Fact vs Thesis: What the Record Settles and What It Cannot

A fact is something the published record already states, or something arithmetic on that record produces. A thesis is a claim about what happens next, or about a cause the record never names, and no document available today can settle it. The margin between revenue and the cost of materials at Sarvani Coatings Limited, an invented maker of decorative and industrial coatings, rose 3.0 points across two years, and the rise is a fact. Why it rose is not.

The four parts of a written view are set out under that heading, and what an assumption is, and the order a forecast gets built in, follow from there. Taking a view apart into the handful of variables it actually rests on is covered separately. One job comes before all of them, and it quietly decides whether any of the rest can be argued with: every sentence a research note contains sorts into one of two piles.

The sorting is already done, and done well, outside finance. If a neighbour says the bus fare went up to Rs 25/- last Monday, anybody can walk to the stop and read the board. If the same neighbour says the fare went up because diesel got dearer, that cannot be checked anywhere, and if they add that it will go up again by Diwali, there is nothing at all to check yet. Three sentences, one voice, three completely different kinds of thing. A research note is that conversation written down in a more confident font, and the only reason the sorting feels harder there is that the sentences are longer and carry numbers.

What counts as a fact in equity research?

A fact is a figure sitting in a lodged document, or a figure that ordinary arithmetic on lodged documents produces. Revenue of Rs 2,415 crore in year three is a fact. Cost of materials of Rs 1,304 crore in the same year is a fact. Gross profit of Rs 1,111 crore is a fact too. No line in the accounts is labelled that way, but the figure is the first minus the second, and anybody subtracting gets the same answer.

The definition does not mention whether the figure is important, whether it is interesting, or whether anybody believes it. A fact in this sense is simply a sentence whose truth is settled somewhere a stranger can reach. The working definition is that a fact is checkable today by somebody who disagrees with the author, and that hostility is the point rather than a complication. A figure that can only be confirmed by asking a person who already agrees has not been checked at all.

Two consequences fall straight out. First, everyone who opens the same filing has all of them, so facts are cheap. Second, facts are not safe from error, only from dispute: a lodged figure can be restated later, and when it is, the record changes and so does the fact. Neither consequence weakens the definition. The two consequences only stop the fact pile from being treated as the valuable half of the work.

What counts as a thesis, and where does it begin?

A thesis is a claim about what happens next, or a claim about a cause that the record does not state. Both halves matter, and the second half is the one people miss. A sentence can be entirely in the past tense and still be a thesis, if what it asserts is a reason rather than an amount.

Take the stall outside a college gate that sold more tea this term than last. The extra cups are countable. Saying the extra cups came from the new hostel opening is a claim about a cause, and it competes with at least two others: the rains were late, or the stall two doors down shut for a month. The takings book settles the first sentence completely and says nothing whatever about the other three.

A thesis is not a weak fact and it is not an opinion in the loose sense; it is the part of the work that carries the risk, and it is the entire reason anybody is paid to research rather than simply to photocopy filings. Nobody hires a reader to tell them revenue was Rs 2,415 crore. The reported figure is free. A defensible view about which of several causes produced a change, and whether the change repeats, is not free at all. The thesis pile is where all of the value and all of the exposure sit, together, so its boundary is worth marking exactly.

Is arithmetic on a published figure a fact or a claim?

Arithmetic on a published figure is the case most often misfiled, and it gets misfiled in both directions. Somebody who has been told to be rigorous starts calling every computed number an estimate. Somebody working at speed starts treating any number with a decimal point as reported. Both are wrong, and the resolution is clean.

The arithmetic itself is reproducible, so the arithmetic itself is a fact. Anybody starting from the same lodged figures and running the same steps lands on the same answer. Work the year three numbers on Sarvani Coatings and watch it happen. Revenue in year three came in 13.9 per cent above year two. Units sold over those same twelve months rose 6.0 per cent. Divide one by the other and realisationRevenue divided by the number of units sold, so the average money collected on one unit. The average moves when prices move and also when the blend of what was sold changes. rose about 7.5 per cent. The rupee spend on materials rose 9.9 per cent over the same twelve months, so on the same 6.0 per cent more units the materials cost per unitThe total spent on materials divided by the units produced, so the average input cost carried by one unit. The per unit cost can rise even as the materials share of revenue falls. rose about 3.6 per cent. Every one of those steps is a division somebody who dislikes the conclusion can repeat.

And the tie closes. Materials were 56.0 per cent of revenue in year two. Moving that by the two per unit rates, 1.03638 divided by 1.07467, lands on 54.0 per cent. The result is exactly the published year three figure. Nothing was assumed anywhere in that chain.

So the arithmetic is a fact. The decision to run that particular arithmetic is not a fact. Choosing to split revenue growth into volume and realisation, rather than into decorative and industrial, or into first half and second half, is a judgement about what is worth looking at. A ratio is therefore a fact presented inside a frame somebody chose, and both halves of that sentence have to survive: refusing the first turns rigour into paralysis, and forgetting the second lets a framing travel disguised as a measurement.

A short algebraic check shows how carefully the frame has to be read. The gap between the two per unit rates, 3.69 per cent, does not depend on the volume figure at all. The volume cancels on both sides, so any growth figure put into both divisions leaves the gap the same. The two individual rates, the 7.5 and the 3.6, do depend on it. So one number in that paragraph would survive a correction to the volume figure and two would not, and nothing on the face of the note says which is which.

Where exactly does the line fall in the margin case?

Now the useful part. On this case the line falls at a single word.

The margin rose, and the rise is in the record. Realisation outran materials cost per unit, and the comparison is arithmetic on the record. Which cause produced that gap is not in the record anywhere, and the moment a sentence reaches for a cause it has crossed over. The line falls exactly at the word because, and everything to the left of it in a sentence can be checked this afternoon while everything to the right of it cannot be checked at all.

Three candidates fit the gap and the lodged statements separate none of them. One: a pricing environment reaching the whole field, in which every maker was able to price ahead of its inputs for a couple of years. Two: the company's own pricing, meaning something it did that its rivals did not. Three: a mix shiftA change in the blend of what was sold rather than in the price of anything. Selling more of a dearer line lifts the average, and no price tag has to move. towards industrial work, which lifts realisation and input intensity together and can widen the gap without any pricing decision at all. All three produce the same published margins. The silence is not a gap in the disclosure. A profit and loss account is a set of totals, not a set of reasons.

The margin case, sorted in full

Here is the whole thing laid out, so the sorting stays visible rather than being asserted. Two different periods are in play, so read the period beside every figure. Mixing them is the fastest way to produce a sentence that is wrong in a way nobody notices.

One row above holds the trap, and it is worth pausing on. Explaining a rising margin by saying inputs got cheaper is tempting. Here they did not. Materials cost per unit rose about 3.6 per cent in year three, so the entire margin gain is that realisation outran input cost, and any sentence describing falling input costs is not an unsettled claim but a plain error. A wrong fact and an unproven thesis are different failures and they get corrected differently: one by re-reading the filing, the other by waiting for evidence that may never arrive.

Now the sharpest case of the lot. Sector evidence exists, and it moves the question a long way without closing it. Over the same two years Nandivarman Paints Limited gained 2.4 points of margin and Kesaria Surface Solutions Limited gained 3.6 points. All three makers rose. Three simultaneous gains are a real narrowing. An effect reaching the whole field explains them comfortably, and one company's private cleverness explains them badly. But Sarvani Coatings sits at 3.0 points, between the two, with 1.2 points of spread across the three, and nothing in that arrangement separates its own contribution from the shared one.

One curiosity, named as a curiosity so it is not read as evidence: 3.0 is the exact midpoint of 2.4 and 3.6. The midpoint is a coincidence of three made up figures and it means nothing. A reader who starts to build an argument on it has just watched the failure happen at first hand, and watching it teaches more than being told about it.

Why do a fact and a claim end up in the same voice?

Because writing well produces flat declarative sentences, and a flat declarative sentence carries no marking for how its truth was arrived at. The margin rose. The margin rose because the company raised prices. The margin will hold at this level. Same length, same rhythm, same confidence, three entirely different kinds of statement.

The overlap is not carelessness and it is not dishonesty. Fluent prose produces it as ordinary output. The check therefore has to be mechanical rather than a matter of trying to be careful. Care fails here for a simple reason: the author already knows which of the sentences are claims, so nothing feels wrong while writing them. The reader has no such knowledge and no way to acquire it from the note. A mechanical pass, sentence by sentence, asking one question of each, catches what care cannot.

How is a sentence rewritten so its status is visible?

Not by hedging. Hedging makes a claim vaguer. Vaguer means less checkable, and that is the wrong direction entirely. The move is the opposite: make the claim more specific, and attach to it the three things that let a stranger settle it later.

Take each sentence and ask who could settle it today. If the honest answer is nobody, the sentence needs a stated condition, a named period, and an owner. Watch it happen on the worst offender in this material. The company has pricing power is four words that assert a cause, name no period, and belong to nobody. Rewritten: in year three, realisation rose about 7.5 per cent while the cost of materials carried by each unit rose about 3.6 per cent, and whether that gap reflects the company's own pricing rather than a shared environment is not settled by the published statements.

A view written this way reads less confident and is worth considerably more. Such a view can be argued with, and a sentence nobody can argue with has taught nobody anything. There is a real cost, and it is worth stating plainly: this style is harder to sell. The four word version sounds like somebody who knows. The long version sounds like somebody hedging. The long version hands the reader the arithmetic and hands back the one judgement the writer could not make for them.

Can a thesis ever become a fact?

Yes, and this is the only route by which any research process gets better. A claim about the coming year sits in the thesis pile for twelve months and then the year is reported, and at that moment it moves. The claim becomes a fact about a past year, either the fact that was claimed or a different one.

There is a condition, and it is strict. The conversion only happens if the claim was written specifically enough for a published year to settle it. Margin will hold at 46.0 per cent through the four quarters of next year converts cleanly: the quarterly filings arrive and the sentence is settled either way. The business has good momentum never converts, in any year, ever. The sentence cannot be wrong, so it can never be checked, and so it can never teach its author anything. Vague claims are permanently safe and permanently worthless, and the specificity that makes a claim risky is the same specificity that lets it eventually pay its author back in learning.

One caution belongs beside this, and it is treated in full separately. When the year closes and the claim turns out to have held, that does not by itself mean the reasoning was sound; a claim can be right for a reason its author never wrote down. Judging the quality of the thinking separately from how the period turned out is called resultingJudging how good a decision was purely by how it turned out, rather than by what was known and reasoned at the time. The term belongs to Annie Duke, Thinking in Bets, 2018. when it goes wrong, and it belongs with the behaviour material rather than here. Only the mechanics belong here: the claim was written, the period ran, the filing arrived, the pile changed.