Language tells

Most of the annual report is audited. The words are not. The financial statements pass under an auditor who must, however imperfectly, sign that they are true and fair; the chairman's letter and the management discussion and analysis pass under no such test. They are the one substantial part of the document where management writes freely, in its own voice, about a year it is describing to the people it answers to — and precisely because nobody checks the prose, the prose leaks. It leaks conviction the numbers are dressed to project, doubt the guidance is written to suppress, and the quiet editorial decisions — what to boast about, what to fall silent on, whose fault the bad quarter was — that a set of statements can never show you. illustrative

This module teaches you to read that leak. Not as literary appreciation, and not as the parlour trick of "reading between the lines", which usually means projecting a suspicion you already held onto whatever words are handy. The method here is stricter and duller than that. It has one governing rule, and everything else follows from it: language is a signal only against a baseline, and the only honest baseline is the same company's own prose in earlier years. A superlative, a hedge, a passive sentence, a missing number — each means nothing read once, in isolation. Each means a great deal read as a drift: this year's letter laid beside the letter from three years ago, from the same pen, describing the same business, so that what has changed in the telling stands out against what has stayed the same.

So the object of interest is never the vocabulary itself. It is the movement of the vocabulary over time — the year the hedges thicken, the year "we will" softens into "we hope to", the year bad news slides into the passive voice, the year the metric management was proudest of simply stops being mentioned. This is language drift, and it is to the words what a broken reconciliation is to the numbers: not a verdict, but the precise place to point the next question.

You have spent the earlier parts of this shelf learning to distrust a number until you have tied it to the number that should agree with it. The commentary demands the same discipline, turned on words, and for the same reason: the parts of the report a company controls most freely are the parts where it can most easily manage your impression. The statements are constrained by accounting rules, by the auditor, by the tax authority, by the brute arithmetic that ties them together. The narrative is constrained by almost nothing but the writer's judgement of what they can get away with saying. That freedom is exactly why it is worth reading — the constraint on the numbers is what makes the numbers trustworthy, and the absence of constraint on the words is what makes the words revealing.

There is a second reason, particular to India and to this part of the shelf. Under promoter primacy, the quality of the controlling family is the first filter on an investment, ahead of the business itself — and character is not on the balance sheet. What you have instead are behaviours, disclosed over years: how capital was allocated, whether promises were kept, and how management talks about its own record when the record is mixed. The commentary is one of the few windows onto temperament that a filing offers. A promoter who, in a bad year, names what went wrong and what they got wrong is telling you something durable about how they will treat you as a minority in the next bad year. A promoter who, in the same bad year, produces four pages of ownerless optimism is telling you something too.

The discipline this module adds is to make that reading systematic rather than impressionistic. Left to instinct, a reader is swayed by fluency — smooth, confident prose feels like a healthy business, and a stumbling, defensive letter feels like a sick one, when often the opposite is true, because the sick business is the one that can afford the best writers. The drift method exists to defeat that instinct. It does not ask whether the prose is good. It asks what has changed in how this company writes about itself, holds the change up against the numbers, and treats any gap between a brightening narrative and a dimming set of statements as the thing to explain.

There are seven recurring movements worth watching. None is a verdict alone; each is a place to look. They earn their weight when several move together, in the same direction, in the same year — the linguistic version of a clustered signal.

Hedge density. Count, roughly, the qualifying words: may, could, expect, hope, intend, subject to, in due course, we believe, going forward. A little hedging is honest — the future is uncertain and a candid management says so. What matters is hedge density and its trend. When the qualifiers thicken year over year, the writer's own confidence in what they are describing is falling, and the prose is quietly transferring risk from the statement to the reader.

The tense of a promise. Watch the verbs attached to the future. "We will commission the plant by March and reach 80% utilisation in the first year" is a commitment you can hold them to. "We hope to ramp the plant in due course as conditions permit" is an aspiration that has pre-written its own excuse. The slide from the indicative to the conditional — from will to hope to, from a dated target to "the medium term" — is one of the clearest tells there is, because it is the sound of management widening the goalposts before the shot.

The grammar of blame. This is the sharpest of the seven and deserves the most attention: attribution asymmetry, the mismatch between how good news and bad news are written. Good news arrives in the active voice with a proud human author — "we delivered a record margin", "the team grew volumes 14%". Bad news arrives in the passive voice with no author at all — "margins were impacted", "inventory was built", "the environment remained challenging". Read those bad-news sentences and ask a child's question: who did this? When the grammar refuses to say — when profit is something that happens to the company rather than something the company produced — you are reading a narrative built so management owns every win and disowns every loss.

Superlative inflation. Track the density of promotional words — best-ever, world-class, transformational, robust, marquee, phygital. A stable level tells you the house style. A rising level — superlative inflation — often runs inversely to the substance beneath it: the year the results are thinnest is frequently the year the adjectives are thickest, because the words are recruited to do work the numbers cannot. The tell is not the superlative; it is the superlative appearing where last year there was a figure.

Blaming the weather. Every business faces genuine external forces — input costs, interest rates, monsoons, a demand slump — and honest commentary names them. The tell is one-sidedness: when bad years are always the world's fault and good years are always management's skill, you are watching a system for evading accountability, not describing it. The honest version sounds different — it credits the tailwind in a good year ("demand and pricing were both favourable") and owns the misstep in a bad one ("we mis-read the inventory cycle and over-produced").

The disappearing metric. The quietest and often the most damning: a disappearing metric. A number management boasted about for years — an order book, a market share, a return ratio, a subsidiary's growth — simply stops appearing in the narrative the moment it turns unfavourable. Nothing is retracted; the number is not discussed as having fallen; it is just no longer there. The silence is the signal, and you catch it only by remembering what last year's letter chose to feature.

Jargon as fog. When plain facts are available, plain language is used. Jargon and abstraction thicken exactly where a concrete number would embarrass — "we are driving synergies across the value chain to unlock stakeholder value" is what "we cannot show you a margin improvement" sounds like when dressed for the page. Fog is not neutral; it is chosen, and it is chosen to cover.

Here is the discipline that separates this from opinion. A single year's commentary is nearly unreadable as a signal, because you have no baseline for this writer. Every management has a house style — some are congenitally promotional, some austere to the point of terseness — and judging one letter in isolation measures the style, not the change. The signal lives in the difference between years, from the same pen.

So the practice is mechanical. Pull the chairman's letter and the MD&A for this year and for three or four years ago. Read the old one first, deliberately, before the new one, so the old voice is fresh and the new prose has to sound itself against it. Then ask a short, fixed set of questions. Are there more hedges now, or fewer? Have firm, dated commitments become conditional and undated? Is bad news written in the same voice as good news, or has it retreated into the passive? Which metrics did the old letter lead with that the new one has dropped — and did those metrics turn unfavourable? Has the adjective count risen while the disclosure has thinned? Each answer is a small vector; the question is whether they point the same way.

Read any one row and you have a shrug. Read all seven and you have a company whose narrative has brightened in tone while, almost certainly, its numbers have dimmed — because that is the only condition under which a management systematically reaches for more hope, more passive voice, more adjective and less metric. That divergence, narrative up while substance down, is the exact gap the forensic reader lives for. It does not tell you the business is failing. It tells you where the next hour of work goes: into the statements the words are steering you away from.

Now the inversion this shelf turns on, applied to language. The temptation, once you have the seven tells, is to build a blacklist — treat world-class and transformational as red flags wherever they appear. That is exactly wrong, and wrong in the characteristic way this book keeps warning against: it reads a signal without its context. Promotional, superlative-heavy language is the native register of a consumer-brand story and tells you almost nothing there — a company that sells to millions with words and images writes to its investors the same way it writes to its customers, because that is the only voice it has. The same register in a business whose culture is to under-promise is an event, because it marks a departure from a deep convention. The words are identical; the reading inverts on the baseline register of the sector.

The rule underneath all four cells is one you have met before in this book, wearing a different costume each time: a signal is only legible against the right baseline, and the baseline moves with the business. For a metric it was the sector's economics; for language it is the sector's register. A reader who carries a fixed dictionary of "bad words" will condemn the consumer brand for writing the only way it knows how, and absolve the industrial in the very year its plain voice cracks into promotion. Carry the method, not the word list: find how this company, in this kind of business, normally writes — then read what has changed.

Take a composite mid-cap industrial — call it Meridian Industrial — that makes process equipment for other manufacturers. Its investor letters, for years, were plain to the point of dryness: capacity in tonnes, order book in rupees, utilisation in per cent, a dated target for the new line, and a flat sentence when something went wrong. This is the register of its sector, and it was the register of its own past. illustrative

This year's letter reads differently, and the difference is the whole point. The order book — quoted proudly on the first page for four straight years — is not mentioned at all; it is not said to have fallen, it has simply left the narrative. The plant that "will commission in Q2 and reach 75% utilisation in year one" three years ago is now expected to "ramp over the medium term as end-market demand recovers". The one weak segment is described in a sentence with no subject: "realisations were impacted by a challenging pricing environment." And scattered through it, where there used to be numbers, are the words the old letters never contained — "transformational partnerships", "a world-class platform", "unlocking value across the ecosystem." Six of the seven tells have moved, all in the same direction, in a single year, in a company whose baseline is the opposite of promotional.

None of this proves anything about Meridian's business, and the forensic reader does not pretend it does. What it does is locate the questions with precision. Where is the order book, and what did it do? — go to the notes and the earlier disclosures and reconstruct the number the narrative dropped. Why did a dated utilisation target become an undated hope? — check the capital-work-in-progress and the interest being capitalised on that plant. Who built the inventory and cut the realisations? — put the passive sentence back into the active voice and see whether the numbers support "we misjudged demand" or something management would rather not author. The language did not give you an answer. It gave you the four places in the statements where the answer is hiding, and it did so before you had read a single figure — which is exactly what a good tell is for.

The habit to build is to read the words first, in this drift-against-baseline way, and let them direct the reconciliation — rather than reading the numbers first and using the words as reassuring background music. The commentary is written to be read last and skimmed; read early and read comparatively, it becomes a map of where management would prefer you not to look.

Language is corroboration, never proof, and the ways it fails are worth as much attention as the ways it works.

It can be bought. The most dangerous letters are not the clumsy defensive ones but the ones written by a skilled investor-relations team or an outside agency, engineered to hit every note of candour and confidence without a word of it being the promoter's own. Polished prose that says all the right things — owns a setback here, names a real external factor there — can be pure craft, and a reader who scores tone alone will rate the ghostwritten letter above an honest, awkward one. The defence is the same as everywhere on this shelf: the words only earn trust when the numbers they describe corroborate them. A letter that owns a hard year is worth something only if the accounts show the hard year being genuinely dealt with.

It cuts the wrong way as often as the right one. A blunt, ungracious, badly written management can run an excellent business, and a fluent, warm, self-aware one can run a failing or dishonest one. Fluency is a fact about the writer, not the company. Some of the plainest, least quotable letters in the market belong to superb capital allocators who simply do not care to perform, and some of the most moving annual-report prose has prefaced a collapse. Reading tone as competence is a category error, and it is the error the confident-sounding fraud relies on.

And it does not survive being read once, or read as a checklist. A single hedge, a single passive sentence, a single superlative is nothing — the method depends entirely on the drift against the company's own baseline, and a reader who has not read the earlier letters has no baseline and is merely projecting. Worse, the tells become known: managements learn what analysts flag and coach the language to avoid it, so a letter can be scrubbed of every obvious tell and still describe a hollow year. The commentary is a lead, not a verdict. It tells you where to point the reconciliation; the reconciliation, and only the reconciliation, tells you what is there.

The first way people get fooled is by reading fluency as health. Smooth, confident, well-organised prose produces a feeling of a well-run company, and the feeling arrives before any number is checked. But the sick business is precisely the one with the budget and the motive to hire the best writers, and the honest small firm often writes plainly and badly. Grading the prose instead of its drift against the numbers is how a reader talks themselves into the most polished disasters on the exchange.

The second is carrying a fixed word list across sectors. A reader who has learned that superlatives are a "red flag" will flag a consumer brand for its native voice and wave through an industrial in the year its plain register cracks into promotion. This is the inversion failure in miniature: a signal read without its baseline is not a signal at all. The register that means nothing in one business is the loudest tell in another, and only the sector's convention — and the company's own history — tells you which you are looking at.

The third is reading a single report instead of the drift. Almost everything in this module is invisible in one year's letter, because one letter measures the house style and nothing else. The disappearing metric can only be noticed by someone who remembers last year's featured metric; the softened promise only by someone who read the firm one; the thickening hedge only against a thinner past. A reader who opens this year's annual report cold, with no earlier letter beside it, will see confident prose and a company describing a year — and will miss every signal the method is built to catch. The tells are differences, and a difference needs two points to exist.