How to Read an Earnings Release Without Being Misled

The work runs as a fixed order of checks on the document itself. Both periods go onto the same basis, the quarter is placed in its year before being compared with anything, volume and realisation are rebuilt behind the revenue line, every adjusted figure is reconciled to the reported one in both directions, any mix explanation is tested against what mix can arithmetically deliver, and a single quarter is never annualised. Anything the release did not disclose is then written down.

Each check below names one thing to open and one thing to record. All three of those have their own places and are taken as read here, so no check stops to explain what a segment is, why an adjusted figure exists at all, or how a forecast gets built out of any of it. The output is a worksheet somebody else could rebuild from the same filing, and its checkability is the whole of its value.

A check run out of turn produces an answer that looks finished and rests on nothing, so the order is doing as much work as any single check. Two periods compared before anyone has confirmed they were measured the same way leave every percentage computed afterwards carrying the mismatch without showing it. Margin commentary read before the revenue line has been rebuilt turns into a test of the explanation against a memory of the numbers rather than against the numbers. The sequence exists to stop both.

Check one: which document is actually being read?

The one the company filed, pulled from the exchanges, together with whatever presentation went up beside it. Not a wire story, not a screenshot, not a table somebody rekeyed. The filing is opened, and before a single number is read, what this kind of document carries and what it leaves out is written down. The second list is what the analyst's own work has to supply later.

A quarterly release is short. The document carries revenue for the quarter, a small set of cost lines grouped the way the company groups them, the profit lines down to profit after tax, a revenue split by segment, and the comparative quarters. Those lines are close to all of it. Sarvani Coatings Limited, an invented paint maker, reports revenue of Rs 700 crore, earnings before interest, tax, depreciation and amortisation (EBITDA) of Rs 145 crore and profit after tax of Rs 92 crore for its third quarter of year three.

Now the absences, and they are the same absences every quarter. No release states how many litres were sold, what the average price per litre was, or what a litre of raw material cost. The three inputs any revenue build needs are exactly the three the document never supplies. Consider a shop that shows the day's takings but never the till roll. The money that came in is known. Whether more people came, or the same people paid more, is not, and no amount of staring at the takings will separate them.

Check two: were both periods measured the same way?

Three things have to match before any comparison is available, and all three are confirmed before anything is computed. The share count, after any corporate actionA company action that changes the shares themselves rather than the business, such as a bonus issue or a split. The listings material settles what each kind does to a per share figure. in between. The segment reportingHow a company carves its business into parts when it reports, so revenue and profit show up for each part as well as for the whole. The company chooses where each carve line falls, and can redraw it. boundaries, which a company can redraw. And the accounting basis, meaning which costs sit under which heading and under which policies.

A restated line wearing its old name is a different measurement, not a different result, so if any one of the three moved, the comparison is abandoned and the movement reported. This is the check people skip, and skipping it is expensive in a particular way: it never announces itself. The arithmetic still runs. The percentages still print. Nothing anywhere on the screen goes red.

The shape of this is familiar from a kitchen. Somebody weighs the rice in a steel bowl one week and in a plastic one the next, then reports that the household is eating more. The number went up. The measurement changed. Neither week's figure is wrong, and the difference between them means nothing at all until somebody weighs the bowls.

Check three: where does this quarter sit inside its year?

Before a quarter is compared with anything, all four quarters of the year go on one line and are looked at together. Only then does the comparison get chosen, and the choice is written at the top of the working note. Sarvani Coatings' year three ran Rs 590 crore, Rs 545 crore, Rs 700 crore and Rs 580 crore of revenue, and Rs 106 crore, Rs 88 crore, Rs 145 crore and Rs 107 crore of EBITDA. The four quarters sum exactly to the published Rs 2,415 crore and Rs 446 crore.

The eight figures give quarterly EBITDA margins of 17.97, 16.15, 20.71 and 18.45 per cent against a blended marginThe margin computed on the whole period's totals rather than averaged across the sub periods. Because the sub periods differ in size, it is not the average of them. for year three of 18.47 per cent. Look at where the peaks fall. The third quarter is the festive one and carries the highest revenue of the four, at Rs 700 crore, or 28.99 per cent of the year against the 25.0 per cent an even year would give. The festive quarter also carries the highest margin. The second quarter carries the monsoon and is the weakest on both.

Revenue and margin peak in the same quarter, so the strongest quarter is strong twice over. Any comparison that crosses seasons is comparing two different things wearing the same label. A tea stall outside an examination hall takes more in March than in June, and its best days also carry its best margin, because the queue is long enough to sell the expensive thing. Nobody at that stall thinks June is a business in decline.

What happens when the same quarter is read against two different bases?

Sarvani Coatings' fourth quarter of year three, set first against the quarter immediately before it, gives the following. Revenue fell 17.14 per cent, from Rs 700 crore to Rs 580 crore. EBITDA fell 26.21 per cent, from Rs 145 crore to Rs 107 crore. The EBITDA margin gave up 2.27 points, from 20.71 to 18.45 per cent. Written out like that, the quarter reads as a collapse.

Now compare the same fourth quarter with the year it sits in. Its margin of 18.45 per cent sits 1.96 basis pointsA hundredth of a percentage point. The step from 18.45 to 18.47 per cent is two of them. Margin gaps are often too small for a decimal place to carry comfortably, and that is why the unit exists. below the full year's 18.47 per cent, which rounds to two. Its Rs 580 crore of revenue is 24.02 per cent of the year against the 25.0 per cent an even year would give. Written out like that, the quarter reads as entirely ordinary.

Both readings come from the same four numbers, and the comparison chosen produced the story rather than the numbers producing it. This is why the comparison chosen goes at the top of the working note in writing. A reader who cannot see which base was used cannot tell the finding from the framing, and neither, six months later, can the analyst who wrote it.

Check four: what is behind the revenue line?

Split it into volume and realisationRevenue divided by units sold, meaning the average price actually achieved after discounts and mix. Building revenue from volume and realisation is covered separately. before a word of the margin commentary is read. The release supplies neither, so they are assembled from whatever the company discloses elsewhere, from the segment split, and from standing assumptions, and every one of the three sources is labelled beside the figure it produced.

Do the assembly first and the commentary second, in that order, without exception. The commentary was written after the numbers were known, by people who already knew which explanation they preferred, so it is a claim to be tested rather than evidence to be used. Read before an independent split exists, it stops being a claim and quietly becomes the frame inside which support is then sought.

For Sarvani Coatings across year two into year three, the record carries volume up 6.0 per cent on revenue up 13.92 per cent, so realisation rose about 7.5 per cent. The single division of growth into volume and realisation is the whole of check four, and it is already more than the release disclosed.

Check five: does every adjusted figure reconcile back, in both directions?

Find every figure the company describes as adjusted, underlying, normalised or like for like, and walk each one back to the reported figure it came from. Then walk it forward again, this time adding back everything that hurt the period and removing everything that helped it. Two directions, always, on the same figure. The second pass is the reconciliationThe itemised bridge showing how an adjusted figure was reached from the reported one. Which items are legitimate to adjust for is settled in the earnings quality material. most releases do not print.

Sarvani Coatings' year three EBITDA is reported at Rs 446 crore. Adding back a Rs 6 crore charge gives Rs 452 crore, a margin of 18.72 per cent on the year's Rs 2,415 crore of revenue. Also removing a Rs 4 crore write back that went the other way gives Rs 448 crore, a margin of 18.55 per cent. The one sided version lifts EBITDA by Rs 6 crore and the honest version lifts it by Rs 2 crore, so two thirds of the improvement was a direction nobody ran.

A household budget makes the two directions easier to feel. The month may fairly be called unusual because the scooter needed repairing. The same month, though, is the one in which the landlord returned a deposit. Counting only the repair describes a month that did not happen, and describes it in the direction that flatters.

Check six: can a mix shift arithmetically deliver what is being claimed?

When a company attributes a margin move to a mix shiftA change in the proportions of the segments making up revenue, so the blended figure moves even when nothing inside either segment changed., do not argue with the story. Multiply. The most a mix shift can contribute to a blended margin is the size of the shift multiplied by the gap between the two segments' margins, and that product is a ceiling rather than an estimate.

Run it on Sarvani Coatings. Industrial revenue went from Rs 510 crore of Rs 2,120 crore in year two to Rs 604 crore of Rs 2,415 crore in year three, so its share moved from 24.06 to 25.01 per cent, a shift of 0.95 points. Over the same one year the gross margin moved 2.00 points, from 44.0 to 46.0 per cent. For a 0.95 point shift to deliver 2.00 points, the two segments' gross margins would have to sit 209.7 points apart. No pair of margins can. At a generous 15 point gap the shift delivers 0.14 points, or about 7 per cent of the move.

Even an arithmetically impossible 100 point gap between the segments would deliver only 0.95 points, under half the move. Mix cannot be the main explanation, and where the margin actually came from still stands open. Real evidence usually does exactly that. Evidence narrows a question honestly without closing it, and any treatment that pretended otherwise would teach a habit that fails the first time it meets a real company.

Check seven: what happens if a quarter is scaled up to a year?

A quarter is never scaled up to a year. If a figure must be scaled at all, the scaling factor and the reason for it are written down, and the scaling is applied to a full year rather than a quarter. The instruction is that blunt because the error it prevents is large, is signed, and repeats in the same direction every single year.

Every setting overstates or understates. The season lifts the margin as well as the revenue, and the two errors compound rather than adding, so the EBITDA bar always misses by more than the revenue bar. The festive quarter is the worst of the four: Rs 2,800 crore of revenue against an actual Rs 2,415 crore, an overstatement of 15.94 per cent, and Rs 580 crore of EBITDA against an actual Rs 446 crore, an overstatement of 30.04 per cent. The monsoon quarter fails as hard in the other direction, at minus 9.73 and minus 21.08 per cent.

The fourth quarter is the interesting exception and worth sitting with. Its errors are minus 3.93 per cent on revenue and minus 4.04 per cent on EBITDA, so the EBITDA error is still the larger of the two but by only 0.10 of a point. The near match is not luck. The fourth quarter's margin of 18.45 per cent sits almost exactly on the year's 18.47 per cent, so there is very little margin error left to compound with the revenue error. The gap between the two errors is a direct reading of how far that quarter's margin sits from the year's. The compounding is severe in the festive quarter and nearly absent in the fourth for that reason alone.

Read the fourth column on its own. One published year, four honest quarters, and annualised EBITDA readings running from Rs 352 crore to Rs 580 crore. The span is Rs 228 crore, wider than half the actual figure, and which end of it somebody quotes depends on nothing but which quarter they happened to pick up.

Check eight: what did the release not disclose?

Write the list. Take the profit ladderThe stack of profit measures a statement reports, each one reached by taking a further group of costs off the one above it. How that stack is assembled is covered under accounting. and the handful of items that sit outside it but still carry the period, and note every one the release did not address. Volume. Realisation. Input cost per unit. Price separated from mix. Margin by segment rather than revenue by segment. Anything about the quarters ahead.

The list barely changes from quarter to quarter. Its sameness is not a reason to skip it. The list is nearly identical every quarter precisely because it marks the places where the analyst's own assumptions are doing the work, and an assumption that stops being written down is an assumption that stops being visible. Six months later nobody will remember which figures the company gave and which the analyst supplied, and neither will anybody reading the file.

Think of a rented flat with no electricity meter reading on the bill, only a flat charge. Every month it is the same absence, and every month the same estimate quietly gets treated as a measurement. Nobody decided to believe it. The estimate just stopped being labelled.

What can two years of quarters settle, and what can they not?

The record here carries both year two and year three by quarter, so the year on year comparison is available and is the one to run. The year on year comparison puts each quarter against the same season a year earlier, and no other comparison removes the season instead of arguing with it. Year two ran margins of 15.38, 13.81, 18.03 and 16.44 per cent against a blended 16.04 per cent, and the same seasonal shape is there, with the third quarter highest and the second lowest.

Run the two years against each other and something useful appears. The margin gain shows up in every quarter, at 2.58, 2.33, 2.68 and 2.01 points, against a full year gain of 2.43 points. Revenue growth by quarter came in at 13.46, 12.37, 14.75 and 14.85 per cent against 13.92 per cent for the year. A gain present in all four quarters is not something that happened in one unusual quarter. The finding is genuine, and about as far as two years of data can honestly carry the conclusion.

Two years cannot separate the season from a trend. Two years give exactly two observations of each season, and two points establish a line through anything at all. Whether the third quarter is reliably the strongest, or was strongest twice by accident, is not a question this record can answer, and saying so is worth more than producing a seasonally adjusted figure the data cannot support.

What does the finished worksheet actually contain?

The complete sequence runs as follows on Sarvani Coatings Limited's year three. Every row is either a document somebody else can open or a division they can redo on the back of an envelope.

The run refuses to produce a good deal. The sheet does not say whether the margin gain was earned or handed over by input prices. The sheet revises nobody's forecast. Nothing in it reaches a view on the company, and certainly nothing reaches a view on the shares. The refusals are what make the sheet usable by somebody who disagrees with its author, and that usability is the only test of this kind of work that matters.

Who runs this, and on what morning?

A sell side analyst runs the whole eight on results morning, in order, before writing a word. The first four checks take longer than the arithmetic does. The imbalance feels wrong the first few times, and stops feeling wrong the first time check two catches a redrawn segment boundary that would otherwise have made a paragraph of confident nonsense.

A lender's credit analyst runs a narrower version and cares most about checks seven and eight. A borrower who quotes a run rate built from the strongest quarter has effectively asked to be lent against Rs 580 crore of EBITDA when the year produced Rs 446 crore, and the difference is Rs 134 crore of coverage that was never there. The absences from check eight matter to a lender for the same reason: the lines a release omits are often the lines that decide whether the cash actually stayed in the business.

Somebody holding a handful of shares, with one free evening in the week, runs the shortest version of all and keeps most of what it is worth. Open the filing instead of the headline. Put the four quarters on one line before reading any single one of them. Never multiply one by four. The work costs nothing but a willingness to write down what is not known before explaining what is, so more time and better tools make it faster rather than better.