One-Off Items: How Frequency Undermines the Label

A one-off item is an amount the issuer describes as belonging to a single period rather than to the ordinary run of the business. The label is a claim, not a fact, and counting is what tests it. Line the same note up across several years and count the ones carrying an item of the same kind. Four years out of five is a running cost wearing a label that keeps it out of view.

Three things sit outside the test itself. The split between what repeats and what does not is set out under recurring and non-recurring earnings. How a provision is recognised, measured and later released belongs to the accounting standards, which the Institute of Chartered Accountants of India holds. Naming an item is enough for the count. How it reached the statements is not. And every rupee below belongs to one invented issuer, Sarvani Coatings Limited, whose year three ladder and whose two disclosed items can be worked on paper. The test itself is made entirely of counting.

What is a one-off item, and who actually hands out that label?

A one-off item is an amount presented as belonging to one period alone. The definition stops there, and what it leaves out matters. Nothing in it says the amount is large. Nothing in it says the amount is unusual in the market. All it says is that somebody has presented the amount as belonging to one period, and who that somebody is matters more than anything else about the item.

No accounting framework hands out this label: it is applied by the issuer, in the issuer commentary and in the adjusted figures the issuer chooses to publish, and no rule anywhere requires an amount to be described that way. The standards settle that the amount gets disclosed at all, and which line it must sit under. The standards do not decide what the amount gets called in a presentation, and they do not decide whether a reader should take it out of a base. The second decision belongs to the reader, and most readers hand it straight back to the issuer without noticing.

Sarvani Coatings Limited, in the year ended 31 March that this record calls year three, disclosed a restructuring chargeMoney booked against reorganising the business, say when a plant is shut or two depots are merged into one. Whether it may go into the accounts at all, and at what amount, is fixed by the standards. of Rs 6 crore sitting inside other expenses of Rs 460 crore, and a provision write backWhat shows up in a later year when money kept aside for something turns out not to be needed, and the earlier charge is unwound. When that unwinding is permitted is governed elsewhere. of Rs 4 crore that reduced the same line. Both are in the notes to the accountsThe detailed sections that follow the headline statements and break each summary line into its parts. What has to appear there is decided by law and by the standards. and neither appears on the face. Neither is wrong, hidden or improper. The question arrives only afterwards: given that the charge is there, correctly, and described as belonging to year three alone, is the belonging believable?

What is the label actually worth on its own?

Very little, and the reason is structural rather than personal. A label saying one-off is a statement about years that have not happened yet: it asserts that this amount will not come round again. Nobody can know that. An amount taken out of the base makes the base look steadier and often larger, so the label is also a statement made by the party whose reported figure improves when a reader accepts it.

Said that plainly, the reasoning usually goes wrong at the very next step. Naming an incentive is not making an accusation, and the correct response to an incentive is a test rather than a suspicion. Anybody sitting in that seat would face the same incentive. A chief financial officer who genuinely believes a plant closure will never recur writes exactly the same word as one who does not, and nothing in the wording distinguishes them. So the person is not what gets read. The count is.

The household version is quick. A relative asks to borrow money and says the shortfall this month was one-off, the school fee and the scooter repair landing in the same week. The claim may be completely true. Asking what last month looked like, and the month before, calls nobody dishonest. The question is asked because the word one-off is a forecast, and forecasts are checked against records rather than against faces. If the last four months each had a different one-off, that says something about the household budget and nothing whatever about the relative.

What is the frequency test, and what does it need before it can run?

The test is one instruction. Line up the same note across three to five years and count how many of those years carry an item of the same kind. The instruction is the whole test. There is no formula, no threshold and no clever adjustment. The bareness is one reason the test gets skipped: it does not look like analysis.

Now the requirement, and it is worth saying flatly rather than in passing. The frequency test needs several years of the same disclosure, so a reader holding one release cannot run it at all, and that single sentence explains why the test is so rarely run in practice. A quarterly result arrives, the pressure is to publish a view within a day, and one release contains exactly one year of notes. The test that would settle the question is not available on the day the question arrives.

Which puts the real work earlier than most readers expect. The count only exists if somebody wrote the first item down, in a year when there was nothing to count and no reason to bother. The pattern a reader wants to check is made of years already gone, so a reader who starts recording only once a pattern looks suspicious has started too late by definition.

What does the same charge in four years out of five actually mean?

It means the amount is a cost of doing business, and it means nothing else. An item of the same kind appearing in four years out of five is not an exception to the ordinary running of the business; it is a description of the ordinary running of the business. A company that reorganises continuously genuinely incurs reorganisation costs continuously. Nothing improper has to have happened for the pattern to appear, and usually nothing has.

The reading is about where the amount belongs, never about who put it there, and a reading that slides from the first to the second has stopped doing analysis and started doing something else entirely. The finding is a classification finding. The amount looks like a running cost of the business, so it belongs in the base a forecast starts from rather than outside it. The claim is checkable, and it is the only claim the count supports.

A delivery business replaces a vehicle every year out of a fleet of ten. Each replacement is genuinely a separate decision, each one is a distinct event with its own reason, and calling any single one a one-off is not a lie. But a budget for next year with no vehicle in it will be wrong. The fleet has ten vehicles and they wear out on a schedule. The item is one-off. The class is not, and the base has to carry the class.

Does the same count run on gains, or only on charges?

The count runs identically on a gain, and almost nobody runs it. Ask a reader to check whether a restructuring charge keeps reappearing and they will do it. Ask the same reader to check whether a provision write back keeps reappearing and it does not occur to them, because a credit does not feel like something that needs policing.

But a write back that appears every year is a recurring feature of the business exactly as a charge that appears every year is. So is a disposal gain that appears every year. So is an insurance recovery that appears every year, which is worth noticing here because Sarvani Coatings disclosed one of Rs 9 crore inside other income of Rs 38 crore in year three. One appearance says nothing. The count is the only thing that would.

Running the test on charges alone rebuilds, inside the analyst's own work, exactly the one-sidedness this whole sequence exists to catch. The one-sidedness is the uncomfortable part. A reader can be diligent, sceptical and completely one-directional at the same time, and the result looks like rigour while producing a base that is systematically too high. The rule is simple to state and slightly annoying to follow: the count runs on both directions, or it runs on neither.

How large does an item have to be before it is worth the trouble?

Wrong question, or rather, incomplete. An item is not small because it is small against revenue. Size is meaningless until the reader names the denominatorThe number being divided by. A change in it can make the same amount look negligible or substantial without anything about the amount itself having changed., and the denominator has to match the question actually being asked.

Take Sarvani Coatings Limited in year three and put the Rs 6 crore charge against four different bases, all of them from the same published year. Against revenue of Rs 2,415 crore it is 0.25 per cent, near enough to nothing. Against other expenses of Rs 460 crore, the line it sits inside, it is 1.30 per cent. Against reported earnings before interest, tax, depreciation and amortisation, or EBITDAA rung on the profit ladder taken while depreciation, interest and tax are all still to come. Its contents were fixed long before this sequence and are simply used here. of Rs 446 crore it is 1.35 per cent. And against the Rs 106 crore by which EBITDA rose between year two and year three, it is 5.66 per cent.

The same Rs 6 crore is 0.25 per cent of one thing and 5.66 per cent of another, a difference of nearly twenty three times, and the arithmetic did not change once. Only the question changed. If the question is whether the company is large, revenue is the right denominator. If the question is whether this item is doing work inside the growth rate a forecast will start from, the increase is the right denominator, and it is the only one of the four that answers that question.

One period rule, and it is not pedantry. The Rs 106 crore is a one year movement, year two to year three, and the Rs 6 crore charge is a year three amount. Both belong to compatible spans and can sit in the same fraction. A two year or three year change with a single year item placed over it manufactures a percentage that describes nothing at all. Stating the period beside every figure used makes this mistake hard to make.

What does the test return on Sarvani Coatings Limited, run honestly?

Now put the two halves together on the case record and be exact about where the work stops. Take the sizing first: that part is fully available. The restructuring charge is Rs 6 crore and the write back is Rs 4 crore, both year three amounts, and the useful denominator is the Rs 106 crore by which EBITDA rose over that one year, from Rs 340 crore to Rs 446 crore.

About one rupee in eleven of the year three increase in EBITDA is accounted for by two items that each look negligible beside revenue, which is a long way from immaterial. The finding is a statement about arithmetic and about nothing else. Nothing in the arithmetic says the items were wrongly described, wrongly measured or wrongly presented. The share says only that a reader forecasting from this year of growth is forecasting from a number that contains them.

Now the count itself, and here the honest answer is short. The Sarvani Coatings record carries one year of item disclosure. The count is therefore one out of one, and the frequency test returns nothing yet, a result rather than a gap in the work. One out of one is not evidence of a pattern and it is equally not evidence against one. A not yet is the reading returned when the test has not had enough years to run on, and writing that sentence down is the correct output.

A not yet leads to a date rather than a shrug. The charge and the write back are recorded with their year, their size, their direction and the note they came from, and beside them a line saying the test becomes runnable at the next release. Two more years of that and the count is worth something. Started now, the record gives the reader in year five evidence; skipped now, it leaves that reader four empty boxes.

What happens to an item accepted as genuinely one-off?

Two things, and readers reliably do the first and skip the second. A base is what can be expected again, and a genuine one-off is not that, so the amount comes out of the base. Then it gets written down.

The record needs four fields and no more. The year, so the count has something to count. The size, so it can be put against a denominator later. The direction, charge or gain, so the count runs both ways. And the note reference: the reader who picks this up in three years can then find the original wording rather than trusting a summary of it. A fifth line is optional, the description the issuer used, in the issuer words rather than the analyst's.

The list is the instrument that makes the frequency test possible in two years, so a reader who removes an item without recording it has quietly destroyed their own future evidence. The recording is where the habit usually fails. Removing an item feels like the work. Removal is half the work, and it is the half that produces nothing durable. The recording is what turns a private judgement into something a second reader can check and a later reader can count.

What does a repeated item not support as a conclusion?

Anything about a person. The pull in the other direction is strong: a reader who has just watched a strip fill up to Rs 30 crore feels that something has been caught.

A pattern raises a question about where an amount belongs, it says nothing whatever about intent, honesty or judgement, and writing it as though it did is both wrong on the evidence and a conduct exposure for the writer. Wrong on the evidence, because the count cannot distinguish a business that genuinely reorganises every year from any other explanation, and it was never built to. A conduct exposure, because published research about a named issuer sits inside obligations that are real and enforced, and the route to those obligations is in the block further down.

There is also a plain professional reason. The classification finding is defensible in front of anybody: here is the note, here is the count, here is the base used and why. The conduct claim is defensible in front of nobody, because the evidence is not there and no amount of counting will produce it. Giving up the sentence that cannot be supported keeps the one that can.

Where every figure came from, and which body holds each rule

Where the numbers come from. Every rupee above belongs to the Sarvani Coatings teaching record: the year three ladder, the two disclosed items and each percentage worked from them. The five-year strip above is a drawing built to show what a count would catch, not a history of this issuer. The four bodies below hold the wording and the rules behind an item of this kind, and the rows say which body holds which.