How to Analyse Profit Margins: A Procedure in Six Steps

Analysing margins is a procedure, not a judgement. Six steps, in order: check the basis, compute the ladder for both years, locate the rung where the shape changes, reconcile in rupees, test the alternative explanations, and write down what would settle it. Run in order, it produces a list of questions with the evidence each one needs. The procedure never produces a verdict, and a run that reaches one has skipped a step.

Underneath that sits the following. Each of the six steps hands something specific to the step after it, and the handover is what fixes the order. Step one hands over a pair of figures that can honestly be put side by side. Step two hands over eight numbers. Step three hands over one rung. Step four hands over a set of named rupee movements. Step five hands over a candidate cause for each movement, or the absence of one. Step six turns whatever is left into questions with named evidence. Remove any step and the step after it has nothing to work on. The six steps are a sequence, not a checklist that can be ticked in any order.

How a margin is computed, what gross profit measures, how a reconciliation is built and what any of these numbers mean about a business are each settled in their own right. The procedure adds the order, the stopping rule, and a plain statement of what comes out at the end. The most common way margin analysis goes wrong is not an arithmetic error at all. The right steps get run in the wrong order, or the analysis stops one step early and calls the result a finding.

Why is checking the basis the first step rather than the second?

Step one is to establish that the two years were prepared the same way. Four things get read, and not one of them is computed. The accounting policy noteThe note in a set of accounts that states the measurement choices the business has applied, such as which cost formula it uses to value inventory. gives the cost formula and whether it changed. The description of what goes into cost names which expenses sit above the gross line and which sit below it. The revenue line states what it is net of. And the cover of the statements says whether the set in hand is a standaloneA set of accounts covering the parent company by itself, with no subsidiary added in. set or a consolidatedA set of accounts that adds a parent and the subsidiaries it controls together and presents them as though they were one business. one. Step one is first because every step after it compares two figures, and two figures prepared on different bases are not a comparison at all, whatever the arithmetic says.

The household version comes before the accounts. An electricity bill was Rs 2,400 last quarter and is Rs 3,600 this quarter, so it is up half again. Then it emerges that last quarter's bill covered one flat and this quarter's covers the flat and the shop downstairs, taken over in April. Nothing about the arithmetic was wrong. The two numbers simply were not the same kind of number, and every conclusion built on their difference inherits that. Checking what a figure covers before subtracting it from another figure is not caution, it is the whole of step one.

Run step one on Anjani Stationers Private Limited and four answers come back. The cost formula is first-in-first-out and it did not change between the two years. Both years are presented on the standalone basis, the parent company by itself. The basis matters more than it sounds. Chitra Binding Works was bought at the start of year two, and on a standalone basis that business's own payroll and overheads are not inside any of these lines at all. Year two's balance sheet carries one thing year one's did not: an investment of Rs 21,00,000 where the purchase went. Revenue is stated net of goods and services tax in both years. The prescribed format has no cost of goods sold line, so the published gross profit of Rs 1,21,50,000 and the 45.0 per cent that comes out of it are a materials margin rather than a full cost of goods sold margin. A build on the location and condition test gives 44.1 per cent instead. The same basis applies to both years, so the analysis proceeds. But the materials basis fixes what the 45.0 per cent may afterwards be compared with, and step one is where a reader writes that down rather than discovering it at step five.

The basis check has just done something for the rest of the procedure, and it is the clearest illustration of why the order is what it is. Knowing the figures are standalone establishes what the cost lines cannot contain, and that is worth as much as knowing what they do contain. A reader who assumed the statements were consolidated would spend step five looking for a subsidiary's payroll inside employee benefits, and it is not there and never was. The acquisition still reaches these lines, but by a different route, and finding the right route is only possible because step one settled the basis first.

What does step two hand to step three?

Step two computes the whole ladderThe sequence of profit figures from gross profit down to profit after tax, each rung carrying more deducted costs than the rung above it. for both years. Four rungs, two years, eight figures, and no shortcuts. The temptation at this step is to compute the rung already suspected and move on. Step three does not examine any rung on its own, so the temptation has to be resisted. Step three examines the difference between one rung and the rung above it, so a ladder with a rung missing has a hole exactly where the finding would have been. Step two is not where anything is found; it is where the material step three examines gets assembled, and a partial ladder cannot locate a change because locating is a comparison between rungs and not a reading of one.

How each of the eight figures is computed is settled in its own right. Step two is a pointer: all eight are computed, both years, and laid out in one place where the two years sit against each other. For Anjani Stationers, laid out that way, year one reads 45.0, 24.2, 22.1 and 15.8 per cent, and year two reads 45.0, 19.8, 15.4 and 11.1 per cent. The two rows repay a moment's attention before anything is done with them. Six of the eight numbers are lower in year two, one is unchanged and none is higher, and stopping here yields exactly one true sentence: profitability fell at every rung except the top one. Such a sentence is worth very little, and step three is what turns it into something.

Which rung does step three point at?

Step three compares the movement at each rung with the movement at the rung above it, and marks the first rung where the two differ materially. The marked rung is where the change entered the business. The reasoning is one sentence long: each rung carries everything the rung above it carried plus one more set of costs, so if a rung moved and the rung above it did not, the movement was produced by the costs that sit between them and nowhere else. Step three establishes where the change entered and not what changed, and treating a location as a cause is the most common way this procedure gets misread.

Run it on Anjani Stationers. The gross rung moved 0.0 points. The earnings before interest, tax, depreciation and amortisation (EBITDA) rung moved 4.4 points down, a divergence of 4.4 points from the rung above it. The first divergence is the location, and the EBITDA rung is the first. Below it, the earnings before interest and tax (EBIT) rung moved 6.7 points down in total, a further 2.4 points beyond the EBITDA rung, so something else entered between EBITDA and EBIT as well. Then the net rung does something a reader watching only the direction of travel will miss entirely. The net rung moved 4.7 points down, less than the EBIT rung's 6.7, so between EBIT and net the shape moved back up by about 2.0 points.

Three divergences, then, and one of them runs the other way. Step three has delivered two things and withheld a third. Step three has named the cost groups worth reconciling: the ones between gross profit and EBITDA, the ones between EBITDA and EBIT, and whatever sits between EBIT and net profit. The gross rung did not move at all, so step three has established that the trading itself, measured at that rung, is not where the deterioration came from. And step three has said nothing whatever about why. A reader who stops here and says the business is spending too much on overheads has converted a location into an accusation across one silent step.

Why does step four insist on rupees when step three has already found the rung?

Step four builds a bridgeA line by line reconciliation from one period's figure to the next period's figure, in which every named movement adds up to the whole of the change with nothing left over. from last year's figure at the located rung to this year's, naming every movement, until the total closes exactly. The bridge is a discipline rather than a discovery. Percentages point at a place; rupees prove that something is there and say how much of it there is. Until the movements sum to the whole of the change, something unnoticed cannot be ruled out as the thing doing the work. A location that does not reconcile in rupees has not been located.

The value of the closing is that it is falsifiable. Three named movements coming to Rs 4,20,000 against a fall of Rs 4,50,000 is not a small error; it is a fourth movement not yet found, and the arithmetic is saying so. Step four is where a reader who has been guessing finds out. The same fact is why step four is unpopular. The other direction is the one readers get wrong, so it has to be stated exactly. Once every line the statement puts between the two rungs has been named, those lines have to sum to the change; the statement is built that way, and the bridge does not discover it. Closing therefore says one thing, that no line was left out and no arithmetic slipped, and it rules out neither of the two things a reader would like ruled out: that two movements never named cancelled each other, and that whatever is really doing the work is sitting inside a movement that was named. Step five exists because of the second one. Building the bridge itself, the ordering of the lines and the treatment of the sign, is settled elsewhere. Step four is a pointer: the bridge is built, and nothing proceeds until it closes.

On Anjani Stationers, at the located rung, EBITDA was Rs 58,00,000 in year one and Rs 53,50,000 in year two, a fall of Rs 4,50,000. Three movements close it. Gross profit rose Rs 13,50,000, employee benefits rose Rs 6,00,000, and other operating costs rose Rs 12,00,000. Rs 13,50,000 up against Rs 18,00,000 down is Rs 4,50,000 net, exactly. Carry on down and depreciation and amortisation adds Rs 7,00,000 more of cost to reach EBIT, finance cost adds Rs 50,000 to reach profit before tax, and then the tax charge falls Rs 4,00,000. The tax fall is the 2.0 points the net rung recovered. Now look at what the rupees ranked that the percentages could not. Other operating costs at Rs 12,00,000 is the largest single movement below the gross line, nearly half of the Rs 25,00,000 of extra cost, and it grew 86 per cent on a revenue base that grew 12.5 per cent. Other operating costs is the movement the rest of the procedure has to deal with.

What does step five test before any of it is believed?

Step five takes each movement step four named and asks whether something other than trading could have produced it. Five candidates get checked every time: a change of cost formula, a change of classification between heads, an inventory write-down, an acquisition, and a one-offAn amount arising from something not expected to repeat, such as a settlement, a relocation cost or a single large legal fee.. Each of the five has its own note in the accounts that would confirm or rule it out, and step five is done when all five have been looked at for each movement. Step five exists because every one of those five produces a cost movement that looks exactly like an operating one on the face of the statement, and the face of the statement is not where they are distinguished.

The everyday shape of it is familiar. A household's monthly grocery spend has gone from Rs 9,000 to Rs 16,000, and the question is whether whoever runs it has lost their grip. Before an answer comes the questions of whether anyone moved in, whether a weekly market was swapped for a monthly bulk buy, and whether one of those months included a wedding. None of those questions doubts the arithmetic. The questions ask whether the two months are the same kind of month. Step five is that instinct, applied in a fixed order so it does not depend on whether it happened to come to mind.

Run it on Anjani Stationers and the five come back unevenly. The cost formula is ruled out at step one, first-in-first-out both years. The closing inventory figure of Rs 28,00,000 is a clean cost figure with nothing charged against it, so the inventory note rules out a write-down. A cost moving between employee benefits and other operating costs would leave the gross rung untouched, so a classification change cannot be ruled out entirely. Gross margin held at exactly 45.0 per cent, so nothing crossed the gross line. The notes break other expenses into components, and those components have not been read yet, so a one-off remains open. And the fourth candidate fires: Chitra Binding Works was bought at the start of year two, and the Rs 21,00,000 investment sitting in year two's balance sheet where year one's carried nothing is the flag that says so.

Step one already fixed what the acquisition finding can and cannot mean here. The figures are standalone, so the binding operation's own payroll and its own overheads are not inside employee benefits or other operating costs and never were. The acquisition reaches these lines by two other routes instead. The first is the transaction itself: the professional fees, the due diligence and the stamp duty of buying a business are expensed as they are incurred, so they land in other expenses in year two and appear nowhere in year one's comparative. The second is the relationship: work Anjani Stationers used to do in its own shed, or used to buy from somebody unconnected, may now be bought from a business that became a related partyA person or business connected closely enough to the reporting business that their dealings are disclosed separately, because the terms may not be the terms a stranger would have got. in April. Every rupee of that work sits in other operating costs looking exactly like an ordinary overhead. An unknown part of the Rs 12,00,000 rise is therefore the price of a transaction and of a new related party relationship rather than the cost of running the existing operation. The face of the statement quantifies neither, and an unquantified part is precisely the shape of thing step six exists to write down.

The readings from the walker are these. At step one, six basis items are settled and nothing is open. Nothing has been found yet. At step two, two rows of four margins are settled and the open list is still empty. Computing feels like progress, so the empty list surprises people. The open list first moves at step three, to two questions, and again at step four, to four. Naming a movement in rupees is precisely what turns a vague sense that something happened into a specific thing that cannot yet be explained. Step five does not shrink the list either: it re-points all four questions at a named candidate and adds a fifth, so the walker finishes with five open questions and zero conclusions. The open-questions count rises at every step from three onwards and finishes at its highest. A reader expects a good procedure to do the opposite, and the rising count is the correct behaviour. Run it with step five skipped and the count instead falls to zero at step six, replaced by one confident sentence about overheads that names the wrong entity, which is exactly how an analysis feels most satisfying at the moment it becomes wrong.

When does step six stop, and what does it hand over?

Step six writes down, for every question still open, the specific evidence that would settle it, and then stops. Not the answer, not a best guess, not the most likely explanation weighted by experience. The evidence: a named note, a named disclosure, a named figure that would have to be obtained. The stopping rule is exact and does not depend on the time available. Stop when every movement is either reconciled in rupees or written down as an open question with its evidence named. Why the business made the choices it made is not in the statements, and no amount of further work on the statements will put it there, so the procedure does not continue into it.

The reason the rule is written that way, rather than as a length of output or a time budget, is that it is the only version another person can check. Handed to a colleague, the output can be verified: is every movement either bridged or listed? If yes, the work is complete whatever it concluded. If no, it is not complete however confident it sounds. A stopping rule that can be failed is worth having. A rule of stopping once the thing feels understood is worth nothing. The feeling arrives earliest for the reader who knows least.

Anjani Stationers ends with five questions. How much of the Rs 12,00,000 rise in other operating costs is now being paid to Chitra Binding Works, for which the related party transactions note is the evidence. How much of that same Rs 12,00,000 was the one-time cost of making the purchase, for which the components of other expenses and any exceptional items line are the evidence. Why employee benefits rose Rs 6,00,000, a rise of 16.7 per cent against revenue growth of 12.5 per cent, for which the employee benefits note is the evidence. Where the Rs 7,00,000 rise in depreciation and amortisation came from, and specifically the amortisation, since the line was called depreciation alone in year one, for which the fixed asset and intangible notes are the evidence. And why the effective tax rate fell from 24.0 per cent to 21.1 per cent, for which the tax reconciliation note is the evidence. Five questions, five named places to look, and not one conclusion about the business.

What must never be a step in this procedure?

Three things look like natural next steps, feel like the payoff, and are not part of this procedure at all. Comparing the margin to an industry average is the first: two businesses are two bases, and step one exists precisely because two bases are not a comparison, so an average of many bases is not one either. Deciding whether a level is good is the second. Good requires a standard, and no standard is present in a set of accounts. Forming a view on the business is the third, and it is the one that arrives dressed as the obvious conclusion of everything just done.

The procedure produces questions, and treating its output as a conclusion is the single most common way it is misused. The output looks so much like a conclusion that the last step gets taken without anybody noticing they took it. That is not an argument against ever reaching a view. The argument is that reaching one is a different activity, needing inputs this procedure never touched, and that the join between the two is where a reader has to be honest about which one is being done. The moment the sentence the business has lost control of its costs gets written, a procedure that runs on published figures has been left behind for one that does not, and that sentence carries none of the evidence the previous six steps so carefully assembled.

Who runs this procedure, and what do they do with the output?

Three people run these six steps on the same set of accounts in the same week, and none of them wants the same thing out of it. A lender runs the procedure to find out whether the cash cushion above its interest is thinning, an analyst runs it to separate what will repeat from what will not, and Vaidehi Rao, sitting inside the business as its finance controller, runs it to check whether the story she is about to tell matches the arithmetic somebody outside will do.

Watch each of them use the output. A fall concentrated in a line that will keep falling is a different exposure from a fall caused by a single year's integration costs, so the lender starts at step three's location and goes straight to the rupee bridge. The lender reads the five open questions as five things to put in a letter. A lender can ask, so the answers arrive, and asking is the advantage a lender has over everyone else. The analyst uses the output differently. She cannot ask, so the value of step six for her is knowing exactly which parts of her model are resting on an assumption rather than a figure. The Rs 12,00,000 in other operating costs is not one number to her; it is a number whose split between what will repeat and what was the price of a transaction she has assumed, and step six is where that assumption gets written down where she will see it again next year.

Vaidehi Rao has the strangest use of all. She runs the procedure against her own accounts before anybody else does. She knows the answers to all five questions. She has the ledgers, so she can see exactly how much of the Rs 12,00,000 was professional fees on the purchase and how much is now being paid across to Chitra Binding Works. She is checking whether a reader who cannot see any of that could reach a defensible reading from what is published, and where they would go wrong if they hurried. If the procedure run from outside lands on the wrong cause, that is a disclosure problem she can fix in a note this year rather than an argument she has to have next year. The procedure is the same six steps in all three hands; what differs is what each of them can do about the questions at the end.

So what is the last line of this analysis?

Not a verdict. Anjani Stationers' gross margin held at exactly 45.0 per cent while every rung beneath it fell, the deterioration reconciles to the rupee across employee benefits, other operating costs and depreciation and amortisation, and a subsidiary bought at the start of year two for Rs 21,00,000 sits underneath the largest of those movements by a route the published statements do not quantify. Six steps, run properly, come to rest on that reading. The procedure runs on published figures, and the published figures genuinely do not contain the answer. Ending on a question is therefore the correct outcome of the procedure and not a failure of it. Any last line that supplied an answer would have been supplied by the analyst rather than by the evidence.

The difference between the two endings is the whole point of the six steps. One ending says the business has lost control of its overheads. The sentence sounds finished and is not supported. The other ending says other operating costs rose Rs 12,00,000, part of which was the price of buying a business and part of which may now be going to that business, and names the note that would settle how much. The second is harder to write, less satisfying to read, and is the one that survives being checked. So the last line of the analysis is not a statement at all. The honest last line is the question, written down exactly as it stands: how much of the Rs 12,00,000 rise in other operating costs was the one-time cost of buying Chitra Binding Works or is now being paid across to it, and how much is the business that was already there simply costing more to run?

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