Written Communication: The Key Message and What Carries It

Finance writing carries one key message and everything else supports it. Plain language means removing the term and keeping the number, never simplifying the claim. The sequence runs message, then support, then detail, so a reader can stop at any point and not be misled. The executive summary is that first layer; the detailed analysis is everything under it.

One uncomfortable fact sits beneath all of that. A reader stops somewhere, and the writer does not get to choose where. Not the reader the writer imagined, not the one who asked for the work, not the one whose name is on the covering note. Somebody puts the document down after the second paragraph and acts on what they have. If stopping early leaves a reader holding something true but incomplete, the writing has worked. If stopping early leaves them holding something wrong, nothing further down repairs it. The reader never arrives there.

One assignment runs through everything below, from top to bottom. The Kavery research desk, inside Kavery Capital Services Private Limited, looked at Meenakshi Tubes Private Limited, an invented borrower it already lends to, and produced a set of documents: 180 words of executive summary, sitting on a 1,400 word memo, sitting on a working file of 21 documents and a workbook of six tabs. The writing was done by Sharada Iyer, the review by Prakash Nadar, and the decision on 12 March by Latha Menon.

The product Meenakshi Tubes Private Limited actually manufactures does not matter anywhere below this line. How the numbers were produced, and what they mean as finance, is a question for financial analysis rather than for writing. A bond, a fund, a warehouse or one machine in its place leaves every rule below intact. The subject here is how writing carries a finding, not what any finding was. Six moves carry the whole craft: stating a key message in one sentence and testing whether it has actually been found, stripping a technical term without losing the number underneath it, ordering a document so that every stopping point is safe, saying exactly where the top of a document stops, telling a narrative apart from a story, and building the one table that carries what prose cannot.

What is the key message, and how is it recognised once found?

The key messageThe single sentence that everything else in a document is there to support. is one sentence stating what the work found, written so that somebody could disagree with it. The requirement that somebody could disagree is the whole test. A sentence nobody could argue with is not a message, it is a topic, and a document built on a topic gives its reader nothing to hold.

A leak in a ceiling makes the point before any memo does. A tenant messages the landlord. The key message is that the leak is coming from the bathroom upstairs and it needs a plumber this week. The message is not that the tenant first noticed a damp patch on Tuesday, that the patch grew, that the bookshelf was moved, and that the landlord's attention would be appreciated. All of that is true and all of it is support. The message is the sentence that, if the landlord read nothing else, would still get the plumber to the door.

Three tests find it, and they are worth running in this order. First, can it be said in one sentence without joining two claims with an and that is doing structural work? Second, does everything else in the document give a reason to believe that sentence? Third, would anybody behave differently if the sentence turned out to be false? A sentence that fails the third test is decoration however carefully it is written.

On the Kavery assignment the key message was this: the operating margin has fallen below the internal floor of 11 per cent, and the concentration in one December order is what could move it further. The sentence is contestable. Somebody can say the fall is seasonal, or that the floor is set wrongly, or that concentration in one order is normal for this borrower. All three of those are arguments, and a sentence somebody can argue with is a message. A message nobody could be imagined arguing against is a message that says nothing.

Why is everything that is not the message support, including the writer's favourite part?

Once the key message exists, every other block in the document has exactly one job. The job is to give the reader a reason to believe the message. The rule is harsher than it sounds. A paragraph being true, well written or hard won counts for nothing. The only question is whether removing the paragraph would make the message easier to doubt. If the answer is no, the paragraph is not support.

Barbara Minto set this out in The Pyramid Principle in 1978: the answer goes first, with everything beneath it arranged as reasons, grouped so that each group answers the question the reader would ask next. The idea is older than most of the people using it and it is still the most reliable structure available for writing that has to persuade somebody who is short of time.

The block that always breaks the rule is the one the writer enjoyed producing. On the Kavery assignment it was a careful reconstruction of how the borrower's order pattern had shifted over three years. The reconstruction was accurate, it took two days, and it gave nobody any additional reason to believe that the margin had fallen below the floor. Interesting and unsupporting is precisely the combination that makes a document longer without making it stronger, and it survives because nobody enjoys cutting their own best work.

Two disposals exist for such a block and neither is deletion into a bin. The block moves down into the layer below, where somebody who challenges the finding will meet it, or it becomes a separate short assignment with its own key message. The block may not sit in the supporting section pretending to be a reason.

What does plain language mean when the subject is genuinely technical?

Plain languageThe same claim with the technical term taken out and the number left in. means removing the term and keeping the number. The definition ends there, and almost everything written about writing gets it wrong in one direction: plain language is treated as a softer version of the claim rather than the same claim without the jargon. Simplifying a claim and simplifying the words used to state it are opposite operations, and only one of them is honest.

Take the sentence the Kavery desk had to communicate. In technical form it reads: the operating margin fell from 12.0 per cent to 10.8 per cent. In plain form it reads: for every Rs 100/- of sales, Rs 10.80 is left after running costs, against Rs 12.00 last year. Nothing quantitative has been lost. Anybody unfamiliar with the phrase operating margin can now act on the sentence, and anybody who knows the phrase perfectly well loses nothing by being handed rupees instead.

Compare that with what usually happens instead. The margin is a bit weaker than last year. Profitability has come under some pressure. Both of those read as plain language and neither is. The term has gone and so has the number. The reader has been handed a direction of travel and no size. A sentence that drops the figure has not been made accessible, it has been made vaguer, and vagueness is a claim about the writer's confidence rather than about the business.

Which four substitutions do most of the work, and which one goes too far?

Four moves account for most of the plain writing this work ever requires, and each one follows the same shape: the thing is named in rupees or in counts, and the denominator stays visible. The table below sets them out, with the fifth row showing the move that looks like the same operation and is not.

Row four is the one that catches careful writers. Variable cost of Rs 11,34,00,000 against total cost of Rs 18,90,00,000 is exactly 60 per cent, and the sentence is true as long as the words of total cost stay attached to it. Drop those three words and the reader hears 60 per cent of revenue. Sixty per cent of revenue would be Rs 12,72,00,000. The real figure against revenue is 53.5 per cent. A percentage without its denominator is not a shortened sentence, it is a different and usually wrong one, and this is the single most common way a plain rewrite introduces an error.

Row five fails for a different reason. Nothing was translated. A measurement was swapped for a judgement, the number vanished, and the reader now holds the writer's temperature rather than the borrower's margin. Verdicts are shorter than evidence, and plain language is always tempted in that direction.

What sequence carries a message, and why does order do so much of the work?

Message, then support, then detail. Three layersOne depth of a document, written so a reader can stop there and still hold something true. in that order, and the reason for the order is that the reader stops where they stop. Minto's ordering puts the answer first inside a single document, and layering is the same principle applied downwards through a stack of documents rather than across the paragraphs of one.

A household handling a hospital bill works the same way. The top of the bill says what is owed and by when. Under that sits the ward and treatment summary. Under that sits an itemised sheet with every dressing and every test. Nobody reads upward from the dressings. Almost everybody stops at the first line, some go to the second, and the third exists so that the first two can be challenged. Every layer has to stand up by itself. For a large share of readers, that layer is simply the last thing they will see.

The Kavery stack ran the same way. The first layer was 180 words. The second was the 1,400 word memo underneath it. The third was the working file of 21 documents and the six-tab workbook. Every number in the two layers above can be rebuilt from scratch inside that file. A reader who stopped at 180 words came away with the answer, two reasons behind it and the single condition capable of changing it. A reader who stopped at the memo held the working, the range and what had been rejected. Neither of them was misled by stopping.

Where does the summary stop, and what is the first layer allowed to contain?

The first layer stops at the exact point where a reader would need to see the working. The boundary is easier to hold than it sounds. Any sentence whose purpose is to show how a number was produced belongs below the line. Any sentence a reader needs in order to act correctly on the finding belongs above it, however inconvenient its length.

On this assignment the compression from 1,400 words to 180 is a factor of 7.8, so call it eight to one. The ratio hides how uneven the compression is. Two whole sections of the memo contributed nothing at all to the layer above. The answer sentence went the other way. Up at the top it has nothing around it to lean on, so it was rewritten longer than it had been in the memo. The top layer is a different selection from the document rather than a shrunken copy of it, and the parts that grow are the parts that must survive being read alone.

The same boundary settles what the first layer may not contain. Not the method. Method is what a reader turns to once they have decided to challenge the finding. Not the history of the work. Not anything unresolved, unless the unresolved thing is the condition itself. Everything else in the memo earns its place by being reachable, not by being present.

What belongs in the detailed analysis, and who actually reads it?

The detailed analysisEverything below the summary layer, where the working is shown rather than asserted. is everything below the summary layer: the working behind each support, the sources each figure came from, the range around the answer, what was considered and rejected, and what would have to change for the conclusion to move. The detailed analysis is not the leftovers. The detailed analysis is what makes the layer above it something other than an opinion.

Who reads it? Almost nobody, most of the time, and this is where new writers make a decision they regret later. If very few people read the detail, why write it carefully? Because the readership is not small and random, it is small and specific. The detailed analysis is written for the one reader who disagrees with the finding, and that reader arrives without warning, sometimes months later, sometimes after the decision has already gone badly.

On the Kavery assignment the detailed layer had to answer three challenges nobody had made yet. Where did the quarterly split come from, and does it reconcile to the filed revenue figure? Which of the numbers on the inputs sheet were filed by somebody and which were chosen by the desk? And what happens to the margin if the December order does not repeat? Each of those has an answer sitting in the layer below, and each of those answers is why the 180 words at the top can be short.

There is a second readership worth naming: the writer, four months later, when somebody asks why the line was reduced. The detailed layer is the only defence available that does not depend on memory, and memory goes first.

Executive Summary vs Detailed Analysis: which questions does each layer answer?

Setting the two layers side by side settles most arguments about what goes where. The split is not about importance and not about length. The split is about which question the layer is answering. The top answers what was found and what should be done about it. The bottom answers why the finding should be believed and what would change it.

What is a narrative in finance writing, and where does it stop being one?

A narrativeAn order of facts that makes them followable, without adding anything to them. is the order that makes a set of facts followable. Nothing more than that. Three facts placed in an order a reader can move through without going backwards is a narrative, and there is nothing suspect about it. Ordering is a service performed for a reader, in exactly the same way that a recipe puts the oven on before the tray goes in.

The Kavery narrative was short. Revenue rose. Margin fell. One order supplied 14 per cent of the revenue that rose. In that sequence a reader can hold all three at once, and the third one lands with weight because the first two set it up. Shuffle them and the same three facts arrive as a list. Order is doing real work here, and none of that work involves adding anything the record does not already carry.

So a narrative is legitimate and useful, and building one on purpose is part of the work. The question worth asking is not whether to have a narrative. The question is where the narrative quietly stops being an ordering and starts being an explanation. The two look identical in the writing, and only one of them has evidence behind it.

Where does a narrative become a story, and which move crosses the line?

One move, and it is always the same one. A storyA narrative that has been given causes the evidence behind it does not carry. is a narrative that has been given a cause the evidence does not carry. The word that does it is usually because, and its relatives are driven by, as a result of, reflecting and on the back of. Each of them converts a sequence into a mechanism, and a mechanism is a claim that needs its own source.

The test takes ten seconds, and it points at the join rather than at the facts: which line in the source record graded that join? If the answer is a line in the record, the sentence is a narrative with a sourced cause in it. A narrative with a sourced cause is the strongest kind of writing there is. If the answer is that the join simply felt obvious, the sentence has turned into a story, and nothing in the writing lets a reader work out which of the two they are holding.

What is a RAID log, and what does it carry that prose cannot?

A RAID logA fixed table of risks, assumptions, issues and dependencies, one row for each item. is one table of risks, assumptions, issues and dependencies (RAID), in four fixed sections. One row per item, the same four sections every week, and a count on the face of the table. On the Kavery assignment the log ran to 18 rows: 4 risks, 9 assumptions, 2 issues and 3 dependencies.

The four categories are not interchangeable and the distinction is worth holding. A risk might happen. An issue already has. An assumption is something chosen to be treated as true so the work could proceed. A dependency is something awaited from somebody else. Prose blurs all four into the same hedging register, and a reader cannot tell from a paragraph whether a thing is feared, occurred, assumed or awaited.

The table shows absence, and prose cannot. An empty section is a reading in its own right: two issues this week means two things have already gone wrong, and a section with nothing under it says nothing has gone wrong yet. A paragraph cannot express that without sounding like reassurance. Deleting the empty section to tidy the table removes the ability to notice the week it fills up.

The nine assumption rows are not written for the log, and they deserve a paragraph of their own. The rows are the nine rows of the assumption register, moved across unchanged. The same nine are the nine numbers on the workbook inputs sheet that the desk chose rather than found, out of 31 numbers in total, the other 22 having been filed by somebody. One set of nine rows appears in three places and is edited in exactly one of them. No other arrangement stops three documents quietly starting to disagree with each other.

What does a lender, an analyst or somebody writing alone do with a document built in layers?

A lender reads the top and nothing else, unless something in the top disagrees with what they already believe. The boundary at the bottom of the first layer therefore matters more than any sentence inside it. A credit officer who reads 180 words and acts is not being careless; they are allocating attention across eleven files and one afternoon, and the writing either survives that or it does not.

An analyst uses the layers in reverse. The analyst reads the top to find out what they are being asked to accept, then goes straight to the layer that would let them refuse. The first thing a good analyst looks for is the condition. The condition tells them where the writer thinks the answer is weakest. The second thing they look for is whether the numbers behind a support can be rebuilt without asking anybody.

Anybody working entirely alone, without a committee and without a reviewer, takes the most from this and has the greatest need of it. Layering is a way of arguing with oneself on paper. Writing a 180 word top layer forces an uncomfortable question: is there actually a finding, or only a collection of interesting observations? If the top layer cannot be written, the work is not finished, and that is a far more useful signal than any amount of rereading. A household deciding whether to repair a ten year old car or replace it can run the same discipline on one sheet: the answer, two reasons, the one thing that would change it, and the quotes filed underneath.

One last practical note that costs nothing. The layers are written in the order they will be read and finalised in the reverse order. The top layer is a selection from a finished argument, and a selection cannot be made from an argument that is still moving. Written communication carries no jurisdiction of its own. The only local element is the record-keeping and conduct duty sitting under any document a registered person issues.

The Kavery research desk, Kavery Capital Services Private Limited, Meenakshi Tubes Private Limited, Sharada Iyer, Prakash Nadar and Latha Menon are invented.
Educational material. Not advice on any investment, tax, budget or market position.