How to Research a Listed Company, in Order: Eight Steps

Research a listed company in a fixed order. Settle what it sells, read the field it sells into, read three years of its own reporting oldest first, build the revenue line, separate what the company did from what happened to it, and only then build a model. Finish by writing down the two or three questions the answer turns on. The model is step six, not step one.

The order is not a matter of taste. Each step hands the next one something it needs, and a step taken out of turn still produces an output, just with the checking quietly removed. Reading the field before the company is what stops a shared condition from looking like an achievement. Separating shared movement from unshared movement is what establishes which findings belong to the company at all. And a model built before that separation will compute, faithfully and to two decimal places, the consequences of an assumption nobody ever looked at.

In what order is this actually done?

Eight steps, and their sequence is the whole of the method. Five of the eight sit before the one most people begin with. Everything below runs on Sarvani Coatings Limited, an invented maker of decorative and industrial coatings.

By the time a model gets built, which of the company's numbers are its own is already known, and the order exists to produce exactly that. Take that away and the steps become a checklist, and a checklist can be worked in any order without complaint.

1. Say what the company sells, and to whomTwo sentences, no adjectives. What the products are, who buys them, and how they reach the buyer.
   Checking: can this be said out loud to somebody who has never heard of the company, without reading?
2. Read the field before the companyHow big it is, how fast it grew, who else sells into it, and what moves all of them at once.
   Checking: can what the whole field did last year be stated in one number?
3. Read three years of the company's own reporting, oldest firstThe filed statements and the company's own commentary, taken in the order they were written.
   Checking: is the direction each main line has been moving clear, and since when?
4. Build the revenue line before anything else in the accountsWhat is sold, how much of it went out, at what price, to which kind of buyer.
   Checking: can last year's revenue growth be split into volume and into price and mix?
5. Separate what the company did from what happened to itEvery movement that matters goes beside the same movement at the other makers. Shared in one pile, unshared in the other.
   Checking: for each movement, which pile is it in, and how big is the unshared part?
6. Only now build a model, and only against a stated questionWrite the question down first, then build the smallest model that can answer that question and nothing else.
   Checking: what will this model establish that is not already known?
7. Write down the two or three questions the answer turns onEach one with the evidence that would settle it, and where that evidence would show up.
   Checking: could somebody else go and look for the evidence without asking what was meant?
8. StopThe field is read, the record is read, the two piles are separated, and the questions are written with their evidence.
   Checking: is the work continuing because something is unresolved, or because the model is not finished?

What gets settled before anything is opened?

Step one asks what the company sells, and to whom. Nothing else. No spreadsheet, no filing, no price chart. Two sentences are enough: Sarvani Coatings sells decorative coatings to households through a dealer network, and industrial coatings to manufacturers on contract. Two sentences, two buyer types, one company.

Naming what a company sells gets skipped more than any other step, and it gets skipped because it feels too obvious to bother with. The step is not obvious. Every figure that arrives later has to be interpreted against something, and this sentence is the something. A rise in the average price received means one thing if the company sells more to households through dealers this year and quite another if it does not. A figure means nothing until there is something to read it against, and a reader who cannot state what a company sells in two plain sentences has nothing.

The parallel is a shop on the street. Before anybody can say whether the sweet shop had a good Diwali, they need to know that it sells sweets, mostly to walk-in households, mostly in one fortnight of the year. Nobody writes that sentence down for a shop because everybody already knows it. For a listed issuerA company whose shares are quoted on a stock exchange, so its reporting and its announcements are filed publicly rather than kept private. nobody already knows it, and the two sentences have to be written down.

Why does the field come before the company?

Because a company read first makes shared conditions look like achievements, and there is no way to un-see that afterwards. If Sarvani Coatings is opened first and its margin has improved, the improvement arrives with the company's name attached to it, and the name sticks. Read the field first, and coatings makers in general are already known to have had a better year on input costs. The same improvement then arrives as something to be split rather than something to be credited.

The field yields four small things: its size, its growth, who else is in it, and what moves all of them at once. For the field Sarvani Coatings sells into, that is Rs 48,300 crore of revenue against Rs 43,500 crore the year before, growth of 11.0 per cent, three makers holding 38.0 per cent between them and the remaining 62.0 per cent spread across many smaller ones. Nandivarman Paints Limited alone is 30.0 per cent of it.

The field is not background reading; it is the yardstick every company figure gets measured against, and a yardstick collected after the measurement is not a yardstick at all. An hour on the field before the company is opened is the cheapest hour in the whole process.

Why read three years oldest first?

Because the order a record is read in decides what can be seen in it. Read oldest first, Sarvani Coatings gives a direction of travel: revenue Rs 1,840 crore, then Rs 2,120 crore, then Rs 2,415 crore. Materials took 57.0 per cent of revenue, then 56.0, then 54.0, and gross marginRevenue less the direct cost of what was sold, shown as a percentage of revenue, and the first margin on the profit ladder. ran 43.0 per cent, then 44.0, then 46.0. Three years, one direction, and a question forming about whether the direction continues.

The same read in rupees rather than in ratios is harder to argue with. Materials cost Rs 1,048 crore, then Rs 1,187 crore, then Rs 1,304 crore. Gross profit was Rs 792 crore, then Rs 933 crore, then Rs 1,111 crore. The gap between the two widened in every one of the three years, and no percentage of anything had to be taken to see it.

Read the same company newest first and the answer is 46.0 per cent. One number, no shape. A single observation always looks like a level, and 46.0 per cent read alone is a single observation. Reading newest first is how a peak gets written down as a level, and the mistake is invisible afterwards because the note that results does not record which way the record was read.

Why build the revenue line before anything else in the accounts?

Because everything underneath revenue is a proportion of it, and a proportion that cannot be explained is a proportion that will be explained wrongly. With the revenue line built first, what was sold is known: how much of it, at what price, to which kind of buyer. Sarvani Coatings grew revenue 13.9 per cent in the latest year, made up of about 6.0 per cent volume growthGrowth measured in units sold rather than in rupees, so a price rise on the same quantity does not count as growth. and about 7.5 per cent from price and mixWhich products, grades or customer types made up the sales in a period. A change in mix moves an average without any single price moving. together.

The revenue line has consequences further down. Materials cost Rs 1,187 crore in year two and Rs 1,304 crore in year three, a rise of 9.9 per cent on 6.0 per cent more volume. Per unit of what went out of the door, the input bill went up about 3.6 per cent, not down. The materials ratio still dropped from 56.0 per cent of revenue to 54.0 per cent. RealisationThe average price actually received for a unit sold, after discounts and after whatever combination of products went out of the door. rose faster than input cost per unit did. Most confusion about margins turns out, once the revenue line is built, to be confusion about mix and price rather than about costs at all.

How is what the company did separated from what happened to it?

Every movement goes beside the same movement at the other makers, and the results sort into two piles. Anything the others also did is a condition, something that happened to the company. Anything they did not do is a candidate finding, something the company may have done. The sorting is the entire technique, it costs an afternoon, and it changes conclusions more often than any other step.

Run it on the margin first. Sarvani Coatings gained 3.0 percentage points of gross margin over the two years. Nandivarman Paints Limited gained 2.4 points over the same stretch and Kesaria Surface Solutions Limited gained 3.6. All three rose, and Sarvani Coatings sits between the other two. The margin gain goes in the shared pile, and it goes there no matter how good the company's own commentary about pricing discipline sounds.

Now run it on growth. Sarvani Coatings grew 13.9 per cent against a field growing 11.0 per cent, so 2.9 percentage points of that growth were not shared with anybody. In money, growing at the field's rate would have taken Rs 2,120 crore to about Rs 2,353 crore, and the company reported Rs 2,415 crore. The unshared part is roughly Rs 62 crore of revenue. In market shareOne seller's revenue as a percentage of the revenue of the whole field it sells into, measured over the same period. that reads 4.87 per cent one year and 5.00 per cent the next, so what changed hands is 0.13 of a point. On volume alone the gap is 1.5 points, 6.0 per cent against the field's 4.5 per cent.

Two movements, one company, one year, and they belong in different piles. The step that sorts them is the one that changes conclusions most often. The sorting also has a limit. The sorting establishes that the margin gain was not unique to Sarvani Coatings. Why the field as a whole gained margin stays unestablished, and the sorting step does not require an answer.

When does a model get built, and what is it for?

Sixth, and only sixth. Five things are known by then: what the company sells, what the field did, the direction of its own record, how its revenue is built, and the movements that are its own. A model needs exactly those five as inputs, and there is no way to obtain them from inside a model.

A model's purpose is narrower than most people expect. A model tests whether a set of assumptions is internally consistent, and shows what those assumptions imply when carried forward. A model is not a device for producing a number. A model built before the separation step computes the consequences of an unexamined assumption perfectly. The output then carries the authority of arithmetic, which is worse than having no model at all.

So the question gets written before the first formula. On Sarvani Coatings the question is worth stating out loud: what happens if the shared margin gain reverses while the unshared growth gap continues? The question needs a model with three or four lines in it, not thirty tabs, and it was already in hand before anything was opened.

What gets written down at the end?

Two or three questions, each with the evidence that would settle it, and where that evidence would appear. Not a conclusion. On Sarvani Coatings the two questions almost write themselves once the five earlier steps are done. Is the 46.0 per cent gross margin a level or a peak? And does the 0.13 point share gain repeat?

Each one gets an evidence line. The margin question is settled by watching input cost per unit through one full move in input prices. The evidence shows up in the cost of materials against volume, not in the commentary. The share question is settled by next year's volume against the field's volume, 6.0 per cent against 4.5 per cent being the figure to beat. The output of research is a set of questions with their evidence attached, and a process that produces a conclusion instead has skipped the part that made it checkable.

When does the work stop, and what must never be a step?

The work stops when the field is understood, the company's own record is read, shared and unshared movements are separated, and the questions are written with their evidence beside them. The work does not stop when the model is finished. A model is never finished. There is always one more segment to split out, one more year to extend, one more sensitivity to run, and none of it changes the two questions already written down.

And one thing must never appear as a step at all. The procedure produces no target for a share price, no rating, and no statement that a company is cheap or dear. The eight steps run right up to the point where a view would be stated, and stop there on purpose. The stopping is not modesty. A view is a different act with different obligations attached to it, and the eight steps are what somebody would need to have done before that act would mean anything.

What does the whole order look like run on one company?

Here is all eight steps on Sarvani Coatings Limited, with what each one produced and what it deliberately left open. Read the right hand column as carefully as the middle one. The open questions are the part a finished note usually loses.

Look at what step five did to the reading. Before it, a reader had a company whose margin rose 3.0 points and whose revenue grew 13.9 per cent, and both looked like achievements. After it, one of them is weather and the other is worth about Rs 62 crore and 0.13 points of share. The same statements, read in a different order, produce a different note, and only one of the two orders makes the difference visible.

Who actually works in this order, and what do they get out of it?

A credit officer at a bank looking at the same maker runs almost exactly these steps and stops in a different place. Steps one to five are identical. Whether a margin gain was earned or handed over decides whether the cash flow that services a loan is repeatable. The officer then swaps steps six and seven for a coverage test and a covenant, a different sixth step resting on the same five.

An analyst covering twelve makers uses the order to ration attention. Steps one and two are done once for the whole field and reused for every company in it. Field work is the only part of the job that gets spread across twelve names instead of one, and that is the practical reason the field comes first. Step five is then done twelve times, quickly, and it is what tells the analyst which two of the twelve are worth a model at all.

The order protects a small amount of time. A person with an ordinary evening and one company they are curious about gets the most out of it. Steps one to five can be done in an evening with published filings and a calculator, and they are the steps that decide whether the rest is worth starting. Skipping to the model is what turns a curious evening into a week, and the week does not produce a better answer, only a longer one. A household deciding whether to put money into a friend's shop does the same thing without naming it: they ask what the whole street did this year before they ask what the shop did.

One more use, and it is the one that saves the most rework. When somebody else's note invites disagreement, the order shows where to look. A note that never states what the company sells is thin at step one. A note whose margin discussion never mentions the peer setThe small group of other sellers a company is compared against, chosen because they sell into the same field rather than because they are the same size. has skipped step five, and everything after that inherits the gap.

Where the material behind this order comes from

Conduct, meaning who may circulate a written view and what has to be disclosed alongside it, is set by the regulator, and requirements and the periods attached to them get revised. The two exchanges below matter for one reason: step three goes to a company's own filed reporting, and that is where the filing sits.