How to Prepare for an Earnings Call Using Public Disclosures

Preparing for an earnings call is a procedure that turns published disclosures into questions someone can actually answer. Six steps: establish the basis, read the statements and the notes before any commentary, list every divergence between two published figures, convert each into a question with a checkable answer, rank by what would most change the analyst's understanding, and write down what would be accepted as an answer.

Two things decide what the whole exercise is for, so both are worth fixing before the first step. The output of this procedure is a short written list: a question, and beside it the sentence that would be accepted as an answer. The written list is the entire product. The list is not a view, not a case, not a suspicion, and not a theory that the questions have been arranged to support. Nothing in this sequence produces a conclusion about any business, and the moment a conclusion appears in the preparation, the preparation has stopped being preparation and become a search for confirmation. The second thing is smaller and just as practical: the whole procedure runs on documents anybody can obtain. No relationship, no access, no private call. The filings alone are enough to fill a sheet with questions that a person who knows the business can answer in a sentence each.

Four things are taken as settled. The nature of an earnings call, who is on it and how the time is divided is covered separately. So is the ordering of sources by reliability. So are the two habits of mind this procedure is built out of: the tendency to notice evidence that fits a story already held, and the discipline of writing down in advance what would satisfy the questioner. The tests for whether reported profit is a durable base for a forecast are covered separately as well. The divergences used in the worked run below were each established elsewhere in this subject, and the arithmetic behind them belongs there. The order is the subject, and only the order.

Why does the order of these six steps matter more than the steps themselves?

Any reasonably careful person, given a set of accounts and an afternoon, will end up doing most of these six things. Such a person will look at the numbers, notice some things, write some questions. The difference between that and a procedure is not effort, and it is not intelligence. The difference is that a procedure fixes the sequence, and in this particular sequence three of the six positions are load bearing in a way that is invisible while the work is under way.

Each position in this order is earned: the basis comes first because a question asked across two different bases will be answered by a definitional difference rather than by a fact, the reading order comes second because whatever is read first decides what gets noticed afterwards, and the acceptance criterion comes last because writing it after the call is not writing it at all. The three middle steps could in principle be shuffled without much damage. The outer three cannot. Three moves are never steps at any point in the sequence, and the red panel below sets them out under the six numbered rows. The three excluded moves are not advanced technique to be graduated into. Each sits outside the procedure entirely, and a preparation that drifts into any of them has quietly turned into something else.

Step one, what exactly is the preparation about?

Two relatives argue for twenty minutes about whether this year's wedding cost more than the one three years ago. One of them has been counting the catering. The other has been counting the catering, the hall and the buses. Neither is lying, neither is bad at arithmetic, and the argument cannot be settled because it is not one argument. The argument is two people using the same word for two different things. Every minute spent on it is wasted, and both of them walk away thinking the other is being difficult.

Step one writes down four things and calculates nothing: which entity is being read, which two periods are being compared, whether the figures are the entity by itself or the whole group, and what changed in the accounting or the structure between the two periods. The last of the four is the one that gets skipped, and it is the one that ruins the most questions. If a business bought something, disposed of something, changed the length of its reporting period or restated a comparative figure, then some part of every movement about to be noticed belongs to that event rather than to trading, and a question that ignores it will be answered by a definitional difference in about eleven seconds. The answerer is not being evasive. The question was badly built.

Run the four questions on Anjani Stationers Private Limited, an invented business making school notebooks and exercise books. Anjani holds 70 per cent of Chitra Binding Works, also invented. The four answers take a minute. The entity is Anjani Stationers. The periods are year one and year two. The basisThe footing on which a set of figures has been prepared: which entity, which span of time, and whether the numbers cover that entity alone or that entity together with everything it reports as one with itself. for the working capital and margin figures used below is the entity on its own, the footing on which they were published. And the fourth answer is the one that matters most here: Chitra Binding Works was bought at the start of year two, so year two carries costs that year one could not have carried, and Chitra became a connected supplier in the same month. Write that down. Anybody who asks about the cost increase without knowing it will get an answer about the acquisition and will have burned a question learning something already printed.

Step two, what is read first, and what waits until afterwards?

A parcel arrives four days late. Before the tracking record is opened, the shop calls to say the delay was a transport strike. Read after that call, every scan and every gap arranges itself around a strike. There is already somewhere to put each one. Read cold, the same record shows that the parcel sat in the same warehouse for three of the four days and never entered transport at all. The record did not change. The order in which the record and the explanation arrived did.

Step two reads the statements and the notes first, completely, and opens the management commentaryThe written sections a business puts around its accounts in which its own people describe the year, explain what they think happened and set out how they see the period ahead. Useful, and written by people with a view. only after the question list has been drafted, because commentary frames what gets noticed and quietly decides which questions occur at all. This is exactly the tendency to favour evidence that fits a story already held, arriving in practical form, and the practical form is more useful than the abstract one. Nobody can decide not to be framed. The one thing that can be controlled is which document arrives first, and that is a scheduling decision rather than an act of willpower.

The ordering is not saying that commentary is unreliable, and it is certainly not saying that the people who wrote it are trying to steer the reader. The people who wrote it are describing their own year, from inside it, with more knowledge of it than any outside reader will ever have. The commentary is often the single most informative document in the pack. The point is narrower and entirely about sequence: read second, it answers questions the reader brought; read first, it supplies the questions as well as the answers. Confirmation biasThe habit of noticing and weighting whatever fits a view already held, while what does not fit slips past unregistered. Bias does not feel like bias from inside. Procedure handles what effort cannot. is not defeated by knowing about it. Confirmation bias is handled by putting the documents in an order that gives it less to work with.

Step three, which pairs of published figures moved apart?

A vegetable seller supplying two hotels mentions that takings are up about a tenth on last year. Later the same evening he adds, without connecting the two, that the money the hotels still owe him has gone up by half. Neither number is alarming on its own. Set beside each other, they give something specific to ask about, and both figures came from him.

Step three writes out every divergenceTwo figures that would normally move together but did not, set out as a pair. The pair is the finding, not either figure by itself. as a pair of published figures, and the pair is what makes it worth asking about, because a gap between two numbers the business itself put in print cannot be waved away as somebody's outside assumption. That last part is the whole reason this step exists in this form. A question built on an outside model invites a discussion about that model. A question built on two of the business's own numbers cannot go anywhere except to the answer. Four kinds of pair recur often enough to look for by name: an amount owed by customers growing faster than sales, stock growing faster than the cost of what was sold, a margin that is flat above one line and falling below it, and a provision moving by a different order of magnitude than the balance it sits against.

Anjani Stationers, read cold, produces exactly those four. Receivables grew 21.8 per cent against revenue growth of 12.5 per cent, a gap of 9.3 percentage points. Inventory grew 47.4 per cent while cost of materials consumed grew 12.5 per cent. Gross margin was 45.0 per cent in both years while the margin at the operating line fell from 22.1 per cent to 15.4 per cent, a drop of 6.7 points that therefore sits entirely below the gross line. The provision for doubtful debts tripled, from Rs 3,00,000 to Rs 9,00,000. The receivable book it sits against grew by 21.8 per cent. Write all four down as pairs, in that form, and resist the pull to explain any of them yet. Explaining is step four's problem and it is not even the same activity.

Step four, what turns a divergence into a question someone can actually answer?

A car goes to a mechanic and the owner asks why it feels bad. The answer is five minutes of general remarks about the age of the vehicle, most of it true, none of it usable. Asked instead how many kilometres it has run since the belt was last changed, the mechanic gives a number, immediately, from a man who knows. The second question is not cleverer. It is smaller, more specific, and answerable without a speech.

An answerable questionA question built so that a person who knows the business can answer it completely in one sentence, because it points at named figures and asks for a fact rather than a view. passes three tests: it names the figures it is about, it asks for a fact rather than an opinion, and someone who knows the business could answer it fully in a single sentence. All three, independently. A draft can name the figures beautifully and still ask for an opinion. A draft can ask for a hard fact and still be so broad that answering it honestly takes four minutes. Each failure has a different repair. The tests are applied one at a time rather than collapsed into a general sense of whether a question is any good.

Watch the conversion happen on the receivables pair. The draft that arrives first is almost always some version of why the business is struggling. The draft names no figure, so nobody knows which part of the year is being asked about. It asks for a judgement rather than a fact, so the honest answer is a view. Answering it properly means describing the whole year, and that is a speech. Now the rewrite: of the Rs 17,00,000 increase in gross receivables, what share sits with the largest customer group. The rewrite names both the amount and the group, and asks for a fact. The answer is a number and a fraction, deliverable in one sentence by anybody at the business who has seen the ledger. Same underlying curiosity, entirely different instrument.

Step five, which questions have actually earned a place on the list?

Eight minutes with a doctor, and six things written down beforehand. Two of them will change what happens next. Four of them are things already half known, asked mostly to hear somebody say them out loud. Everyone does this, and the fix is not discipline in the room. The fix is deciding, before walking in, which two go first.

Step five ranks the drafted questions by how much the answer would change the analyst's understanding, and most call time anywhere is spent on questions whose answers change nothing at all. Two categories eat that time. The first is the question whose answer is already printed in the pack, asked because it feels like a safe opening. The second is the question designed to sound searching to the other people listening, a performance rather than an enquiry and usually obvious as one. Both are perfectly answerable. Both pass all three tests from step four. Neither earns its place, and this is the reason answerability and value are ranked separately rather than rolled into one judgement about whether a question is good.

Step six, what would be accepted as an answer?

A buyer sets out for a fan having decided beforehand to buy one only if it is under a fixed price and has a spare part available in town. The decision made at home is what protects the buyer from a persuasive twenty minutes in the shop. Decided in the shop instead, the conclusion comes out differently: a slightly more expensive fan with no spare part available becomes the sensible choice, and the buyer believes it.

Step six writes an acceptance criterionA sentence written before the question is asked, setting out what an answer would have to contain for the question to be treated as closed. Written afterwards, it simply describes whatever was said. beside every question on the list, stating what an answer must contain for the question to count as closed, and the criterion has to exist before the answer does or it is not a criterion at all. The reason is uncomfortable and worth saying plainly: a fluent, confident answer delivered by somebody senior is extremely convincing in the moment, and the feeling of being satisfied is not evidence of having been answered. Writing the criterion in advance separates the two. Afterwards the test is not whether the answer felt satisfying. The test is a sentence checked against a sentence, a much smaller and much more honest task.

Criteria are specific and short. For the receivables question, an answer counts if it gives a share of the Rs 17,00,000 and identifies the group, and does not count if it describes the receivables position in general terms. For the cost question, an answer counts if it names the largest two or three items making up the Rs 12,00,000, and does not count if it says costs rose because of investment in the business. Both of the non qualifying answers are perfectly honest things for someone to say. Neither closes the question, and the criterion is what makes that noticeable at the time rather than three days later.

When is the preparation finished?

The stopping rule contains no judgement, and a rule that contained judgement would not stop anything. The work is finished when every divergence on the list has become a question, every question passes all three answerability tests, and every question has a written criterion beside it, and it is finished at that moment even though no conclusion about anything has been reached. There is no further pass over the same figures that improves this. Reading the same statements a fourth time does not produce a seventh divergence, only the same six with more confidence attached to them, and that is worse than the starting position. An answer adds something, and an answer comes from a person, not from another hour with the pack.

Anjani Stationers finishes in four entries, and finishing with four open questions and no view is the correct output rather than an incomplete one. If the temptation to add a summary line is strong, notice what that line would be made of. The line would be made of the same four questions, none of which has been answered, arranged into an impression. Impressions are easier to repeat, so the impression would travel further than the questions did, carrying the credibility of six steps of careful work that never came near it.

What must never appear anywhere in this sequence?

Three things are excluded at every position, and none of them is excluded because it is impolite. Each is excluded because it destroys the thing the procedure was built to protect.

The first is settling the conclusion and then drafting questions that support it. Settling first is the reverse of the six steps performed in the same clothes, and it is nearly undetectable from outside because the questions look exactly like real ones. The tell is internal: the drafter already knows which answer would please. The second is asking anything that presumes wrongdoing. Beyond the obvious problem that the presumption is not supported, a question built that way is unanswerable in a specific and interesting sense. Both a business that did the thing and a business that did not will deny it, so the answer carries no information whatever. Every accusing draft can be rewritten as a request for a fact, and the rewrite is strictly better in every respect.

The third is treating a refusal to answer as an admission, and it is the most important of the three: a business may decline for reasons of competitive sensitivity, contractual confidentiality, an ongoing negotiation or a legal constraint, and a refusal establishes precisely nothing. A question about the share of a receivables increase belonging to the largest customer group asks a business to describe, in public, the payment terms it has agreed with its biggest customer, where every competitor bidding for that customer is listening. Declining that is not evasion. Declining is ordinary commercial care, and a reader who scores it as a point against the business has learned nothing and has damaged their own judgement in the process. The question is recorded as remaining open, and the list moves on.

The cost of getting this wrong lands somewhere real. A wrong reading of a refusal, repeated to two other people, becomes a rumour about an honest business that no arithmetic ever supported, and the person who started it will not be there to correct it. The mistake also costs the reader, and more than they expect: someone who reads refusals as guilt accumulates a set of conclusions that cannot be defended when challenged, and the first time one of them is challenged properly, everything else they have said gets re-examined too.

What does the whole sequence produce on one set of figures?

Run all six on Anjani Stationers and the output fits on one sheet. Step one fixes the basis: Anjani Stationers on its own, year one against year two, with Chitra Binding Works bought at the start of year two and becoming a connected supplier in the same month. Step two reads the statements and the notes with no commentary open. Step three finds the four pairs. Step four converts each into a question that names its figures and asks for a fact. Step five ranks them. Step six writes what would close each one. The table below is the whole product.

Notice what is not in that table: any statement about what Anjani Stationers did, any ranking of the four by how worrying they are, and any conclusion of any kind. The business is an ordinary one that has done nothing wrong, and every one of the four divergences has a plain explanation already sitting in its own disclosures. A plain explanation is not a twist but the normal case, and that is why the output of six steps of careful work is a list of questions rather than a view.

How does somebody with money at stake actually use a list like this?

A credit officer at a bank looking at a working capital limit for a business like this one uses the list in a way that surprises people who have only seen it used by analysts. She does not need answers to all four. She needs the answer to the first one. If most of the receivables increase sits with one large, slow paying customer group, then the limit she is being asked to extend is really a loan against one relationship, and that is a different kind of credit from the one described in the application. The other three questions matter to her mainly because they might change that first answer.

An equity analyst uses the same sheet differently. His interest is which parts of the movement repeat. A cost that recurs changes every forecast year. A cost that arrived once changes only this one. The accounts cannot tell him which it is, so the Rs 25,00,000 of extra cost below the gross line is the centre of his sheet. A supplier deciding whether to extend terms uses a third slice again, and typically only two of the four questions, both about how the business pays rather than how it is paid. The same six steps and the same four questions serve three different decisions, and none of the three needs a conclusion about the business, only an answer to the questions each of them cares about. That is the practical argument for producing a list rather than a view: a view serves whoever formed it, while a list of answerable questions serves anybody who picks it up.

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