Governance Signals: What Research Can Legitimately Read

A governance signal is something disclosed about how a company is controlled that can be pointed at, given a size, and tested against an alternative account. Sarvani Coatings Limited, an invented coatings maker, counts four independent directors among nine, and buys Rs 18 crore from a connected party, 1.38 per cent of its cost of materials. Both are readable. Neither comes with a reason attached.

Sizing comes first, and sizing does most of the work of keeping a reading honest. A single transaction is sized before it is discussed, and a change in a stake is sized before it is discussed. The unsettled part is what comes after the sizing. Once three or four sized governance observations sit in front of an analyst, what is that analyst actually entitled to conclude from them, and what happens to the rest of the work when the answer is not very much? The second question has a precise answer that most treatments of governance never reach.

What makes a governance observation readable at all?

The word signal is doing real work here and not decoration. A signal is an observation that changes what a reader believes. The definition immediately splits everything that can be said about a company's governance into two heaps, and the smaller heap is the useful one. Most governance observations, honestly examined, change nothing that was believed before they were made, and noticing that is a finding rather than a disappointment.

Consider a landlord a tenant is thinking of renting from. The building has a written maintenance agreement pinned up in the lobby. Three neighbours, over chai, convey that the landlord is difficult. The first thing sits in a document that could be photographed. The second is an impression: real, possibly accurate, and completely unusable if the tenant ever has to explain to somebody else why a higher deposit was paid. The distinction is not about which one is true. The difference is which one somebody else can go and check.

The same line runs through a research file. A filingA document a listed issuer sends to the exchanges so that it reaches every holder at the same moment rather than one person at a time. from Sarvani Coatings Limited, or its shareholding patternThe statement a listed issuer publishes at intervals showing which class of holder holds what share of the company., or a note attached to its accounts, is a place a second reader can go. An impression absorbed at a conference is not. The moment an impression enters a research file with no source attached, the file has stopped being checkable, and checkability was the whole point of writing anything down.

Which three tests does an observation have to pass?

Three questions, asked in order, and an observation that fails any one of them is not used. The three are quick to ask, and they throw out most of what usually gets written about governance.

Is it observable? Is the thing actually in a published document, or has it been inferred from a reputation, a tone on a call, or a general sense in the market. Observability is the crudest test, and it removes the largest volume of material.

Is it sizable? There must be a denominatorThe figure that something is divided by when a size is put on it. Chosen badly, the resulting percentage is arithmetically correct and completely useless. for it. Four directors is not a size. Four directors out of nine is. Rs 18 crore is not a size. Rs 18 crore against a cost of materialsThe single line that gathers everything a maker spent on the physical inputs it bought and consumed during the year. of Rs 1,304 crore is. An observation that no denominator can be put under will silently expand to whatever size the reader's mood suggests.

Is it testable? The competing account has to be named, along with the published thing that would tell the two apart. Testability is the test almost nobody applies, and it is the one that fails most often. An observation that cannot separate two competing accounts has not established which world the company is in, however carefully it has been sized. Such an observation is not a reason to hide it. The right response is to record it, state that it separates nothing, and stop using it as though it did.

What does a board of nine actually establish?

Sarvani Coatings Limited has nine directors. Four of them are independent directorsBoard members who are neither employees of the company nor part of the group that controls it. What qualifies somebody as one is set by rule, not decided by a reader., two come from the promoter groupThe holder or holders recorded as controlling a listed issuer, counted together with the parties connected to them. and three are executive. A proportion nobody has worked is a proportion that gets quietly exaggerated, so the proportion has to be computed before a single word is said about any of it.

Four of nine is 44.4 per cent. Two of nine is 22.2 per cent. Three of nine is 33.3 per cent. Each of 44.444, 22.222 and 33.333 rounds downward at one decimal place, so the three printed figures add to 99.9 rather than 100. The underlying shares do add to exactly 100 and the nine seats add to exactly nine. The rounding is worth saying out loud rather than nudging one figure up to force the sum: a set of rounded shares that lands on 99.9 is telling the truth about rounding, and forcing it to 100 is a small dishonesty that trains a writer for a larger one.

Now the reading, and this is where most governance commentary goes wrong within one sentence. The 44.4 per cent supports a factual description of who sits on the board of Sarvani Coatings. The 44.4 per cent does not support any claim about how decisions there are taken, how a difficult question is handled, or whose view carries a room. The composition is disclosed. The conduct is not. No published document anywhere connects the first to the second for this company. The observation fails the third gate outright.

The minimum composition, and who counts as independent for the purpose of meeting it, is set by rule and not by a reader's judgement. Whether the composition described here satisfies anything is a question for those rules: sebi.gov.in carries the current text, and mca.gov.in carries the company law side of it.

How big is a related party purchase, and is size the whole question?

Sarvani Coatings buys Rs 18 crore of materials from an entity connected to its promoter group. Size it first, and size it against the line it actually belongs to. The cost of materials for the year is Rs 1,304 crore, so Rs 18 crore is 1.38 per cent of it. Size it a second way, against revenue of Rs 2,415 crore, where the same rupees are 0.75 per cent. Both are legitimate. The rupees were spent on the cost line, and that makes it the more honest denominator. The cost of materials is 54.0 per cent of revenue for the year, so it is also the larger of the two claims on the business, and a question landing anywhere inside it is worth sizing carefully.

Now the part that most treatments of related party dealings skip. Scale is the first question and it is not the only one. A small amount can matter enormously if it sits where the price is set, and a large amount can be entirely ordinary if it is a routine supply relationship at an ordinary price. A researcher who stops at 1.38 per cent has answered how much is at stake and has not touched whether anything is unusual.

Here is the everyday version. Ten shops in one mall all buy their packaging from a supplier connected to the mall's landlord. The arrangement is worth knowing about, and knowing it does not establish that anybody is being overcharged. Two entirely mundane figures would settle it: what the nine other tenants pay, and what the same packaging costs from an unconnected supplier down the road. Until somebody publishes that, the arrangement is a fact with an open question attached, and describing it as anything worse is somebody's opinion dressed as analysis.

In a research context, the mundane thing that would settle it has a name: the arm's lengthPriced the way two parties with no connection to each other would have priced it, rather than on terms a shared interest could have shaped. pricing basis. The disclosure gives an amount and a counterparty. The same document usually does not give the method by which the price was set. So the honest reading is a sized fact plus a named missing input, and the named missing input is the more valuable half.

How Governance and Disclosure Quality Affect Research Confidence

Almost everybody gets the direction of the next mechanism wrong, so it is worth stating plainly before it is demonstrated.

Disclosure quality determines the share of the reported figures whose basis can be independently verified, and everything that cannot be verified has to be carried as a range rather than as a single number. That sentence bears reading twice. The sentence says nothing whatsoever about whether the reported figures are right. The claim is about how much of them can be checked.

Work it on the cost line. Of Sarvani Coatings' Rs 1,304 crore of materials, Rs 1,286 crore is bought from unconnected suppliers, and the price paid there could in principle be compared against what that material costs in the market. The unconnected suppliers account for 98.62 per cent of the line. The remaining Rs 18 crore is bought from a connected party with no published pricing basis, so it cannot be compared against anything. The connected party accounts for 1.38 per cent. The two shares add to 100.00 per cent exactly, and that is the check that nothing has been lost.

So what may now be said? Exactly this and no more. The cost of materials is Rs 1,304 crore. The figure has not moved, and no evidence in the disclosure gives a reason to think it is wrong. And 1.38 per cent of it carries a question this disclosure cannot close. The statement is modest, specific and checkable, and it is completely different from the verdict a reader was probably expecting.

The everyday version helps here too. A buyer of a second-hand scooter is shown service bills for eleven of the twelve months and told that the twelfth was done by a friend. Nothing in that establishes that the scooter is worse. The gap establishes that one twelfth of its service history is unverifiable. A sensible buyer widens what they are prepared to be wrong by, asks for the missing bill, and does not silently knock a third off the price to punish the gap.

Why does weaker disclosure widen a reading instead of lowering it?

Here is the counter-intuitive part, and it repays sitting with for a minute. If a company discloses less, what has actually been learned about its numbers? Nothing. The learning is about the record, not about the business the record describes. So the reported figures stay exactly where they were, and what changes is how much confidence they can be held with.

Weak disclosure widens the range around a reading and leaves its centre precisely where it was, and a researcher who quietly shifts the centre downward has substituted a suspicion for evidence. The move feels prudent. It is not. Prudence that expresses itself as a number nobody can reconstruct is not prudence, it is an unrecorded opinion with a decimal point on it. Genuine evidence that a figure was overstated would justify moving the centre, with the evidence written down beside it. Having no evidence is the reason for width, not a licence for direction.

There is a second consequence, and it is the practical one. Widen a range far enough and some questions simply become unanswerable from the published record. Unanswerable questions are not a failure of the work. Saying which questions the record cannot answer is itself an output, and often the most useful one anybody can be handed. A reader who is told that a particular question cannot be closed from what has been published knows exactly what to go and ask for. A reader who is handed a point estimateA single number offered as the answer, with no width around it. Presenting one where the evidence supports a range hides how much of the answer was actually established. with the doubt quietly baked into it knows nothing, and cannot even tell that there was doubt.

What does the whole record look like, read in this order?

Now put it together on Sarvani Coatings Limited, from a standing start, in order, and end where the evidence ends rather than where an opinion would be comfortable.

Take the board first. Nine directors, four independent, two from the promoter group, three executive: 44.4, 22.2 and 33.3 per cent, printing to 99.9 for the reason set out earlier. Run the gates. Observable, yes, it is in the disclosure. Sizable, yes, four of nine. Testable, no. The competing account is that composition and conduct are simply unrelated at this company, and nothing published separates that account from any other. So the composition is recorded, and it is used for nothing further. The treatment is complete and honest, and it takes one line.

Then the connected party purchases. Rs 18 crore against a cost of materials of Rs 1,304 crore is 1.38 per cent, and against revenue of Rs 2,415 crore it is 0.75 per cent. Run the gates. Observable, yes. Sizable, yes, twice over. Testable, partly. The two competing accounts are an ordinary supply relationship priced as any unconnected supplier would have priced it, and a relationship priced on some other basis. Only the pricing basis separates them, and the pricing basis is not in this document. Partly testable earns a recorded observation plus one named open question, and no more.

Then the confidence step, worked as arithmetic so that it is demonstrated rather than asserted. Rs 1,286 crore of that cost line, or 98.62 per cent, is bought from unconnected suppliers whose pricing could be compared against the market. Rs 18 crore, or 1.38 per cent, cannot be. Therefore: the cost of materials is Rs 1,304 crore, that figure has not moved, there is no reason on this record to think it wrong, and 1.38 per cent of it carries a question the disclosure cannot close. The centre did not move and the range around it widened by a small, computable, stateable amount, and that sentence is the entire legitimate output of the governance work.

Every percentage above was worked from rupee absolutes held in whole rupees, never by dividing one rounded figure by another: Rs 18,00,00,000 over Rs 13,04,00,00,000 gives 1.3804 per cent, printed as 1.38; over Rs 24,15,00,00,000 it gives 0.7453 per cent, printed as 0.75; and Rs 12,86,00,00,000 over Rs 13,04,00,00,000 gives 98.6196 per cent, printed as 98.62, so that the two printed shares add to exactly 100.00.

What does a governance reading actually produce?

Four things, and none of them is a verdict on anybody.

A list of sized observations. An explicit statement of what each one supports, and for most of them that will be a factual description and nothing more. A statement of which research questions are now unanswerable from the published record, with the reason attached. And no characterisation of any board, any promoter group, any director or any counterparty. A disclosure states no reason, and this method supplies none.

A governance reading produces a note about the limits of the rest of the work, not a score, and any output shaped like a score has smuggled a judgement inside a number where nobody can inspect it. That is the real objection to a governance score, and it is not squeamishness. A score of six out of ten cannot be argued with. There is nothing there to argue with: no observation, no size, no competing account, no missing input named. Three plain lines and one open question can be argued with by anybody who reads them, and being arguable is exactly what makes them worth writing down.