount lands in one period. Under the expected approach a small amount is recorded from the day the invoice ages past its terms, more as it ages further, and by the time the school actually closes most of the loss is already sitting in the allowance. Add up the two paths over three years and they arrive at the same place. The two paths differ in which year carries the pain, and therefore in which year's profit a reader is looking at.

The timing is why the shift matters to the reader rather than only to the person preparing the accounts. Under the older approach, a business whose collections were quietly deteriorating could report untouched profit for two years and then a single catastrophic year, and nobody reading the first two years had been misled by any individual number. Under the expected approach the deterioration shows up as it happens, in small pieces, in the year it is happening. The information arrives earlier and in smaller instalments. Earlier and smaller is less comfortable and almost always more useful.

How does a provision matrix actually work?

A provision matrix is the practical machine that turns an ageing scheduleA report that splits what customers owe by how long each invoice has been outstanding, usually in bands such as not yet due, one to thirty days late, and so on. The report is produced straight from invoice dates, not estimated. into a number. The receivables book is cut into buckets by how overdue each balance is, a loss rate is attached to each bucket, and the products are multiplied out and added. Cut, rate, multiply, add: that is the entire method. The matrix is popular for ordinary trade receivables because it needs nothing a small business does not already have: invoice dates, balances, and a few years of collection history.

The matrix works because how late an invoice already is turns out to be the single most informative thing known about whether it will ever be paid. The same fact is why the loss rate climbs steeply down the buckets rather than sitting flat. Ordinary life teaches the same lesson. A friend who owes two hundred rupees from last Tuesday will almost certainly pay. The same friend, the same two hundred rupees, ten months later and after four unanswered messages, is a different proposition, and anyone would already treat it as one without doing any arithmetic. A matrix is that instinct written down and applied consistently to every balance instead of only to the ones that happen to be remembered.

Anjani Stationers bills its schools on sixty day credit termsThe period a seller gives a buyer to pay after delivery. Sixty day terms mean the invoice is not late until sixty days after it is raised, so the not yet due bucket holds recent billing that is behaving perfectly normally., so anything sitting inside sixty days of its invoice date is not late at all and belongs in the first bucket. Here is the year two matrix in full. Read the balance column first for the shape of the book, then the rate column, then check that the last column adds to Rs 9,00,000.

No accounting standard, regulator or published study sets a loss rate for anybody. A business derives its own rates from its own collection history and its own view of what is coming. The rates in the table are assumed, and their shape is the realistic one for a supplier selling to schools, where almost everything eventually arrives and the exceptions are old and few.

What did the ageing profile do between the two years?

Whether the rising provision is believable turns on the ageing, and the ageing is a fact rather than an opinion. Put the two years' ageing schedules side by side and one thing jumps out before any arithmetic.

Every rupee of the Rs 17,00,000 growth in Anjani Stationers' receivables landed in a bucket that was already overdue, and the not yet due bucket did not move at all: Rs 34,00,000 in year one, Rs 34,00,000 in year two. Sit with that for a moment, because it is a stronger statement than the days sales outstanding figure ever made. Revenue grew from Rs 2,40,00,000 to Rs 2,70,00,000, a rise of 12.5 per cent, and yet the amount of fresh, perfectly normal, not yet payable billing sitting on the books at the year end was identical. The whole of the growth in receivables is in balances the schools were already late on. The over 180 day bucket alone more than doubled, from Rs 5,00,000 to Rs 12,00,000.

Two details make the picture sharper. The Sunrise Public School group, the largest customer and the slowest payer, accounts for Rs 8,00,000 of the Rs 12,00,000 sitting beyond 180 days, so two thirds of the worst bucket belongs to one relationship. And Rs 2,50,000 of accrued revenue for a delivery made in the last week of March, not yet invoiced, sits inside the not yet due bucket. An earned but uninvoiced balance belongs exactly there. Neither of those is an opinion. Both come out of the sales ledger.

Where does judgement enter, and how much of the number is it?

Judgement enters in exactly one column of the matrix. Vague statements that provisioning is subjective are not much use to anybody, so precision about which column matters. The buckets are a choice, the balances are a fact, the multiplication is arithmetic. The loss rates are the estimate, and everything anybody argues about lives there.

The loss rate for a bucket is built from two things: what actually happened to similar balances in the past, and a forward looking adjustment for what is expected to be different, and only the first of those leaves a trail somebody can check. The historical part is genuinely constrained. If the business has three years of records showing that four per cent of balances that reached six months late were never recovered, it cannot claim thirty per cent without saying why. The forward lookingTaking account of conditions expected in the future rather than only what has already happened. In a provision this means adjusting a historical loss rate for what is known or reasonably expected about the period ahead. part is the loose one. The forward looking adjustment is where a preparer says that the coming year will be harder than the last three, and it is where a reader has to decide whether to believe them.

How much of Anjani Stationers' Rs 9,00,000 is judgement? One counterfactual answers that, and the answer is uncomfortably concrete. Take the year two book, in its year two ageing profile, and apply the year one loss rates to it. The result is Rs 5,23,000. So of the Rs 6,00,000 increase, Rs 2,23,000 came from the book itself getting bigger and older, and that part is arithmetic on facts. The other Rs 3,77,000 came from raising the rates, and that part is judgement. A little over a third of the increase is checkable and a little under two thirds is a decision.

Four things constrain the judged part, in descending order of usefulness. The business's own history comes first. A rate wildly out of line with three years of recovery data has to be explained. Consistency across periods comes next. A rate that moves conveniently in one direction every year gets noticed. The requirement to disclose the matrix comes third. A reader who can see the buckets and the rates can rebuild the number. And the auditor, who will ask for the working. None of those makes the rate objective. All of them make it awkward to be silly with.

What happens when a balance is finally written off?

A provision and a write-off look similar and behave completely differently, and the difference is where readers most often get tangled. A provision is an estimate held against a balance that is still there. A write-offRemoving a balance from the books entirely because the business has given up on recovering it. The customer may still legally owe the money; the business has simply stopped reporting it as an asset. is the removal of a balance the business has stopped expecting anything from at all.

When a balance that has already been provided for is written off, the gross receivable falls, the allowance falls by the same amount, the net figure does not move, and profit is not touched at all. The loss was already taken in the year the provision was raised. The last of those is the whole point. The charge to profit happens when the estimate is made, not when the paperwork catches up. A deposit left with a landlord works the same way. The month it becomes clear the deposit is not coming back is the month it has really been lost. Whatever letter arrives two years later changes nothing about the tenant's finances; it only tidies the records.

Anjani Stationers wrote nothing off in year two, and that is precisely why the movement in the allowance and the charge to profit are the same Rs 6,00,000. Where a business does write something off during the year, the two figures separate: the allowance moves by the charge less the write-off, and a reader who reads the movement as the charge will get the wrong number. Here is the same idea shown forward. Suppose in year three Anjani Stationers gives up on Rs 2,00,000 of the oldest Sunrise invoices, with the allowance standing at Rs 9,00,000 and nothing else changing.

What is Anjani Stationers' provision, and what does the increase say?

The allowance stood at Rs 3,00,000 at the end of year one and Rs 9,00,000 at the end of year two. Nothing was written off in between, so the whole movement is a charge: Rs 6,00,000 against year two profit. Because raising an allowance moves no money at all, that same Rs 6,00,000 is added straight back at the top of the operating section of the cash flow statement, alongside depreciation. The charge is real for profit and invisible to the bank.

Against gross receivables of Rs 95,00,000 the allowance is 9.5 per cent, where a year earlier it was 3.8 per cent of Rs 78,00,000, so Anjani Stationers has gone from expecting to lose about one rupee in twenty six to expecting to lose nearly one rupee in ten. A shift of that size in one year deserves to be stated as plainly as that. Set next to the ageing, the change is also not a surprising one. A business whose overdue buckets absorbed the entire growth in its receivables and whose worst bucket more than doubled would look careless if its provision rate had stayed still.

How does a lender or an analyst actually use this number?

Nobody reads a loss allowance for its own sake. Three different people read it for three different reasons, and knowing which of the three a given reader is amounts to most of the skill.

A lender lending against receivables cares about the allowance because it usually lends against a borrowing baseThe pool of assets a lender is willing to advance money against, after excluding the parts it does not trust. For receivables this normally means excluding balances beyond a stated age and applying a percentage to the rest. that already excludes anything beyond ninety days, so a shift into the old buckets shrinks what the business can draw long before it shows up in profit. Work Anjani Stationers through it. The two oldest buckets hold Rs 27,00,000 of the Rs 95,00,000. On a facility that excludes balances more than ninety days late and then advances eighty per cent of the rest, the eligible pool is Rs 68,00,000 and the draw is Rs 54,40,000. A year earlier the excluded buckets held Rs 15,00,000, the eligible pool was Rs 63,00,000 and the draw was Rs 50,40,000. So the receivables book grew Rs 17,00,000 while the money that could be borrowed against it grew only Rs 4,00,000. The lender is also watching any covenantA condition written into a loan agreement that the borrower promises to keep meeting, such as holding a ratio above a stated level. Breaking one usually gives the lender the right to demand repayment or renegotiate. expressed on net receivables, because a rising allowance pulls that figure down without a rupee moving.

An analyst reads it as a consistency test rather than as a number. Nobody outside the business can know whether 9.5 per cent is the right rate, so the rate itself is never the question. The question is whether the provision rate moved in the same direction as the ageing and by a proportionate amount. A provision rate that stays flat while the old buckets swell is the thing worth asking about, and so is a provision rate that jumps while the ageing is unchanged. Here the two moved together, the unremarkable case.

And Meera Rao, running operations, reads the matrix as a work list. The Rs 12,00,000 sitting beyond 180 days is not an accounting entry to her; it is a set of specific schools, mostly one group, that somebody needs to visit. The matrix is too often treated as a reporting chore. The same table, read on a Monday morning, is the most useful ranked list of who to call that the business produces all year.

Anjani Stationers Private Limited, Chitra Binding Works Private Limited, Anjani Kulkarni, Meera Rao and the Sunrise Public School group are invented.
Educational material. Not advice on any investment, tax, budget or market position.