Market Fragmentation: Share in a Crowded Market

A crowded field does not make a share small. The crowding makes a share unknowable. Setu Bazaar, an invented marketplace, carries 2,000 merchants and Rs 4,80,00,00,000/- reaching them in the year, so an average merchant accounts for 0.05 per cent of that flow. Which of the 2,000 names carries how much of that flow is published nowhere, so 0.05 per cent describes a place in the field and no merchant in it.

What does a crowded field actually do to a share?

The answer most readers arrive carrying is reasonable, and it is wrong. Two thousand sellers stand in a field. Surely each of them is small. Surely the arithmetic writes itself: divide the whole by two thousand, and there is the share, tiny and unremarkable, and the crowding has done its work.

Small is not the finding. A crowded field does not make a share small; it makes a share unknowable, and the second problem is much harder to live with than the first. Small is a claim. Claims can be checked, argued with, revised. Unknowable is a state of the evidence, and no amount of care with a calculator repairs it. Believing every seller in a crowded field is small is a false comfort. The field could just as easily be a handful of large names beside a very long list of small ones, and the count on its own would look exactly the same from outside.

A nearby subject carries the same word in its title, so one separation comes first. How many sellers a field carries, and how that field divides between them, are two separate facts, and both are covered under Consolidation and Fragmentation: How an Industry Concentrates. A different question is what a share is worth once the field is crowded, and the answer sits underneath the line rather than above it. Concentration counts heads. A share divides a total.

The everyday version of this is worth holding on to throughout: a wholesale market with two hundred stalls under one roof. Anybody can count the stalls in ten minutes; the count is a fact lying in plain sight and nobody disputes it. Which stall is the biggest is another matter. No stallholder shows another one their book, so walking the aisles and asking produces two hundred confident and contradictory answers. The count was free. The split is not for sale at any price.

Where does the only population of sellers in these notes come from?

Exactly one published population of sellers has a money figure attached to it. Setu Bazaar carries 2,000 merchants. In its published trading year, goods worth Rs 5,00,00,00,000/- were ordered across it, and of that sum Rs 4,80,00,00,000/- went out to the sellers. The gap between the two figures, Rs 20,00,00,000/-, is the revenue the marketplace kept for itself.

Now do the division that the figures invite. Rs 4,80,00,00,000/- divided by 2,000 gives Rs 24,00,000/- of goods for each merchant across the year. Multiply it back to be sure rather than to be satisfied: Rs 24,00,000/- multiplied by 2,000 returns Rs 4,80,00,00,000/- exactly, so the division is sound and nothing has been rounded away in it.

Then take the same division as a percentage instead of as rupees. One merchant out of two thousand, evenly, is one part in two thousand. One hundred divided by two thousand is 0.05 per cent. The whole derivation is that one division, and a one-line derivation is exactly why the figure is so easy to write down. There is no survey behind it, no estimate inside it and no judgement anywhere in it. Two published figures went in and one number came out, and the whole operation takes about four seconds.

One label travels with every appearance of Rs 24,00,000/-, and it goes in the same sentence rather than in a footnote. The label is gross flowMoney measured as it moves past, with nothing yet deducted. Whether any of it stays with the party it moved past is a separate question, and usually an unanswered one., meaning goods that left. Between that figure and anything a merchant keeps sit the cost of those goods, the cost of shifting them, and the charge for trading there in the first place. Three deductions, no published size for any of the three, and nothing in these notes carries a turnoverWhat a business billed over a stretch of time, before a single cost of doing the billing has been taken off. A turnover figure opens an account rather than closing one. figure across into an income one.

Why is 0.05 per cent an average rather than a share?

The difference between an average and a share is the centre of the subject. A share belongs to a name. A share says what this merchant, this one here with a signboard and a bank account, holds of something. An average belongs to a place. An average says what a merchant would hold if the flow divided evenly. An even split is a statement about the field and about the arithmetic, and never about a person in it.

The reason the two cannot be swapped here is one published fact: nothing anywhere breaks the Rs 4,80,00,00,000/- down name by name. Not partially. Not in bands. Not for the largest ten. Nowhere at all. So not a single merchant's share is known, and 0.05 per cent describes a place in that field rather than a name standing in it.

Careful readers go wrong at exactly this point, and the trap closes here. The figure is not an estimate of a merchant's share. The figure is not a midpoint that the true shares scatter around, and not a typical value, a central case, or a starting point to be adjusted once more is known. An average is not a share with error bars on it; it is a different quantity that happens to share a unit. Error bars would imply the average is aiming at something, and it is not aiming at anything. The average is the answer to a division nobody asked about any merchant.

The everyday version again. Two hundred stalls under one roof, and the market's landlord announces what value of goods left the building this year. The landlord's figure divided by two hundred gives what an average stall shifted. Point at any one stall, though, and nothing about it is known. The landlord's figure was never about that stall; it was about the building. Somebody who reads the average off the notice board and repeats it as what the stall on the corner sells has not made a small error of degree. The reader has swapped one quantity for a different one and kept the unit. A swap that keeps the unit is the swap nobody notices.

What does that average leave free underneath it?

Saying that an average constrains nothing about the largest name is easy, and most readers will nod at it and then go straight back to reading 0.05 per cent as a small merchant. The claim needs working rather than asserting, and the working carries a label. The settings worked below are arithmetic demonstrating a property of an average. The settings have no name, no trade, no country and no year, and none of them describes how the merchants on Setu Bazaar are split.

Hold two things fixed: 2,000 merchants, and Rs 4,80,00,00,000/- reaching them. Both are published. Now take the top two hundred names, one tenth of the count, and put a stated portion of the flow into that tenth. At an even split, the top two hundred carry 10.00 per cent of the flow, exactly their share of the heads, and every merchant on the marketplace averages Rs 24,00,000/- and 0.050 per cent. Nothing is happening yet, and a setting at which nothing happens is the right place to start.

Move the portion in that top tenth to 40.00 per cent. Each of the two hundred now carries Rs 96,00,000/- of goods and 0.200 per cent of the flow. Each of the other 1,800 carries Rs 16,00,000/- and 0.033 per cent. Move it to 60.00 per cent and the readings are Rs 1,44,00,000/- and 0.300 per cent against Rs 10,66,666.67/- and 0.022 per cent. The tallest merchant on screen has grown sixfold and the smallest has shrunk to under half.

Here is what did not happen at any of those settings. The total being divided never moved and the count dividing it never moved either, so the average across all 2,000 stayed at Rs 24,00,000/- and 0.05 per cent, to the last decimal. The average is fixed and the arrangement behind it is entirely free, and nothing published anywhere rules any of those settings in or out. The silence is not a weakness in this particular average. The silence is what a weighted averageA single figure got by dividing a total by the count of things that made it. Larger items pull it further than smaller ones. A weighted average reports the total and the count, and it reports nothing about how the total was shared out. is: a report on a total and a count, and a silence about everything in between.

A share of what, exactly?

Everything so far has been about the number on top of the line. The trouble does not stop when the numerator arrives, and the number underneath the line carries a trouble of its own.

Suppose the difficulty vanished. Suppose one merchant's own selling on Setu Bazaar were published to the rupee, so the top of the ratio was solid, checkable and beyond argument. Dividing it by Rs 4,80,00,00,000/- yields a share of what reaches one marketplace's sellers. A share of one marketplace's flow is a real quantity, correctly computed, and it answers a question nobody asked. A merchant selling on Setu Bazaar also sells elsewhere, or could, and nothing anywhere says how much. The shop may have a street frontage, may supply two other marketplaces, or may sell nothing anywhere else at all. Not one of those three is published, in either direction, for any of the 2,000 names.

So the thing underneath the line is one route to buyers rather than the merchant's market, and the answer is a share of one channel. Note what has gone wrong and where. The numerator was assumed perfect. The division is arithmetic. The trouble is that the only figure within reach was the wrong one, and it was within reach precisely because a marketplace publishes its own year while a market does not publish anything at all.

Then a second difficulty arrives on top of the first, and this one has nothing to do with channels. The same marketplace in the same year carries three published totals, all correct, and each of them is drawn on a stated convention. Rs 4,80,00,00,000/- reaches sellers. Rs 5,00,00,00,000/- is the sum the buyers actually handed over, the marketplace's gross merchandise valueEverything bought across a marketplace over a stretch of time, added up at what the buyers handed over. Gross merchandise value sizes the traffic and says nothing whatever about the toll.. Drawn the other way round, with each seller's whole sale entered as that seller's own revenue, the chain adds to Rs 5,20,00,00,000/-. Take one merchant's Rs 24,00,000/- against each of the three in turn. Against what reaches sellers it is 0.050 per cent. Against what buyers paid it is 0.048 per cent. Against the chain drawn the other way it is 0.046 per cent, that last one rounded from a shade under 0.0462. Three correct answers, three stated conventions, one merchant, one year, and not one of the three is a market share.

One inherited ruling closes the block. A rival is defined by what it sells; a substituteSomething that removes the need for the order rather than competing for it. A substitute is defined by what the buyer was trying to achieve, so it reaches the same buyer from a completely different place. is defined by what the buyer was trying to achieve, and two circles drawn around the same buyer leave the second always the larger of the two. The larger circle is the whole of why any wider figure underneath this line would be larger still, and the whole of why nobody has one.

What is every share in these notes a share of?

The claim is countable, so here it is with the count. Thirteen shares are published across every business, every chain and every household in these notes. Thirteen shares, thirteen figures sitting underneath the line, and each of those figures is named rather than assumed. Read the last column of the register below and the argument makes itself.

Thirteen published shares, thirteen named figures underneath them, and not one of those figures is a market. And the split inside the thirteen turns the observation into a pattern rather than an accident. Sort them by one test: is the figure underneath something a single named business publishes about itself? Eight of the thirteen answer yes, being the two customer shares, the two sets of spend shares, the take rate, the heavy buyer band, the flow reaching sellers, and the utilisation figure. The remaining five all span several parties at once: a chain drawn on a convention, that chain's pool, a two stage pool, and the two readings of one household's wedding.

Then the sentence that turns the register into a finding. Every one of these thirteen has a figure underneath it that somebody actually holds and can produce on request. Not one of them is a market, and that is precisely why every one of them is knowable. A business holds its own revenue. A household holds its own bills. A chain drawn on a stated convention holds every invoice inside the convention. Nobody holds a market, so nobody can produce one. The thirteen shares in this register are therefore the thirteen that could be written down at all.

One figure in that register appears twice in these notes with two different divisions behind it. The 62.50 per cent above is 2,50,000 registers made against 4,00,000 of rated capacity. The division there is Anjani Stationers dividing its own output by its own ability. The very same 62.50 per cent appears in the Setu Bazaar material quoted throughout, where it is Setu Bazaar's standing base measured against Setu Bazaar's own revenue. Two businesses, two divisions, one number, and reading either as confirmation of the other would be reading a coincidence as evidence. Which division produced a figure is what matters, never which figure it is.

So what is a share worth once the field is crowded?

Answer that by asking what a share is normally sent to do. A share is sent to do three jobs. The three do not fail in the same way, and one of them is repairable, so lumping them together and dismissing all three at once would be lazy.

The first job is to say whether a seller is big. Saying whether a seller is big needs a figure underneath the line drawn around a field, and drawing that line is a decision rather than a measurement. Two careful people can draw it in two places, both honestly, and get two different answers. In a crowded field there are more places to draw it and fewer people who agree, so the job fails before any counting begins.

The second job is to say whether that seller is getting bigger. Saying whether a seller is getting bigger needs the same figure underneath twice, on two dates, drawn the same way both times by whoever drew it. A convention nobody agreed on once cannot be held still across two years by two different sets of hands, and a share that moves because the line moved looks exactly like a share that moves because the seller did. The moving line is the worse of the two failures, and it is the one that reaches print most often.

The third job is to say how much room is left. Saying how much room is left does not need the total at all. A total with nothing known about its shape is only a number, and shape is what decides whether any of it can be reached, so the third job needs the shape of the field. Two thousand small merchants and four large ones plus one thousand nine hundred and ninety six small ones can add to the very same total and offer completely different room. The three jobs fail in three different places, and only the third one fails for a reason that gathering more information could actually end. Shape is a counting job. Somebody could go out and count who sold what. The first two fail on a decision about where a line goes, and no survey, however large, settles that.

Say the professional consequence plainly, in one sentence. A share figure in a crowded field is expensive to get, unstable between two dates, and answers the smallest of the three questions it was sent to answer.

What is written instead of a share that cannot be got?

Three lines, and the order is the teaching rather than a convenience. One, the count, said plainly as a count: 2,000 merchants, gathered by asking rather than read off anything. Two, the average, with its label attached inside the same sentence rather than parked in a footnote: Rs 24,00,000/- of goods a merchant, gross flow, describes a place in that field and no merchant in it. Three, the named absence: the split across the 2,000 names is published nowhere, so no merchant's share is known and none is offered here.

A count, a labelled average and a named absence is a finding, and the same three lines with a plausible share in place of the third is a fabrication nobody will ever check. That is not an exaggeration for effect. There is no register anywhere against which the invented third line could be tested, and a line nobody can test is exactly the line that survives.

One of the three always gets deleted, and it is always the same one. The third. The third line is the only one that reads like an admission, and a document being cut for length loses its admissions first. The third line is also the only one of the three that tells the next reader what to go and find, so cutting it removes both the honesty and the instruction in a single stroke. The line should stay. When a reviewer asks for it to be softened, the softening is the fabrication arriving politely.

How does anybody handed a share figure actually use this?

Which part of this is settled somewhere other than here?

Almost none of it. The mechanism described here is not a rule anybody set. A ratio needs a figure underneath it in every market on earth, and nowhere does anybody keep a record of everybody else. Two of the conventions used do come from a particular place, and neither of them is a number.