Sector Research: Reading the Field Before the Company

Sector research reads the conditions every company in a field faces before reading any one of them. Sector research exists because a large part of what happens to a single company is decided outside it, by input prices, demand conditions and the cycle. Reading the company first makes shared conditions look like company achievements.

The ordering error described there does not feel like an error while it is being made. The error feels like insight. The analyst has read a set of accounts carefully, has found something that moved, and has explained it. Nothing in the accounts objects.

Why does sector research come before the company at all?

Ask what a company controls and the list is shorter than most people expect. A company controls what it decides to make, roughly what it decides to charge, whom it hires, where it spends. Raw material costs are not on the list. Neither is whether households are repainting this year, nor whether the manufacturers who buy industrial coatings are building anything. All three arrive from outside, they arrive at every competitor at the same time, and they land in the accounts looking exactly like decisions.

Reading the company first makes shared conditions look like company achievements, and this single ordering error accounts for a very large share of the confident wrong readings in equity research. It is not a subtle mistake with a subtle cost. The error produces a view that will hold for exactly as long as the outside condition holds, and will then reverse for a reason the analysis never named. At that point the analyst has no idea what went wrong, since nothing in their own work has changed.

Take it out of finance for a moment. Ten shops sit inside one shopping centre. One of them has a very good year: footfall up, takings up, the owner delighted and starting to talk about what they did differently with their window display. A new metro station opened two hundred metres from the entrance, and walking the corridor reveals that all ten shops had the same year. The window display may still have helped. But that cannot be known until the other nine have been counted, and an observer who never walks the corridor will write down the window display as the reason and believe it.

Sarvani Coatings Limited, an invented maker of paints and industrial coatings, appears throughout, and its year three accounts contain exactly that shape. Its gross marginRevenue less what the goods sold actually cost to make, expressed as a share of revenue. How it is built from the statements is settled in the accounting material. rose from 43.0 per cent to 46.0 per cent across two years, a gain of 3.0 points, and profit after tax rose 41.1 per cent in the most recent year alone. Read on its own, that is a company that has learned to price and to run itself. Read against the field, it may be something quite different. The corridor has to be walked.

Top-Down: what does starting from the whole field provide?

The top-down route starts wide and narrows. The top-down route begins with the economy or with a field inside it, establishes what conditions that field is under, and only then looks for the companies those conditions favour. The order matters: by the time a single set of accounts is opened, what a normal year looked like for everyone is already known, so there is a benchmark to read the accounts against rather than reading them cold.

The top-down route is genuinely good at noticing that a whole field is moving, and poor at identifying which company inside that field will actually capture the movement. Those are two different skills and it only has the first. A field can grow 11.0 per cent while a particular maker in it grows 3 per cent, loses market shareOne company's revenue expressed as a share of everything sold in the field. A gain of a fraction of a point is a real gain and a small one, and describing it without its size overstates it. and finishes the year worse off than it started. The top-down reading was right about the field and useless about the company.

The limitation is easy to feel in a familiar setting. Somebody reports that a particular street has become the place people go for evening food, and the report is right: the whole street is busier than it was two years ago. The report says nothing at all about which of the fourteen stalls on the street is making money. Some are packed and pricing badly. One has a queue because it is genuinely better. Two are about to close. The street is the field, and knowing the street is up is a real piece of knowledge that stops well short of the stall.

Bottom-Up Research: what does starting from one company provide?

The bottom-up route starts narrow. The bottom-up route takes one company, reads it in its own right, works out what it makes, what it costs to make, who buys it and on what terms, and reaches the field only as context for what has already been found. The advantage is depth. The analyst ends up knowing the business rather than knowing a category, and every specific thing learned is attached to a real set of accounts rather than to an average.

The bottom-up route understands one specific business properly, and its weakness is mistaking a shared tailwind for a company property. That weakness is not incidental to the method; it is built into it. The company under examination is the only thing in view, so without a sideways look everything found looks like it belongs to that company. A genuine competitive advantageSomething a company has that rivals cannot easily copy. Returns stay above a rival's for as long as the copying stays hard. Competitive advantage is defined and tested in the business analysis material. and a shared condition produce identical evidence in a single set of accounts, and the accounts cannot say which of the two is present.

The full comparison of the two routes, including when a researcher should deliberately pick one over the other, is set out under top-down and bottom-up research. Each route is blind exactly where the other sees. Practitioners therefore rarely use one alone.

Industry Cycle: how is it different from a trend?

An industry cycle is a pattern of expansion and contraction that repeats, without being regular in either its length or its depth. The irregularity is what makes a cycle hard to work with. People hear the word cycle and picture something with a period regular enough to set a watch by, and then when the contraction arrives eighteen months later than the last one did, they conclude there was no cycle after all. There was. Cycles repeat; they do not keep time.

A cycle returns and a trend does not, so mistaking one for the other alters the expectation of what comes next in the most consequential way available. If a contraction is cyclical, the right expectation is that conditions come back, and a company that survives the trough is worth more than its trough earnings suggest. If the same contraction is the start of a trend, conditions do not come back, and the trough earnings are the new normal or the first step down towards it. Same data, opposite conclusion, and the difference is entirely in which of the two the analyst decided was present.

The five year volume record for this field is worth reading slowly. Volume grew 2.1 per cent, then 6.8 per cent, then fell 1.4 per cent, then grew 7.2 per cent, then 4.5 per cent. Averaged, those come to 3.84 per cent a year. One year in five went backwards, and it went backwards inside a field that was growing at nearly 4 per cent a year across the whole stretch. The swing about that average runs from plus 3.4 percentage points in the best year to minus 5.2 in the worst.

The single negative year does something particular to a careless reader in real time. In the year it happened, nobody had the two years that came after it. All that was available was a field that had grown twice and then shrunk, and the honest position at that moment was that one year of contraction cannot separate the two possibilities. The recovery settled it, and the recovery was only available later.

How Sector Drivers Flow into Equity Research, traced once and slowly

Saying that field conditions reach a company is one thing. Pointing at the line in the accounts where they land is another. So here is the path, traced once, with nothing skipped. A shared input price moves. Every maker in the field buys that input, so every maker's cost of materialsThe line in the profit statement that carries what the goods sold actually cost in raw materials. Built and defined in the accounting material. moves. Cost of materials sits directly above gross profit, so gross margin moves. The costs below gross margin did not move with it, so the gain grows in percentage terms as it flows down into operating profit and then into profit after tax.

The transmission is completely visible in the accounts and completely invisible in the story. Finding it takes a deliberate look. Nothing in a set of statements is labelled. The cost of materials line does not carry a note saying which part of its movement came from outside. The line shows one number, and one number is what gets read.

In the case record the input side is a small set of pigments and resins whose prices move with crude oil. A move therefore reaches every maker at roughly the same time rather than one at a time. Sarvani Coatings buys them, Nandivarman Paints Limited buys them, Kesaria Surface Solutions Limited buys them, and Thottam Chemicals Limited sits one step further up the chain and sells them. When that input moves, four sets of accounts move together, and none of the four decided anything.

Watch it arrive in the numbers. Sarvani Coatings' cost of materials was 57.0 per cent of revenue in year one and 54.0 per cent in year three. Gross profit therefore went from 43.0 per cent to 46.0 per cent, a gain of 3.0 points. In year three that is Rs 1,111 crore of gross profit on Rs 2,415 crore of revenue. Further down, in the most recent year alone, profit after tax rose 41.1 per cent while revenue rose 13.9 per cent. A margin gain landing on top of a cost base that did not grow with it makes each line down the ladder grow faster than the line above.

One detail here matters more than it looks, and it is the part most readers get backwards. Over the most recent single year, from year two to year three, input cost per unit of output did not fall. Input cost rose about 3.6 per cent, and realisationWhat a company actually gets per unit of what it sells, after discounts and whatever the mix of products happened to be. Worked out in the earnings material. rose faster, at about 7.5 per cent. The margin improved because the selling side outran the cost side, not because the cost side got easier. A selling side that outran the cost side is a different mechanism with different things that would reverse it, and confusing the two produces a forecast that fails in a way nobody prepared for.

What happens when the whole thing is run on one company?

Now put it together on Sarvani Coatings Limited, field first and company second. The field first. Field revenue was Rs 48,300 crore against Rs 43,500 crore the year before, a rise of 11.0 per cent. Nandivarman Paints Limited holds a 30.0 per cent share of it, Sarvani Coatings 5.00 per cent and Kesaria Surface Solutions Limited 3.0 per cent, so the three together are 38.0 per cent and the remaining 62.0 per cent is spread across many smaller makers.

Only now open the company. Sarvani Coatings grew revenue 13.9 per cent against a field at 11.0 per cent, so it outgrew the field by 2.9 percentage points, and its share moved from 4.87 per cent to 5.00 per cent, a gain of 0.13 percentage points. The share gain is real and it is a small one, and describing it as taking share without stating its size would overstate it by a wide margin.

Then the margin, the interesting part. Sarvani Coatings gained 3.0 points of gross margin over the two years. Walking the corridor shows that Nandivarman Paints gained 2.4 points and Kesaria Surface Solutions gained 3.6 points over the same period. All three rose.

The peer evidence narrows the question sharply and does not resolve it, and holding both halves of that sentence at once is the single most useful habit in the work. It narrows it because a pricing environment shared across the whole field explains a field wide gain and one company's pricing does not: when every maker in the peer setThe handful of comparable companies a researcher reads a company against. Which companies belong in one, and on what basis, is settled in the comparison material. gains, the shared explanation is the one doing the work. Sarvani Coatings sits between the two peers rather than outside them, so the question stays open. Nothing in these three numbers separates its own pricing from a mix shiftA change in what a company sold rather than what it charged, for instance more industrial and less decorative, which can lift the average price achieved without any price rise at all. towards industrial work that would lift realisation and input intensity together.

Research Error: what counts as a failure in the work itself?

A research error is a failure in the work. An assumption used but never stated. Evidence that was available and was not checked. A mechanism read backwards. A figure quoted at one scale and compared against another. A conclusion that does not follow from the material underneath it. Each of those is a specific, locatable defect, and each of them sits inside the document.

A research error is identifiable from the work alone, without knowing anything at all about what happened afterwards. A note can be handed to somebody with the date covered and nothing said about the year that followed, and that reader can still find the unstated assumption. The test is a strict one. If what is wrong with a note cannot be stated without referring to how things turned out, no error in it has been found.

The error on offer in this case is a good one to hold in mind. An analyst reads the 3.0 point margin gain, writes that Sarvani Coatings has improved its pricing and its execution, and files it. The defect is not the conclusion. The conclusion might even be right. The defect is that a company specific conclusion was drawn from evidence that was never tested against the field, and the test was available, cheap and one column wide.

Market Outcome: what actually happened afterwards?

A market outcome is what happened next: to the price, and to the business. An outcome arrives after the work is finished, and everybody remembers it. An outcome is the part with a number attached that anyone can check without reading anything.

An outcome is one draw, and it is affected by a great deal that the work never claimed to predict. A year contains a monsoon, a change in what households do with their savings, a decision by a competitor, a crude price move nobody forecast, and the collective mood of everybody who happened to be transacting in a share on the days that mattered. A research note claims none of that. A note usually claims something much narrower, such as what a margin was and where it came from, and then the year happens to it.

A doctor advising a patient to stop smoking has done good work whether or not that particular patient goes on to develop something anyway, and a patient who smokes for fifty years and stays well has not proved the advice wrong. The advice was about what raises and lowers the odds. The individual case is one draw from those odds and cannot audit them. Research works the same way and is judged, almost universally, as though it did not.

Research Error vs Market Outcome: why are these two independent?

Because they are answers to different questions. The first asks whether the work was done properly, and it is settled by reading the work. The second asks what happened, and it is settled by waiting. Nothing forces the two to agree, and in practice they disagree constantly. Careful work is followed by a bad year. Careless work is followed by a good one. All four combinations occur and all four are ordinary.

Judging research by how it turned out has a name, resulting, and the name belongs to Annie Duke, Thinking in Bets, 2018. Naming it is more useful than it sounds, because once a habit has a name it can be caught in the act. The outcome is loud and arrives on its own. The quality of the work is quiet and has to be gone and looked for. Resulting is hard to catch for that reason alone.

Outcomes on their own cannot identify which parts of the work were sound, so a research process improves only if errors are found by reading the work. Grading by outcome keeps whatever was done in the years that went well and discards whatever was done in the years that did not, and since both sets contain a mixture of good and careless work, the reinforcement is random. Years of that leaves a process nobody can defend and nobody can fix.

The most dangerous of the four boxes is not the obvious one. Careless work followed by a bad year gets reviewed. The bad year forces the review. Careless work followed by a good year gets filed away as a success, and the method that produced it gets used again, and nothing anywhere in the process raises a hand.

What does sector research produce, and where does it stop?

Sector research produces three things and no fourth. A statement of the conditions the field faces, written plainly enough that somebody else could disagree with it. The evidence for each condition, enough of it that the statement can be checked rather than believed. And a list of what would change each condition, naming in advance what is being watched for and leaving nothing to be reinterpreted after the fact.

A conclusion about which company will do well requires the company work that comes afterwards, so sector research never produces one. Stopping short is not modesty and it is not hedging. The stopping point is the boundary of what the method can support. The field evidence in this case narrows Sarvani Coatings' margin question a great deal and cannot close it, and closing it anyway would add an assertion where the evidence stopped.

Ending there is the correct place to end. A reader who wants to be told which maker to back will find the ending unsatisfying, and the alternative on offer is a confident answer that the material underneath does not support. The conditions are stated, the evidence is on the table, what would change it is named, and the company question stays open until the company work is done.

How does an analyst actually work this, in practice?

Meghna Iyer covers coatings and paints and has Sarvani Coatings on her list. Before she reads a single line of the company's most recent statements she does three things, in this order, and they take her a morning. She writes down what the field did on volume and on price, Rs 48,300 crore of revenue and a rise of 11.0 per cent, so she has a benchmark. She writes down what happened to the main shared input, so she knows what pressure everybody was under. And she writes down what the two closest comparable makers reported on the lines she is about to read. When she opens the company she is then reading a difference rather than a level.

A benchmark chosen after the fact is not a benchmark, so the discipline that does the work is establishing it before the company number is seen. Once a reader has seen that a company gained 3.0 points, it becomes remarkably easy to decide that 2.4 was roughly the same thing, or that 3.6 was a special case. Writing the comparison down first removes that option.

A lender does an almost identical thing without calling it sector research. Before advancing against a small manufacturer's receivables, a credit officer asks how the whole trade has been going. A borrower whose sales are down 8 per cent in a trade that is down 12 per cent is in a very different position from one down 8 per cent in a trade that is up 6 per cent. Same number, opposite meaning, and the difference is entirely outside the borrower's own books. A household does it too, in miniature: before deciding whether the electricity bill is a problem, it asks what the neighbours' bills did this summer.

At the end of the morning Meghna Iyer will not state which maker in this field will do well. She has the conditions, the evidence and the list of what would change them. The company work comes next, and until it is done the honest output is the one she has.