How to Find and Rank the Drivers Behind One Company

Finding a company's earnings drivers is a ranking job, not a listing job. The earnings identity is written down to the line the question asks about, a unit goes on every term, and each term is then moved by one per cent to record what it does to that line. The order that falls out is almost never the order the disclosure invites a reader to care about.

Everything below rests on three things settled elsewhere. A term counts as a driver when it carries a unit and is, at least in principle, observable from beyond the company gates, and every assumption gets labelled by where it came from. Revenue is volume multiplied by realisation, and the cost of materials is volume multiplied by input cost per unit. And a driver that reaches an entire field at once behaves differently from one that belongs to a single company. Those three settle what counts as a driver and how the terms are built. Putting the drivers in order is what remains.

Why is this a ranking job rather than a listing job?

Anyone asked to list what drives a paint company will produce twenty items in ninety seconds. Raw material prices, the monsoon, the festive season, dealer incentives, the housing cycle, freight, the new factory, competition from imports. Every one of those is true. A list of twenty true things is worth almost nothing. The list does not say which one to spend Thursday on.

The useful question is not what drives earnings, it is what a small error in each thing costs. That question has an arithmetic answer, and the answer does not care what anybody believes.

Consider a tea stall outside an office gate. The person running it could get three things wrong: the price per cup, the number of cups sold in a day, or the price paid for milk. An error of one per cent on the price per cup lands wholly in the pocket. Selling the same tea at a slightly higher price costs nothing extra to make. An error of one per cent on cups sold lands only in part. Every extra cup drags its own milk and sugar and gas along with it. The two errors sound equally serious described in words. The two errors are not equally serious in rupees, and the gap between them is what a ranking measures.

What are the seven steps, in order?

1. Write the earnings identity down to the line the question asks about, and stop thereIf the question is about operating profitability, stop at earnings before interest, tax, depreciation and amortisation, or EBITDAEarnings before interest, tax, depreciation and amortisation. The operating profit line before the charge for using assets and before anything to do with borrowing. Built in the accounting layer.. If it is about the bottom line, carry on to profit after tax.
   Checking: does every term below the stopping point stay out of the identity?
2. Put a unit on every termUnits of output, rupees per unit, rupees. Anything left holding a percentage is a result rather than a driver, and gets replaced by whatever produces it.
   Checking: could this thing in principle be counted or priced in the physical world?
3. Move each term by one per cent, holding every other term stillThe effect on the target line is recorded in rupees, not in per cent, and recorded once for each term.
   Checking: is exactly one term moving each time?
4. Rank by the rupees recordedLargest first. Nothing else goes into the order.
   Checking: did any opinion about likelihood get into the sorting?
5. Mark every driver disclosed, derived or assumedDisclosed means read straight off the company's reporting. Derived means obtained by dividing two published figures. Assumed means the analyst supplied it.
   Checking: does the mark say where the number came from, and nothing else?
6. Cross the ranking against the markingPut the power order and the disclosure order side by side and read the cell that is powerful and undisclosed.
   Checking: which drivers sit high on power and low on disclosure?
7. For each of the top drivers, write down what outside evidence would move itName the evidence and name where it is found. That list is the output worth keeping.
   Checking: is each line something that can actually be obtained?

Carrying the identity further than the question needs is the quiet way to waste a day. Terms that sit below the stopping point cannot change the answer the question asked for, and every one of them added dilutes the terms above it. If the question is about operating profitability, then depreciation, the finance costThe interest a company pays on money it has borrowed, reported as one line below the operating profit. Settled in the accounting layer, and not rebuilt here. and the tax charge are all below the line and all irrelevant to it, however precisely they can be known.

Step two: what does putting a unit on every term actually do?

Step two is where the work is. A term written as a percentage has no unit, so there is nothing physical to move by one per cent, and it has to be replaced by the quantities that produce it. A term already written in units of output or in rupees passes the test and stays where it is.

Sarvani Coatings Limited, an invented paint maker, reported other expenses of Rs 460 crore in year three against revenue of Rs 2,415 crore. The ratio is 19.0 per cent. Written as 19.0 per cent of revenue, it is not a driver: move it by one per cent and what moves is a ratio whose numerator and denominator both belong to other terms. Written as Rs 460 crore, it is a driver. Rupees are a unit, and a one per cent move in rupees is a definite thing.

Step one and two run on the case: what does the identity look like?

Take Sarvani Coatings Limited, a listed maker of decorative paints and industrial coatings, and take the question to be about operating profitability. Operating profitability means the identity stops at EBITDA. Written down to that line with a unit on every term, it holds volume, realisation per unit, input cost per unit, employee cost and other expenses. Five terms, and no percentage anywhere in it.

Notice what the five cards say about volume. Volume is the term readers find least comfortable. Volume in units of output is not published anywhere in Sarvani Coatings' year three reporting. The identity still needs volume. Revenue and the cost of materialsThe rupee cost of the raw materials consumed to make what was sold, reported as its own line just under revenue. Settled in the accounting layer. both contain it. The five terms can still be ranked without ever knowing what that volume number is.

Step three: how is the effect of a one per cent move measured?

Move one term by one per cent, hold every other term exactly where it is, and write down what happens to EBITDA in rupees. Then put the term back and do the next one. Repeating that for all five terms is the whole of step three.

One per cent is chosen because it makes the five terms comparable, not because a one per cent move is equally plausible for each of them. Nobody is claiming realisation and employee cost are equally likely to be wrong by one per cent. Plausibility is a real question and it comes back at step seven, where it belongs. Letting it in at step three contaminates the only part of this procedure that has no judgement in it.

Run it on the year three ladder, one term at a time.

The one that surprises people is the third row. Volume appears twice in the identity, once inside revenue and once inside the cost of materials, so moving it lifts both. Volume carries its own cost of materials along with it, and only the difference reaches EBITDA. Realisation appears once, on the revenue side only, so all of it reaches EBITDA. The tea stall behaves the same way, with three more zeros on it.

Step four: what does the ranking look like once it is computed?

Sort by the rupees in the third column, largest first, and stop. The ranking is arithmetic and contains no judgement at all. The absence of judgement is exactly what makes the order worth producing before any judgement is applied. Nothing about the paint market, the monsoon or management got into that order. Only the shape of the ladder did.

Two facts fall straight out of that picture. Realisation is worth more than twice volume, at 2.17 times. And the two terms most companies talk about least, realisation per unit and input cost per unit, are sitting in the top two places.

The single most useful number the procedure produces

Turn the top two rows into an exchange rate. If a one per cent error in realisation costs Rs 24.15 crore, and one per cent of volume is worth Rs 11.11 crore, then a volume error of 2.17 per cent costs exactly what a 1.00 per cent realisation error costs. Getting realisation right to one per cent is worth more than getting volume right to two per cent, and almost nobody expects that before they compute it.

The 2.17 is not a coincidence and it is not specific to paint. The multiple is one divided by the gross margin. On the year three absolutes of Rs 1,111 crore of gross profit against Rs 2,415 crore of revenue, the margin is 46.0041 per cent. A company with a thinner gross marginRevenue less the cost of materials, expressed as a share of revenue. The part of a sale left over before any of the running costs of the business are paid. has a larger multiple, because more of each extra unit is handed straight back to suppliers. The relationship is worth carrying to the next company examined.

Every one of the five effects scales in a straight line with the size of the move applied, so turning one per cent into three would stretch all five bars by exactly three and leave the order untouched. The order survives a change of scale. The size of the bars does not. The five numbers stand on their own, and they can be recomputed on the next ladder that comes along.

Step five: how is each driver marked disclosed, derived or assumed?

Three marks, and one question decides which one goes on. Where did this number come from? Disclosed means it was read straight off the reporting. Derived means it was obtained by dividing two published figures. Assumed means the analyst supplied it and nothing published stands behind it.

The mark records the origin of a number and never how confident anybody feels about it. A derived figure can be more reliable than a disclosed one, and an assumed figure can be the best supported figure in the table. None of that belongs in this column. Keeping the marking mechanical lets step six say something. The moment confidence gets in, the marking simply mirrors the ranking and the crossing has nothing to reveal.

On Sarvani Coatings, year three, the marks fall out like this. Employee cost is on the face of the statement, so it is disclosed. Other expenses is disclosed, with two of its components separately named. Volume appears occasionally in commentary, so it is available in words and not in a table. Realisation per unit and input cost per unit are both derived, each by dividing a published rupee line by the volume figure that is often absent. Both inherit whatever weakness that volume figure has.

Step six: what happens when the ranking is crossed against the marking?

The whole procedure exists for step six. Put the two orders side by side and read the cell that is powerful and poorly observed.

The two most powerful drivers on this ladder are the two least observable, and the two best disclosed are the two that matter least. That is the entire output of the procedure, and it is not a complaint about Sarvani Coatings. The pattern is a structural feature of how a profit ladder is reported. Statements are built to report rupees that were spent and earned, not to report the prices and quantities underneath them. So the terms with the most leverage are the ones that have to be reconstructed, and reconstruction is where the error lives.

A procedure that stops at step four produces a neat table of five numbers, and there is nothing to do with it. The crossing is what says where to point the week.

Step seven: what outside evidence would move the top drivers?

For each driver at the top of the crossing, two things are written down: what evidence outside the company would change the view of it, and where that evidence is found. Step seven converts a sensitivity table into a research plan, and those two lines per driver are the only output of the whole procedure worth keeping.

Realisation per unit is a claim about pricing, so the evidence that moves it is evidence about pricing: what the wider paints and coatings field did on price over the same year, what a supplier one step up the chain such as Thottam Chemicals Limited was charging, what the peer set said in its own reporting. Input cost per unit is a claim about the price of resins, pigments and solvents, so the evidence is about those inputs over that same year, and about the lag between an input price moving and a finished product price following it.

Two disciplines matter in writing that list. Every item of evidence stays on the same period as the figure it is meant to move. A two year price move and a one year cost move do not belong in the same sentence. The evidence also has to be something that can actually be obtained. A research plan that names an unobtainable fact is a wish rather than a plan.

What can the ranking not show?

Sarvani Coatings' year three gross margin came in two points above year two, on volume up 6.0 per cent, and it did so because realisation per unit outran input cost per unit rather than because input costs fell. The ranking above shows why that gap matters so much: realisation is the most powerful term on the ladder. The ranking says nothing at all about where the gap came from.

Three explanations fit it. A pricing environment across the whole paints and coatings field that let every maker price ahead of its input costs. Sarvani Coatings pricing ahead of the field on its guidanceWhat management says in advance about how the year ahead is expected to go. Which parts of that amount to a commitment, and which are only a signal, is covered separately. and its brand. Or a mix shiftA change in what is sold rather than in how much or at what price: more of one product line and less of another, which moves the blended figures without any single price moving. towards industrial coatings that lifted realisation and input intensity together. The published statements separate none of the three. Segment reportingThe split of revenue and results between the separately reported parts of a business, published alongside the main statements. Settled in the accounting layer. narrows the question and does not close it. Naming what would separate them is exactly what step seven asks for, and it is where this procedure hands off.

Sarvani Coatings Limited, Nandivarman Paints Limited, Kesaria Surface Solutions Limited, Thottam Chemicals Limited, Meghna Iyer and Ravindra Setlur are invented.
Educational material. Not advice on any investment, tax, budget or market position.