The Variant View: Disagreeing With Consensus, With Reasons

A variant view is a disagreement about how a business works, held against a reading somebody actually wrote down, with the evidence that would settle it named alongside. A variant view is not a disagreement about a price, and it ends at a statement about the business. Exactly one such written reading is held in these notes, and this guide argues with it on three published grounds and then stops.

A figure that is correct, published and reproducible. So what is left to argue with?

The idea is slippery and the object is not, so the object comes first. In a reading already built in these notes, an analyst quotes Anjani Stationers Private Limited's degree of operating leverageA multiple that says how hard the profit line reacts to a change in sales. It is worked out as contribution divided by the operating result at one stated level of sales. as 2.78313. Five decimal places. Anjani Stationers, an invented maker of paper registers, has its second year published in full, so the division can be done in one line: contributionThe slice of a sale left over once the costs that sale itself created have been met. Whatever remains goes towards the bills that turn up regardless of sales. of Rs 1,15,50,000/- over an operating result of Rs 41,50,000/- gives 2.78313, and the reading that published it says in terms that the division really does give that figure.

One separation comes before anything else. A reader who reaches the end still confusing the two has misread everything in between. The object built here is a disagreement about how a business works and what the evidence supports, and never a disagreement about what a share is worth or what anybody should do about it. A view held against a market price is a different object with different rules, settled by nothing anybody could go and fetch, and it is covered separately under Variant Perception: Where a View Differs From Consensus. Everything below stops well short of that line, and the last block in this guide is about exactly where the stopping happens.

So: the arithmetic is right, the inputs are published, the writer could reproduce it tomorrow to the last decimal. Three things are left, every one of them itself printed in the same built reading, and not one of them an arithmetic error. The reading rests on an input that was somebody's judgement rather than anything a company presented. The reading is also one number written twice and then read as two agreeing findings. And it describes the position the reader happens to be standing at rather than a property the business carries around with it.

One clause travels with this figure everywhere it goes in this guide, including inside the drawings, inside the panel and inside the answers to the questions. Whoever sorted the cost base behind both the contribution and the operating result was making an estimate, not reading a disclosure. Somebody decided which part of the cost base moves with the number of registers made and which part sits there regardless. No filing anywhere carries that decision. The sorting is a judgement, and every figure downstream of it inherits the judgement without ever saying so.

Here is the same shape at household size. Two people sit down with the same electricity bill. The two agree on every printed figure: the units, the rate, the total, the dates. Then one of them says next month will cost about the same, and the other says it will not. Nothing they disagree about is on the bill. One of them is treating this month's rate as a property of the flat, and the other is treating it as a reading taken at one moment with the fan running. The disagreement is real, it is about the world rather than about the arithmetic, and it can be settled by somebody going and looking at the meter again.

So what is a variant view, and what are its three parts?

Most readers arrive expecting a clever contrary opinion: everybody thinks one thing, the writer thinks another, and the skill is in the thinking. A variant view is not that object. The thinking is the cheap part, and it is cheap because anybody can produce a contrary sentence about anything in about four seconds. The expensive part is saying precisely which reading is being disagreed with, so a second person can check whether the two readings differ at all.

A variant view carries three parts, and all three are written down. The first is the reading being differed from, with a name on it, a place it was written and a date. The second is the writer's own reading of the same fact, stated in the same units. Stated that way the two sit beside each other without anybody having to translate. The third is the evidence that would settle which of the two is right, named, together with a plain statement of whether anybody currently holds it.

Now the part people skip, and it is the first one. A disagreement whose other side is unnamed cannot be checked, cannot be corrected and cannot be settled. The first line therefore decides whether the other two are worth writing at all. Lines two and three filled in beautifully with line one blank produce a well-dressed opinion. Line one filled in honestly turns even a rough second line into an argument. Somebody else can now go and look at the same thing.

There is a practical version of this test and it takes about ten seconds. The question is whether a second reader could be pointed at the reading being argued with. Not described. Pointed at. If not, what is being carried is a view. A view is a perfectly respectable thing to carry, and plenty of good work starts as one. A view is simply not the object taught here, and confusing the two is how a note ends up sounding certain about nothing in particular.

The shopkeeper version: a customer tells a shopkeeper his rice is dear. He asks, quite reasonably, dear against what. And the honest answer turns out to be a figure half remembered from somewhere, from somebody, about a different sack, some months ago. The moment line one has to be spoken out loud, most disagreements turn out to be about a number nobody can produce.

These notes hold nobody's expectation of anybody. So what does a disagreement attach to?

The softened version is worse than the blunt one, so say it plainly. Nothing in these notes publishes what anybody expected of any business in them. There is no aggregated reading of Anjani Stationers, no range of readings, no count of people who have written one, and no gap between what somebody looked for and what arrived. The absence is not an oversight and not a gap waiting to be filled in a later edition. The absence is a fact about invented businesses. Nobody outside these notes has ever looked at this one. There is nothing outside these notes to look at.

The tempting move at this point is obvious. A disagreement needs something to be variant from, and supplying one would take about a sentence. A trade figure, a typical reading, what people generally assume about a business of this shape, and the argument suddenly works. Supplying one would also manufacture the single input that decides the whole answer, and manufacturing that input is the one move an argument from evidence must never make.

So do the honest thing instead. Write down what would have to exist before anybody could print an aggregated reading at all. Five things, and each one is a line rather than a paragraph.

One, more than one person publishing a reading of the same business. An aggregate of one reading is just the reading. Two, each reading stated in the same units. Two figures measuring different things cannot be combined however carefully they are averaged. Three, a date on each. A reading is a statement made at a moment, and a set of readings taken years apart is a history rather than an expectation. Four, a named holder for each. A reader who wants to know what a reading assumed can then go back and ask. Five, some record of how the collection was made. A set gathered by asking whoever happened to answer is a different object entirely from one gathered by asking everybody who writes.

Not one of the five exists here, and the honest report is the absence itself. Written out that way it stops being an embarrassment and starts being information. A reader now knows exactly what is missing, exactly what would fill it, and exactly what would have had to be invented to pretend otherwise.

Here is the sentence the whole method turns on. Published elsewhere in these notes, it is worth repeating because it is doing real work. Write not established on a line and anybody who later turns the fact up can complete it. Write a plausible figure on that same line and no future reader will ever undo it. A year later not one person remembers that somebody simply supplied it. The asymmetry is the entire argument, and it has nothing to do with modesty: the honest blank is repairable and the plausible filler is permanent damage.

The household version is uncomfortable and exact. Somebody asks a householder what rent the flat would fetch. The householder does not know, and says a figure that sounds about right. Six months later that figure is in three conversations, one spreadsheet and a household budget, and not one of the people carrying it around knows it began as a guess in a doorway. The figure did not get more true by travelling. The figure only got harder to challenge. By then it had been passed along by people who never stood in that doorway.

How an aggregated reading is put together, revised, and where it stops seeing things, is covered separately under Consensus: How the Aggregated Expectation Is Built, Revised and Where It Is Blind. Aggregation is a real subject with real machinery behind it. In these notes it simply has no material to work on.

What do these notes hold instead, and is one reading enough to argue with?

The permitted object is smaller than an aggregate and it is enough. Call it a stated reading, meaning a reading somebody actually wrote down, in a place a second person can be sent to. A stated reading has to be nothing more than that. Such a reading does not have to be popular, it does not have to be representative, and it does not have to have been counted alongside anybody else's.

One such reading sits in these notes. An analyst quotes Anjani Stationers' degree of operating leverage as 2.78313, in a reading already built, worked from that business's published second year. Both halves of the description belong in the same breath. The second half is where careless notes go wrong. The quoted reading is not an aggregate of anything. Nobody counted how many people held it. Nobody says who wrote it, where they worked, or on what day. One reading, written down, is enough to disagree with and nowhere near enough to call anybody's expectation.

The distinction is worth holding onto because it inverts what most people assume. The instinct is that a disagreement gets stronger the larger the thing being disagreed with: arguing with everybody sounds braver than arguing with one written reading. In practice it runs the other way. A crowd cannot be quoted, cannot be dated and cannot be gone back to, so a disagreement needs a named other side rather than a crowd. An argument with one written sentence is checkable at every step. An argument with what people generally think has its weakest line first.

There is one more thing worth saying about where this particular reading sits, and it changes the tone of everything below. The sentence quoting 2.78313 was written in that earlier reading as a question to be argued with. The sentence was put there to be picked at. So none of what follows is a discovery, and nobody is being caught out. The work here is to take an argument opened there in one line and run it properly, at length, on published figures. Running an argument out is a different job from finding a mistake.

Ground one: what does it mean to say the reading rests on an estimate?

Across the two published years the multiple moved from 1.94 to 2.78. The obvious story is that the business got bigger and bigger businesses amplify more. The obvious story is wrong. The attribution is worked out properly rather than loosely in these notes, and that is what makes this a ground rather than a quibble.

Hold the standing costThe bills a business carries whether or not it sells anything in a given period, such as rent, salaried people, insurance and the wear on machines already bought. base at the first year's Rs 49,60,000/- and put the second year's revenue and contribution through it. The multiple comes out at 1.75. Growth in revenue by itself was therefore dragging the reading down, 1.94 to 1.75, and the Rs 24,40,000/- added to the standing base then carried it the entire way, 1.75 up to 2.78. The standing bill did more than the entire work, and every rupee of that standing bill sits on one side of a split that nobody presented.

The Rs 24,40,000/- rise opens up into parts, some read off something and one put there by somebody. There are three, and they are worth setting out as rows rather than as a sentence.

Here is the reading that turns arithmetic into a ground. Two of those three rows are movements in heads sitting on the face of the statementSet out as its own line in the main body of a financial statement rather than tucked into a note behind it or assembled afterwards by whoever is reading., and together they come to Rs 13,00,000/-, or 53.28 per cent of the rise. No company is asked to sort its costs by whether they answer to volume, so the middle row is printed by nobody. So 46.72 per cent of a total that closes to the rupee was put there by somebody rather than disclosed by anybody, and the closing to the rupee is exactly what conceals that.

Telling a recorded fact apart from something worked out from it, and both of those apart from a picture of what might follow, is settled at length elsewhere and is not reopened here: it is covered separately under How to Separate Facts, Inference and Scenarios in Company Research. The addition here is narrower: what happens to a disagreement when a recorded fact and something worked out from it are added together and the sum closes exactly.

The conclusion runs against instinct. A total that closes to the rupee feels checked. Feeling checked is not being checked. A sum closing exactly says that the three figures were chosen so that they would close, and says nothing whatever about where each of them came from. A split that was constructed out of a level cannot then confirm the level it was constructed out of. The split can still say which term a movement came out of, a genuine claim and a far weaker one, and whichever of the two is being made gets said out loud or neither gets said.

Stated in the form a disagreement actually uses, ground one is one sentence. Change the sorting and the multiple changes with it, so the quoted reading reads an assumption at least as hard as it reads the business, and that sorting was an estimate, not a disclosure. Ground one is not an accusation. Ground one is a description of the material the figure is made from.

Household version. Ask any household whether its rent is a cost that moves with how much it uses the flat. Obviously not, they will say. Electricity? Obviously yes. Now ask about the annual maintenance charge on the building, or the school fee for a child who is going anyway. The two easy cases are easy and the interesting cases are all judgements, and judgements are exactly what nobody is asked to publish.

Ground two: how does one number get quoted twice and counted as two findings?

This is the tidiest of the three grounds and the easiest to walk straight past. The multiple is contribution over the operating result. There is a companion quantity called the break-even cushion: the share of revenue that could be lost before the operating result reaches nothing. The cushion works out as the operating result over contribution. The same two quantities, one on top of the other and then the other way up.

The second year's pair is published in the same reading: 2.78313 against 0.359307. Multiplied together the answer is one. Not close to one, not one when rounded, but one, at any revenue whatever and any standing bill whatever. Each of the two is the other one inverted. The first year's pair does the same thing: 1.93585 against 0.516569, and their product is one as well. Dividing a number into one does not make a second number, so a reader who reports the rise in the multiple and the fall in the cushion as two agreeing observations found one thing rather than two.

The doubling belongs in a treatment of disagreement rather than one of arithmetic because of the effect the error has on the reader who makes it. The cost of counting one finding twice is not that the note is wrong; it is that the note is more certain than the evidence in it. A single finding invites the obvious next question: what else might explain it. Three findings that agree close that question down. So an error that adds nothing to the evidence subtracts something from the doubt, and doubt was the part doing the work.

Nothing about either figure is incorrect. Both are right, both are reproducible, and the product being one is a property of how the two are defined rather than anything about the business. Being a property of the definitions is what makes this a ground rather than a correction: the analyst is not being told the sum is wrong, but that two of the three supporting observations are the same observation.

The household case takes ten seconds and lands hard. A household gets a pay rise and, in the same month, a smaller deduction at source. Two pieces of good news, everybody agrees. Except the smaller deduction is arithmetically a consequence of how the pay is structured, so the second piece of news is the first one seen from the other side. Counting both gives that household a picture of its year that is twice as bright as the evidence for it.

Ground three: why does the same measure read 2.39 one row along?

The multiple is contribution over the operating result at one stated level of revenue. So the moment revenue moves, the multiple moves too, and it moves because the operating result underneath the fraction moved further in proportion than the contribution above it did. The recompute is printed rather than asserted. At the row one step up, revenue of Rs 2,97,00,000/-, contribution of Rs 1,27,05,000/- and an operating result of Rs 53,05,000/- give a multiple of 2.39, not 2.78. The cushion at that same row reads 41.76 per cent, being Rs 53,05,000/- of operating result divided by Rs 1,27,05,000/- of contribution, and it is worth saying which division produced it because a figure of that shape turns up elsewhere in these notes out of a completely different one.

The reading is about the very next small step away from the position it was taken at, and it is not something the business carries about with it. The earlier reading says as much in its own words, and that is the whole of ground three. Nothing about the business changed between those two rows. No machine moved, no customer arrived, no cost was renegotiated. Only the position from which the reading was taken.

One further part turns an observation into a ground. Writing a figure to five decimal places tells a reader it was measured. The quoted multiple was not measured. Elsewhere in these notes the fault is named plainly as false precisionReporting an answer with more decimals than the figures feeding it can bear, so it looks measured when what stands behind it was somebody's estimate.: five decimals of apparent measurement standing on a single judgement. The repair is not to trim the decimals. Trimming them would only tidy the problem out of sight. The repair is to print the assumption alongside the answer.

One more limit, worth noting without labouring. A standing bill only stays flat over a band. Beyond the band it moves in jumps rather than slopes. A shed fills and a shift is added. Anjani Stationers took a second warehouse during its second year and its monthly standing cost jumped by Rs 50,000/- the moment it did. A figure worked out on one tread of a staircase says nothing whatever about the tread above it, and nothing published anywhere states the volume at which a tread gives out.

The everyday case is on the dashboard of any bus. The speed shown is exactly right about this second. The needle gives a true, precise, instantaneous reading. The same needle says nothing whatever about how long the journey will take, and anybody who multiplies it by the hours remaining has confused a reading taken at a point with a property of the trip.

Where does this stop, and what is the last sentence it is allowed to write?

The finished thing written out lets its shape do the arguing. A completed disagreement of the kind this guide teaches carries four lines, and it is short enough to copy onto the back of a sheet of notes.

Line one, the reading differed from: the degree of operating leverage is 2.78313, quoted in a reading built earlier in these notes, from the published second year, with no name, no house and no date attached to it by whoever wrote it. Line two, the different reading: that figure is a reading of an estimated sorting of costs at least as much as a reading of the business, taken at one position, and two of the observations usually quoted alongside it are the same observation. Line three, the grounds, one each: it rests on an estimate; it is one number written twice; it describes a position rather than a property. Line four, the evidence that would settle it: a statement of which costs answer to volume and which do not. No filing carries such a statement, only the business itself could give it, and nobody currently has it.

Then stop. A statement about the business is the last sentence any fact could settle, so the view ends there. The stopping is not modesty and not caution dressed up as principle. The difference lies in the two kinds of sentence. Line four names a disclosure that, if it arrived tomorrow, would close the argument for everybody. No disclosure anybody could publish would close a sentence about what the business is worth, so no equivalent fourth line exists for it. A sentence with no line four is not a weaker version of this object. Such a sentence is a different object, and it belongs to a different question.

Show it to yourself with two sentences about the same shop. The first: this shop's standing costs went up more than its takings did. The shopkeeper's own books could confirm or contradict that tomorrow morning, and once they had, the argument would be over. The second: this shop is a bargain. No book that shopkeeper will ever keep can settle the second sentence, and every book in the world could be opened without moving it one inch.

So: a view expressed against a market price is covered separately under Variant Perception: Where a View Differs From Consensus. A stated figure for what a share is worth, and the practice of issuing one, is covered separately again. Both are real subjects. Neither is this one.

What five lines travel with any disagreement worth writing down?

How does a figure that reconciles perfectly become the hardest kind to challenge?