Disclosure Quality: What a Willingness to Explain Says

Disclosure quality is how readily a company explains its own figures, and it is judged by behaviour rather than by the length of the document. The observable tests are whether an item that flatters the result is disclosed as plainly as one that hurts it, whether a measure keeps its definition from year to year, and whether a direct question receives a quantified answer.

Three things are settled elsewhere. The adjusted figures on Sarvani Coatings Limited, an invented manufacturer of industrial coatings, were worked to the last rupee under earnings adjustment, so the Rs 446 crore, Rs 452 crore and Rs 448 crore figures are used below rather than derived again. The reading of a call, and the idea of a question that gets an answer without a number in it, come from the analysis of earnings call transcripts. And the machinery that puts any item into the statements at all, how a provision enters and later leaves, what a note has to carry, sits with the accounting material. Disclosure quality adds a way of reading the behaviour around those figures, and a very firm line about how far the reading goes.

What is disclosure quality, and what is it not?

The everyday version is a test most people already run on each other. Two shopkeepers on the same street each hand a customer a bill. The first writes one line, "goods", and the total. The second writes each item, the rate, the discount he gave and the two rupees he rounded off. Neither is cheating anyone. But only one of them has handed over something that can be checked, and a customer coming back tomorrow with a question about the discount knows in advance which of the two will have an answer ready.

The two bills are the whole idea, moved into a listed company. Disclosure quality is a reading of behaviour: what an issuer explains, how plainly it explains it, and whether it explains the awkward thing as readily as the flattering one. Disclosure quality is not a measure of volume, and volume and quality are confused so routinely that a longer report is very widely taken for a more forthcoming one. Volume is easy to see and quality is not, so the eye settles on the thing it can see.

Keep one more distinction close, because all three tests lean on it. Disclosure quality is about the document and the conversation around it. Disclosure quality is not about the person who signed either. None of the three tests reads a person, and that boundary is a hard one rather than a nicety.

Is more disclosure the same as better disclosure?

No, and the reason is worth seeing rather than agreeing with. A note to the accountsThe detailed sections published behind the main statements, where one summary line is opened up into the items sitting inside it. Law and the accounting standards between them fix what must appear there. can run to great length, describe a movement as broad based, softer than expected or largely offset, and never once give the figure that would settle it. The reader finishes it knowing the direction of something and not its size. In a model, direction without size is the same as knowing nothing.

So here is the practical test, and it takes seconds. The test is whether the passage could go straight into a model without a further question. If it could not, the length of the passage has not helped at all. That test is blunt, and bluntness is the point, because the failure mode it guards against is being impressed. A long, fluent, carefully written passage feels like disclosure while it is being read. The test asks what the reader was left holding at the end.

The same test runs on Sarvani Coatings Limited. The record carries other expenses of Rs 460 crore for year three, the twelve months to 31 March. Inside that line sit two items already named: a restructuring chargeAn amount charged when a business reorganises itself, folding two operations together or closing a unit down. Whether it may enter the accounts, and at what size, is decided by the standards. of Rs 6 crore, and also a provision write-backMoney once put aside for an expected obligation and released again later, because it is no longer expected to be spent. The conditions permitting a release of that kind sit with the standards. of Rs 4 crore. Both amounts are in the notes and neither is on the face. Because the amounts are there, the charge can be stated as 1.30 per cent of that line and the release as 0.87 per cent of it, and the release alone as 0.90 per cent of the Rs 446 crore of reported earnings before interest, tax, depreciation and amortisation (EBITDA) for the same twelve months. A version of the same note that said the line included certain one-time costs, partly offset by a release, would have used more words and left a reader unable to write any of those three sentences.

What is the symmetry test?

The first of the three tests asks one question. Is an item that flatters the result disclosed as plainly as one that hurts it?

The household version is a school report that lists every prize and mentions the failed subject in a footnote. Nobody has lied. But the leaning of the document is visible, and visible without knowing anything at all about the school.

The symmetry test is the sharpest of the three because it needs no history and no access. A single set of notes already contains items pointing in both directions, and how prominently each one is carried is visible at once. Last year's document is not needed, attendance at a call is not needed, and nobody has to be asked for anything. The evidence sits in the document already to hand, arranged in a way that can be seen.

The direction of an item matters here, not its size. In year three, Sarvani Coatings has three items in the notes. The restructuring charge of Rs 6 crore pushed the reported result down. The provision write-back of Rs 4 crore and an insurance claim of Rs 9 crore, the latter inside other income of Rs 38 crore, both pushed it up. So the symmetry test has something to work on: two items flattering, one item hurting, all three in the same set of notes, and the test simply asks whether the two groups were treated alike.

What is the consistency test?

The second test asks whether a measure keeps its definition, its reconciliation and its presentation from one year to the next. Same name, same build, same table showing how it was reached.

A shop reports its monthly takings, and one month it quietly starts counting the wedding order it delivers next month because the advance has already come in. The number is bigger and nobody has said anything false. The meaning of the number has changed, and without last month's definition beside it there is no way to see the change at all.

Definition drift is close to invisible inside the current document and becomes obvious the moment two years sit side by side. The consistency test therefore needs a reader who kept the previous year's document. That requirement sounds trivial and it is the reason the test is so rarely run. The current document is the one that arrives unbidden. Last year's has to have been saved by somebody who did not yet know they would need it.

On Sarvani Coatings, the honest result is that the test cannot be run at all. Sarvani Coatings has one year of an adjusted figureA measure an issuer publishes alongside the reported one after taking certain amounts out of it. The arithmetic of the ones used here is covered under earnings adjustment., year three, and a definition needs two points before it can drift. So the output is not a pass and it is not a fail. The entry in the file is a note to run the test at the next release, and an instruction to keep this year's document until then.

What is the responsiveness test?

The third test is the simplest to state. When a direct question is put, does the answer contain a figure?

Not whether the answer was polite, or long, or delivered with confidence. Whether there is a number in it. On the year three call, Meghna Iyer, the analyst, asked Ravindra Setlur, the chief financial officer of Sarvani Coatings Limited, why the Rs 4 crore write-back had not been removed from the presented figure when the Rs 6 crore charge had been added back. The answer described the adjustment policy the company applies. The answer contained no figure and no reason specific to that item, and nobody in the room returned to it.

The reply is recorded as a non-answerA reply that addresses the topic of a question without supplying the amount or the specific reason the question asked for. The reading of a call that produced this term is covered under the analysis of earnings call transcripts., with its date, and the writing stops there. A non-answer is information about what was available to be said in that room at that moment, and it is not information about anybody's character, and recording it as the first and not the second is the entire discipline of the reading. There are ordinary explanations that cost nobody anything: the person may not have had the number, the question may have been misheard in a queue of fourteen others, the answer may exist in a note nobody reached for. A transcript cannot rank those explanations. One thing can be recorded. The question went in with a figure attached and came back without one.

How do the three tests come out on year three?

Now the whole reading, on one issuer, one year, in the open. Everything below is drawn from the notes, the call and the guidance for Sarvani Coatings Limited, year three, the twelve months to 31 March.

Taken in turn, the first result is not the one a reader would expect.

On the symmetry test, the disclosure passes, and that is worth stating plainly because it is not the finding a reader expects. The notes carry the Rs 6 crore charge that hurt the result and they carry the Rs 4 crore write-back that flattered it, with the amount, in the same place, at the same prominence. The insurance claim of Rs 9 crore also flattered the result, and it is carried in the same way. Nothing was buried. A reader who came here looking for something to be annoyed about has to record a pass.

Consistency cannot be run. One year of an adjusted figure is on the record, and the test needs two. Consistency goes down as unavailable, with a line in the file to run it at the next release, and the pull towards assuming a result in either direction is to be resisted very firmly. An unavailable test is not a soft fail, and it is not a finding at all.

Responsiveness returns a recorded non-answer. The question about the write-back was asked once, answered with a description of policy rather than a figure, and not returned to. The non-answer gets a line, a date and nothing further.

Why are disclosure and adjustment two different decisions?

Disclosure and adjustment are two separate decisions, and the difference between them is the point least likely to have been met elsewhere.

An issuer can disclose an item fully in the notes and then leave that same item out of its own adjusted figure. The three EBITDA figures for year three, all worked under earnings adjustment, are the proof. The reported figure is Rs 446 crore of EBITDA. On revenue of Rs 2,415 crore, that works out at 18.47 per cent. Management publishes an adjusted EBITDA of Rs 452 crore, or 18.72 per cent, having added the Rs 6 crore charge back. Running that same adjustment in both directions, taking the Rs 4 crore write-back out as well, gives Rs 448 crore, or 18.55 per cent. The distance between the published figure and the two-way one is Rs 4 crore. That gap is 0.9 per cent of the reported EBITDA figure, and it shifts the margin by 0.17 of a point.

Both readings are correct about the same item at the same time, so hold them next to each other. The write-back was disclosed: symmetry passes. The write-back was not removed from the presented figure: the adjustment is one-sided. The disclosure test passes and the adjustment test does not, on one Rs 4 crore item, and that is two separate findings rather than one blurred impression about transparency. A reader who mashes them together into "the disclosure is a bit questionable" has thrown away the only precise thing they were holding, and has replaced two checkable statements with one unfalsifiable one.

Why is strong disclosure in an awkward year worth more?

Because explaining is not equally expensive in every year, and evidence is only worth what it cost to produce.

The household version: a friend who says exactly what he earned in the month he was promoted has revealed very little about himself. The friend who says exactly what he earned in the month he lost the contract has revealed a great deal, and the difference is entirely in what the telling cost.

Explaining a good year is cheap and explaining a poor one is not. So the observations that actually carry information cluster in exactly the years a reader least wants to be reading carefully. That is an uncomfortable shape, because attention naturally drifts the other way. Good year, everything is fine, skim it. Bad year, this is grim, put it down. Both instincts point away from where the information is.

Year three at Sarvani Coatings was, on the record, a comfortable year. Revenue rose 13.9 per cent on year two, EBITDA rose 31.2 per cent and profit after tax rose 41.1 per cent, so every line down the ladder grew faster than the one above it. In a year shaped like that, disclosing a Rs 6 crore charge and a Rs 4 crore release costs the issuer very little indeed. So the symmetry pass recorded above is a real pass, and it is also a cheap one. Write it as both.

What did the issuer choose to lead with?

There is a fourth observation available on this record, and it is not one of the three tests. The fourth observation is about emphasis. Emphasis is what an issuer puts first.

At the start of year three, management guided on two measures. Management guided to high single digit volume growth, and said it expected gross marginWhat is left of revenue once the cost of materials has been taken out, stated as a share of revenue. How this line is built for Sarvani Coatings is covered separately. to hold around the prior year level. The record gives the phrase and not the numbers behind it, so high single digit is read here as the 7.0 to 9.0 per cent band, and the reading is recorded as a reading.

Volume grew 6.0 per cent, 1.0 point below the floor of the guided band and therefore outside it. Gross margin went from 44.0 per cent in year two, being gross profit of Rs 933 crore on revenue of Rs 2,120 crore, to 46.0 per cent in year three, being Rs 1,111 crore on Rs 2,415 crore. Gross margin rose 2.0 points against a guide of holding. So one guided measure was missed and the other was beaten, and the release leads on margin.

Now stop. Which measure was emphasised can be observed; why it was emphasised cannot, and the moment the why is written down the writing has turned to intent. Perhaps margin genuinely was the bigger story of the year. Perhaps the ordering was decided by whoever drafted the release on the night. Perhaps the volume miss was so small that nobody thought it worth leading with. A reader has no way to rank those, and does not need to: "guidance was given on two measures, one was missed, one was beaten, and the release leads on the one that was beaten" is a complete, checkable, dated sentence that any other reader can verify against the same two documents.

What is disclosure quality actually evidence of?

Most readings quietly overreach at exactly this point, so being exact is worth the trouble.

Disclosure quality changes what can be CHECKED, and it does not change what can be CONCLUDED. A forthcoming issuer hands over the material for a harder test, so the range narrows because more work was possible. A reticent issuer leaves less to work with, so the range stays wider. The narrowing is the whole of the finding, and it is a finding about the evidence in hand rather than about the company.

The two axes really are independent, and that is easy to lose. Nothing in a fuller set of notes makes a company better run. A fuller set of notes makes the company more examinable. A well disclosed issuer can be examined thoroughly and something objectionable found, or a thinly disclosed one examined and nothing wrong found at all, and both of those outcomes are entirely consistent with everything set out here.

Where does a reader stop?

Right here, and the line is bright rather than blurry.

Nothing in the record supports a statement about honesty, intent or character. Writing one goes past the evidence, and it is separately a conduct exposure for whoever puts their name to it. Every observation above survives being read out loud next to its source: the notes carry these three items, the presented figure removes one of them and not the other, this question was asked on this date and came back without a figure, guidance was given on two measures and the release leads on one. Every one of those is checkable by somebody else against the same documents. "They are not being straight with us" is checkable by nobody, and it is the sentence that ends the work.

The correct output is a list. The list holds what could be checked, what could not, and why not in each case. A list is a duller deliverable than a verdict, and it is the only one that holds up when a second reader picks it up.