How to Update Research After an Earnings Release

A results release is worked in order. The basis comes first, then the questions the view actually rests on, and the headline last rather than first. Most releases move several numbers and no views, and telling those two apart is the whole skill. Finishing an update having changed nothing is a finished update, not a wasted one.

Everything below sits on one distinction, set up in researching a company from a standing start and put to use here. Research on a company ends with a view and with the questions that view is standing on. When the next set of results arrives, the release does not know what those questions were. The release arrives as a list of figures arranged in the order the company chose, and working out which of those figures touch anything the analyst is actually uncertain about is the analyst's job, not the company's. Almost none of them will.

What order does an update actually run in?

Seven steps, and the order is the point. The analyst checks what basis the figures are on, goes to the written questions, sorts what moved into things that changed a number and things that changed a view, names which figures could have borne on a question at all, deals honestly with the gap between the release and the forecast, writes down what the release taught, and then stops.

The headline figure is not step one and it is not step two, and a reader who starts there has handed the ordering of their own research to whoever wrote the press note. That is not a complaint about companies. A results release is written for lenders, employees, journalists, index compilers and shareholders at once, and it leads with what is legible to all of them. The analyst's questions belong to the analyst alone, so nothing in the release is arranged around them.

Consider a school report for a child a parent knows well. The one line every reader can use is the overall grade, so the report leads with it. But a parent is not every reader. The two things that parent was actually worried about are already known, whether the handwriting improved and whether the child is speaking up in class, and a good parent goes to those two lines before the grade. The grade is real. The grade is just not what the parent came for.

Step one: what is checked before reading a single number?

Four things, and none of them is a number the analyst came for. Is the statement standaloneFigures for the parent company by itself, leaving out the entities it holds a controlling stake in. How the two presentations are built is settled in the accounting material. or consolidatedFigures for the parent company and the entities it controls added together as one economic unit, with dealings between them removed.. Then the period the statement covers, and the period it is being compared against. Has any corporate actionA company level event that changes the shape of the equity or the entity, such as a bonus issue, a split, a merger or a sale of a business. Settled in the listings material. occurred since the last set examined. And have the prior period figures been restatedPrior period figures reissued on a new basis so they can be read alongside the current ones, usually after an accounting change or a reorganisation. Settled in the listings material..

The basis comes first because a comparison drawn across a restated period is not a comparison at all, and no amount of careful reading further down can repair it. If last year has been recast and this year has not, or the entity has absorbed a business partway through, then every growth rate calculated is measuring two different things against each other and calling the difference performance. Why restatement happens and what it does to a per share figure is covered separately. A restatement has to be looked for before the reading starts.

There is a household version of this that most people already do without naming it. A household is checking whether the electricity bill went up. Before the two amounts are compared, the number of days each bill covers gets checked. A sixty two day bill against a thirty day bill is not a price increase, it is a calendar. Nobody finds that subtle. The same check becomes subtle only when the two periods are both called quarters and the difference is buried in a note.

Step two: what comes first, the headline or the written questions?

The written questions, and this only works if they were written down. Researching this company from a standing start ended with exactly two for Sarvani Coatings Limited. First, can gross margin sit at 46.0 per cent from here, or has the company just touched its high point on the way back down. Second, was the 0.13 point move in share of the field the start of something or a single good year.

A reader who wrote two questions down at the end of the previous note opens the release holding a filter, and a reader who did not opens it holding nothing and reads in whatever order the company printed. That is the entire difference, and it is available to anybody for the cost of two sentences typed at the end of a note.

Neither question is "will the results be good", and neither is "will profit beat the market". Both are narrow, both name a specific figure, and each can in principle be settled by one named disclosure. Narrowness is what makes step four possible at all. Which figures bear on a question cannot be asked until the question is sharp enough to have an answer.

Step three: how is a changed number told apart from a changed view?

A number changes every single period. Moving is what a number does. A view changes only when something the view was resting on has moved. A changed number and a changed view are different events, they happen at wildly different frequencies, and confusing them is the most common way a research process turns into noise.

Most releases change many numbers and no views, and treating every release as informative is exactly how a research process fills up with movement that means nothing. Revising something after every set of results is not responsiveness. Revising on that schedule is being led, one quarter at a time, by whatever moved most.

The test is mechanical once the two questions are in hand. A line is taken from the release and set against what the view was standing on. If the line is not one of those supports, it changed a number. If it is one of those supports, a second question follows: did it move, or did it merely appear again at the same level? A support that appears again unchanged is comforting, and comfort is not a change of view.

Step four: which figures could bear on a question at all?

Now go the other way. Instead of sorting what arrived, name in advance what would have to arrive for each question to move. The margin question asks whether 46.0 per cent is somewhere gross margin can sit or the top of a climb, and the case record is explicit that the gain came from realisationRevenue per unit sold, meaning what the average unit actually fetched. Realisation moves with both price and the mix of what was sold, so the two have to be separated. outrunning input cost per unit rather than from input cost falling. So the evidence that would move it is a volume and realisation split, or a segment disclosureResults broken out by line of business rather than reported as one total, so a reader can see which part of the company produced the movement. separating decorative from industrial. The share question needs a figure for the whole field over the same period. A single company's release cannot contain one.

The figures that would update a view are usually not the figures a release leads with, and that gap is a fact about who a release is written for rather than a criticism of the company that wrote it. A release leads with revenue, profit and earnings per share because those are the lines every reader shares. The two questions here are shared by nobody.

Step five: the forecast missed. What was actually wrong?

Say Meghna Iyer had written down a profit after tax forecast of Rs 355 crore, and the release printed Rs 315 crore. She is Rs 40 crore adrift, a miss of 11.3 per cent, and a miss that size is bad by any standard. The temptation is to treat that as a verdict on her thinking.

A forecast number is almost never right, so the useful question after a miss is not whether the number was right but whether the mechanism described still holds. Her mechanism was that the cost of materials would sit at 54.0 per cent of revenue. The cost of materials sat at exactly 54.0 per cent of revenue. The cost behaviour she claimed to understand ran precisely as she said it would, and she still missed the profit line by 11.3 per cent. A profit forecast also carries a revenue estimate, an expense estimate, a depreciation estimate and a tax rate, and any of those can be adrift while the mechanism is intact.

A broken mechanism and a wrong number are two different errors, and they carry two different repairs. A broken mechanism means the story told about the business was wrong, and the work has to be redone. A wrong number on an intact mechanism means the arithmetic or the inputs were off, worth noting and not a reason to change the account of what is happening. Confusing the two is what makes sound work look broken and, just as often, makes a lucky number look like understanding.

Step six: what actually gets written down?

Four fields, and they take five minutes. What moved. Whether it touched a question. The new answer to that question, if there is one. Whatever is still open. Four fields are the whole record, and the record is worth more than any single conclusion reached this year.

The outcome cannot show which parts of the work were sound. A record of the reasoning is the only thing that can. Judging research by how a position turned out is covered separately: the outcome is a fact about the world, not a fact about the quality of the thinking that produced it, and grading work by its result rewards luck and punishes care with equal enthusiasm. The record is the alternative. The record puts the reasoning on paper before the answer is known, and that is the only moment at which it can be written honestly.

There is nothing sophisticated about the form. A shopkeeper who writes "took extra stock in before the festival because last year I ran out on day two" has done exactly this, and next year they can tell whether the reasoning was sound separately from whether the festival was busy. Without the note, all they have is a takings figure and a story assembled after the fact.

Step seven: when does an update stop?

An update stops when three things are true. The basis has been checked. Each open question has either been addressed or explicitly left open with a note of what would settle it. The record is written. Three conditions are the whole stopping rule, and none of them is "every figure in the release has been read".

An update that changes nothing is a finished update, not a wasted one, and a process that cannot record a null result will keep manufacturing changes to justify the time it spent. The stopping rule is the hardest of the seven to act on. An hour spent producing no revision feels like an hour wasted, and the feeling is wrong. The field handed over no answer this period, exactly what would have done so is now known, and that stands in writing.

Compare it to a health check that comes back clear. Nobody says the appointment was pointless because nothing was found. The information is precisely that nothing was found, and it is worth having because it was looked for properly.

What must never be a step?

Revising a view because the share price moved after the release. Not softening the view, not hardening it, not quietly adjusting a forecast until the arithmetic agrees with where the price now sits.

A price move after a release is a great many other people updating their own assumptions at once. The move may be informative about what they had assumed, and it is not evidence about the business. If the price falls sharply on a release the analyst read as unremarkable, the useful reading is that others were carrying an expectation the analyst did not know about, and that is a question about positioning rather than about the company. Positioning may be worth pursuing. The price has disclosed nothing about the business, so a price move is no reason to revise the account of the business.

What does the whole procedure look like on one release?

Sarvani Coatings Limited has published a fresh set of results. How the release below was built matters for what it can and cannot show. Revenue is grown at the company's own published rate of 13.9 per cent, every cost line is held at exactly the share of revenue it took in year three, finance cost and other income are held at their published amounts because neither moves with revenue, and tax is struck at the published effective rate. Every ratio on this release is therefore a ratio the company has already reported. Only the period is new.

Step one, the basis. The figures are standalone, they cover a full year against the full year before it, no corporate action has occurred since the last set, and no prior period figure has been restated. Both sides are on the same basis, so a comparison is permitted. ComparabilityThe condition that two periods are stated on the same basis, so that a difference between them reflects a change in the business rather than a change in how it was measured. is not a formality. Being on the same basis is the licence to do everything that follows.

Step two, the questions. Can gross margin sit at 46.0 per cent from here. Was the 0.13 point share move the start of something. Straight past the profit line, straight to those two.

Step three, the sort. Fourteen lines moved and not one of them moved a view. The two lines that come closest are these. The cost of materials sits at 54.0 per cent of revenue, exactly where it sat, and gross margin sits at 46.0 per cent, exactly where it sat. Both of those bear directly on the margin question. Neither of them moves it. Sitting still for one more year fits a margin that can be held, and it fits a high point that has simply not turned yet. The share question stands apart. A single company's release names only its own figures, so no line of this one reaches the field at all. Revenue growing 13.9 per cent is only a share gain against a known figure for the whole field, and the release does not say.

Step four, what was absent. Three disclosures would have moved something and none of them arrived: a volume and realisation split, a segment disclosure separating decorative from industrial, and a revenue figure for the whole field over the same period. Naming those three is the most valuable output of this update. The list says exactly what to look for next time, and it prevents a second reading of the same fourteen lines with more attention.

Step five, the forecast. Meghna Iyer had Rs 355 crore against a printed Rs 315 crore, a miss of 11.3 per cent, while the cost line she had assumed at 54.0 per cent came in at 54.0 per cent. Her mechanism held and her arithmetic did not. A year from now the mechanism and the arithmetic will need to be read separately, so the distinction goes in the record as two separate lines.

Step six and seven, the record and the stop. Fourteen lines moved, three of them touched a question, none of the three moved from where it already sat, no new answer, both questions still open with a named list of what would settle them. The basis was checked, both questions were addressed, the record is written. Stop. The honest output of this update is that no view changed, and that is a successful update rather than a wasted one.

Where do the filing and the obligations actually live?

Filing duties and their deadlines are a conduct and disclosure matter, they get amended, and a written summary is the worst possible place to learn them from. Take the obligation from the regulator that writes it and the filing from the venue that received it, and never carry either into an update out of memory.