Take Rate: What a Platform Keeps of What Passes Through

A take rate is the share of everything crossing a marketplace that the marketplace keeps as revenue. Divide the platform's own revenue by the gross value flowing across it, over the same period and the same transactions. Of the Rs 500 crore crossing Setu Bazaar, Rs 20 crore is booked as its revenue, so the take rate is 4.00 per cent. The number is easy; almost every conclusion drawn from it is not.

Setu Bazaar is an invented two-sided marketplaceA business that does not buy and resell, but instead brings two separate sets of people together and charges for the meeting. How that structure behaves is covered separately. used all through these notes, and it carries the arithmetic below. Setu Bazaar does not buy goods and resell them. Sellers list, buyers pay, goods move, and Setu Bazaar takes a commissionA charge calculated as a share of the value of a transaction rather than as a flat fee. The charge a business should set, and in what form, is covered separately. out of the middle. Standing beside a flow without owning it is what makes a take rate a sensible measure at all: there is a flow, the platform stands beside it, and the question is how much of the flow sticks to the platform on the way past.

Finding revenue on a statement, and reading a margin, are covered elsewhere. The part almost nobody does slowly is the argument about what belongs in the number being divided by.

Which two numbers make a take rate, and what has to be true of them?

A take rate needs exactly two inputs. The first is the revenue the platform itself recognises: what it billed and kept, not what buyers paid in total. The second is the gross value that crossed the platform in the same window. Divide the first by the second and multiply by a hundred.

The arithmetic is a single division. A single division travels easily, and it goes wrong just as easily. Both numbers must cover the same period and the same set of transactions, and the first place a reported rate goes wrong is that they do not. A revenue figure for a full year divided by a flow figure for the last quarter produces a number that looks like a take rate, sits in a table like a take rate, and means nothing at all. A revenue line that includes an advertising business, divided by a flow counting only goods orders, fails the same way. The numerator now contains money the denominator never touched.

The everyday version is worth holding on to. A wedding caterer takes Rs 4,00,000/- of business through a hall this season and pays the hall Rs 40,000/- for the use of it. The hall's take rate on that caterer is ten per cent. Then comes the awkward question: does the Rs 4,00,000/- include the two functions the caterer cancelled and was refunded for? And when the hall counts what passed through it, does it count the drinks the guests brought in themselves from a shop outside? Those drinks crossed the hall and never crossed the caterer's bill. Two people could answer that hall's take rate honestly and differently, and neither would be lying. Two honest and different answers to one question is the whole of the difficulty.

What does the calculation look like when it is done slowly?

Setu Bazaar's year, invented for teaching: goods worth Rs 500 crore were ordered across the platform, and Setu Bazaar recognised Rs 20 crore of revenue. Twenty divided by five hundred is 0.0400. Multiplied by a hundred, that is a take rate of 4.00 per cent. Written the other way round, Rs 500 crore multiplied by 0.0400 gives Rs 20 crore back. The reverse check is worth doing every time, and it catches a misplaced decimal in about two seconds.

Two other bases look plausible and are not. Dividing the same Rs 20 crore by the Rs 480 crore the sellers received gives 4.17 per cent. The figure is close enough to the true one to survive a glance, and wrong in a way nothing alongside it will announce. Dividing it by the platform's own revenue gives 100.00 per cent, and dividing any number by itself gives the same. Neither answers the question a take rate asks, so the panel above turns both readings red when it produces them.

Now the same arithmetic at the level of one buyer. Setu Bazaar has fifty thousand buyers. Each transacts Rs 1,00,000/- of goods across the platform in the year and brings Rs 4,000/- of revenue with them. Four thousand divided by one lakh is 0.0400 again, the same 4.00 per cent. Multiply back up and it reconciles: fifty thousand buyers times Rs 1,00,000/- is Rs 500 crore, and fifty thousand times Rs 4,000/- is Rs 20 crore.

The trap here catches people who are otherwise doing everything right. The per buyer calculation is not a second route that confirms the first one; it is the same relationship written at two scales, and two writings of one relationship cannot disagree. Multiply the buyer pair by fifty thousand and it becomes the platform pair exactly. The two are built from the identical numbers with a common factor cancelled out of the top and the bottom, so an error in the platform rate appears unchanged in the buyer rate.

So what does the match actually show? The match shows that the rate is uniform across buyers. Every buyer on Setu Bazaar is charged on the same basis, so no group of buyers is carrying a heavier charge than another. Uniform charging is a real fact about the platform and it is worth knowing. To see why it is not automatic, take a pair of buyers charged differently: one transacts Rs 1,00,000/- and is charged Rs 4,000/-, the other transacts Rs 1,00,000/- and is charged Rs 2,000/-. Together they bring Rs 6,000/- on Rs 2,00,000/- of flow, a rate of 3.00 per cent, and 3.00 per cent is neither buyer's rate. A platform level take rate is always the value weighted average of the individual rates, and it equals every individual rate only when the charge is uniform.

What does one divided by the take rate show?

Turned upside down, the fraction becomes the reading most worth carrying away. One divided by 0.0400 is 25.00. Twenty five is the multiple by which the flow overstates the revenue. For every rupee Setu Bazaar recognises, twenty five rupees of goods value crossed the platform, and twenty four of those twenty five rupees belong to somebody else.

A platform that reports its flow as its size is reporting a number twenty five times its revenue, and that is the single most common misreading of a marketplace. It is not usually a lie. The flow is a genuine measure of how much activity the platform organises, and for an operations team it is the number that determines how many warehouses and how many support staff are needed. The failure is in the transplant. Every other business the reader met that year reported a number of that shape and it was revenue, so when the flow arrives in the same slot the reader supplies the word revenue without being told to.

A household example gives the size of it. A courier who moves Rs 25 lakh of other people's parcels in a year and charges Rs 1 lakh to do it has not earned Rs 25 lakh. The parcels are visibly not the courier's, so nobody makes that mistake about a courier. On a marketplace the goods are invisible, the money moves through one payment screen, and the same mistake becomes easy.

Why is the denominator the part nobody has agreed on?

Everything so far treated the Rs 500 crore as a settled fact. It is not. There is no rule anywhere that says what a marketplace must count in the flow it publishes, and the choices are not small. Does the flow include the taxes collected on an order and handed straight to the government? Does it include the shipping charge that goes to the courier? Does it include orders that were placed and then cancelled, or delivered and then returned for a refundMoney returned to a buyer after a sale is reversed, usually because the goods came back. Whether the original order still counts in a published flow figure is a reporting choice.? Does it include transactions the platform introduced but never processed, where buyer and seller settled between themselves?

Each of those is a defensible yes and a defensible no, and the label attached to the published figure barely narrows it. Some businesses call it gross merchandise value, some call it gross bookingsA label some marketplaces use for the total value of transactions arranged through them in a period. The label alone does not say which categories were counted., some call it total transaction value, and none of those names carries a fixed definition into the number.

Setu Bazaar's year, broken into its parts, with every component published so any total below can be recomputed.

The last three rows do the work. The revenue is Rs 20 crore in every one of them. Setu Bazaar charges nothing on a cancelled order, nothing on the tax, nothing on the shipping it hands to a courier, and nothing on a transaction it merely introduced. So the numerator sits perfectly still while the denominator moves from Rs 625 crore to Rs 400 crore, and the reported rate travels from 3.20 per cent to 5.00 per cent. Two reported take rates are usually measuring different things, and comparing them without reading both definitions is comparing nothing.

Take the narrowing on its own. Drop the Rs 100 crore of cancelled and returned orders out of the denominator and the flow falls to Rs 400 crore. No revenue was ever earned on those orders, so the numerator does not move. Rs 20 crore over Rs 400 crore is 5.00 per cent, and the inverse falls from 25.00 to 20.00. The reported rate rose by a full percentage point without Setu Bazaar changing one line of what it charges anybody.

What actually makes a reported take rate move?

Three things move a take rate, and only one of them is a decision about price. The first is exactly what a reader assumes: the platform changed what it charges, raising a commission from four per cent of an order to five, or adding a fee that did not exist before. The second is mix. If the categories that carry a heavier charge grow faster than the categories that carry a lighter one, the blended rate rises while every individual charge stands still. The third is definition, the whole of the section above.

A rate can move without anyone changing any charge, and a reader who treats a rate change as evidence of a pricing decision will be wrong more often than right. Mix does it quietly and definition does it in a single footnote. The practical consequence is that the direction of a rate is not, on its own, information about strategy. Movements are usually reported in basis pointsA hundredth of one percentage point. A rate moving from 4.00 per cent to 4.10 per cent has moved ten basis points, and the unit exists so small movements can be stated without ambiguity., and a movement of a few dozen of them is entirely capable of being pure mix.

Think about a household shop that sells both groceries and mobile phones. The shopkeeper never changes a single price, but a month when phones sell heavily has a completely different overall margin from a month when they do not. Nothing was decided. The mixture changed. A marketplace with a dozen categories has the same machinery running underneath it every quarter.

Setu Bazaar's Rs 500 crore splits, for this illustration alone, into two groups. Rs 250 crore of household goods charged 2.00 per cent brings Rs 5 crore, and Rs 250 crore of electronics charged 6.00 per cent brings Rs 15 crore. Five and fifteen is the published Rs 20 crore, and Rs 20 crore over Rs 500 crore is the published 4.00 per cent.

Now move four fifths of the flow into electronics and cut the electronics charge from 6.00 to 4.50 per cent. Rs 100 crore at 2.00 per cent brings Rs 2 crore, Rs 400 crore at 4.50 per cent brings Rs 18 crore, and the year again reports Rs 20 crore on Rs 500 crore, a take rate of 4.00 per cent, identical to the year before it. A take rate can stand perfectly still while the mixture and the charges underneath it both move a long way, and the reader watching only the rate records a year in which nothing happened.

What does a take rate refuse to say?

A take rate refuses to say whether the platform earns anything, and Setu Bazaar is the demonstration. Its take rate is 4.00 per cent and its result for the year is a loss of Rs 2.5 crore. The build is short: Rs 4,000/- of revenue per buyer, of which Rs 2,000/- survives as contributionWhat is left from one customer after the costs that arise only because that customer exists. How to build and read it is covered separately. after the costs that arise only because that buyer exists, times fifty thousand buyers, giving Rs 10 crore. Against that sit fixed costsCosts that do not move when one more customer arrives or leaves, such as the engineering team and the office. Fixed costs sit above the per customer arithmetic. of Rs 12.5 crore. Ten less twelve and a half is a loss of Rs 2.5 crore.

A take rate is a measurement of capture and says nothing whatever about profit. The rate lives entirely above the cost line. The rate describes how much of the flow attaches to the platform and stops there. Whether the amount captured covers what the platform spends is a different calculation with different inputs, and it is covered separately under unit economics, where the fixed cost line and the break even buyer count are worked through properly.

Three more refusals belong in the same list. Whether a rate is sustainable depends on what sellers can bear and what alternatives they have, and a take rate reports neither. Nor does a take rate say what the platform is worth, since valuing a marketplace takes a wholly different set of inputs. A heavier charge on a shrinking flow and a lighter charge on a growing one both produce revenue, and the rate alone cannot separate the two, so a higher rate is not on its own better than a lower one.

What should be asked before writing a reported take rate down?

An analyst reading a marketplace, a lender sizing an exposure to one and a strategy team building a competitive picture all reach for the reported take rate early. The rate is one number, and one number appears to summarise a business. The discipline is to ask one question before the number is written into anything.

The answer changes the number and is almost never volunteered, so what sits in the denominator is settled before the rate is written down. Not after, when the figure has already been settled on and the definition arrives as an inconvenience. Before, so the number goes into the notes with its basis attached to it like a unit.

A lender does the same work for a different reason. If a marketplace is borrowing against the activity crossing it, the flow figure is the collateral story and the revenue figure is the repayment story, and the ratio between them is exactly the take rate. A lender who accepts a flow number without its definition has accepted an unknown fraction of it as the revenue that actually services the loan.

Setu Bazaar and Kadamb Mart are invented.
Educational material. Not advice on any investment, tax, budget or market position.