Customer Economics: What One Buyer Earns, Costs and Repays
Ten fields in, and every step of the build shown
Every field below is a line copied off a set of statements or a disclosure, not a figure somebody had already worked out. The five in the first group were measured and written down by somebody. The four underneath them were not measured by anybody. They sit apart for that reason, and the panel keeps saying so. The tenth is not a figure at all but a choice about where a cost is charged, and it is there so that the mistake worked through below can be reproduced and its effect on a buyer that is perfectly sound observed. Every division the panel performs is walked through by hand further down. Nothing typed here is kept; closing the tab discards it.
| Period | Of 100 buyers, still there | Amount that period | Discounted to today | Running total, plain | Running total, discounted |
|---|
Why can a closed period be divided up at all?
Because every number the division needs has already happened. The rupees came in, the rupees went out, somebody counted the buyers, and all of it was written down beforehand. Because none of these three answers needs a view about next year, anybody holding the same set of statements can check them. Two people handed the same statements should land on the same figures, or find out exactly where they disagreed.
The business being divided up is Setu Bazaar, a marketplace used throughout these notes, and its trading year is set out in full elsewhere in these notes. The Rs 20,00,00,000/- on its top line is what the marketplace kept for itself as its take rateThe share of an order's value that a marketplace keeps as its own revenue, rather than passing on to the seller who supplied the goods. on the orders placed across it, not the value of the goods that moved. The marketplace served 50,000 buyers in the year. Three cost lines moved whenever a buyer arrived. Payment handling accounted for Rs 700/- of that movement, buyer servicing for Rs 500/-, and delivery and returns support for the largest share at Rs 800/-. The costs no single buyer moved came to Rs 12,50,00,000/- for the year. Winning one buyer cost Rs 6,000/-, and that is what a newly won buyer cost rather than a charge laid against all 50,000, most of whom were won in earlier years.
One test decides which costs come out, and the test itself is covered separately. A cost belongs in the per buyer subtraction only if it would have moved had that one buyer never existed. The three named lines pass. The lines inside the Rs 12,50,00,000/- do not. Applying a settled test is a different exercise from building one.
Which four numbers does customer economics need, and where is each one found?
Four, and no more, and the panel above asks for more than four only because it also accepts the things nobody measured. The four come in the order an analyst actually goes looking for them, and the question throughout is always where, never what.
One, revenue for a stated period. Top line of the income statement. Revenue arrives with its reporting periodThe stretch of time a set of statements covers, usually a quarter or a year. Every figure inside them describes that stretch and no other. attached to it. No other input on this list comes with its own dates.
Two, the buyer count. Not on the income statement. Not in the notes. The buyer count sits in an operating metrics disclosure, an investor deck, or a paragraph of management commentary, and it is an operating metricA number a business reports about its activity rather than about its money. No auditor signs it, and the definition behind it is the reporting business's own. rather than an audited figure. Which buyers were counted, and on what rule, is a real question with a real answer, and it is settled in these notes under Average Revenue Per User. The requirement at this step is smaller and firmer: the count used and where it came from are written down.
Three, the costs that move with a buyer. The individual lines are printed, in the cost scheduleThe breakdown behind a single expense total, listing the separate costs that were added together to reach it. The schedule usually sits in the notes rather than on the face of the statement. or in the notes behind it. Which of those lines qualify is printed nowhere. The volume test settles that, not the place the accountant chose to group them.
Four, the spend to win a buyer, together with the fixed base it sits inside. Both come off the same schedule, and they come together on purpose. The spend to win buyers is not a separate pot sitting outside the accounts. The spend is one of the lines inside the Rs 12,50,00,000/-, and on this business it is the largest of them.
One of those four is not printed in the accounts at all, and a second carries a judgement no statement records. Two people working from the same statements can therefore arrive at different answers. Think of working out what one child costs a household. The bank statement exists, so the school fee is not in dispute. The share of the electricity bill is not written anywhere, and two people in the same house will split it differently and both be reasonable.
Customer economics needs four inputs. Which one of them is not printed anywhere in the accounts?
How is revenue for one buyer computed?
One division. Rs 20,00,00,000/- of revenue over 50,000 buyers is Rs 4,000/- of revenue a buyer. Then the multiplication runs straight back where the reader can see it done: 50,000 buyers at Rs 4,000/- each returns Rs 20,00,00,000/-. The check takes two seconds and it catches a slipped decimal. A slipped decimal is the single most common thing to go wrong when a figure has seven zeroes behind it.
Field note. The numerator sits on the income statement and the denominator does not, so the count used and the document it came from are recorded in the same place as the answer. Change the counting rule and this figure moves while nothing whatever about any buyer has changed, and a reader who finds a different number later needs to be able to see who counted differently.
Setu Bazaar's revenue per buyer is Rs 20,00,00,000/- over 50,000, or Rs 4,000/-. Somebody rebuilds it using a count of only the buyers active in the final month and gets a higher figure. What happened?
How is contribution for one buyer computed?
Start at the Rs 4,000/- and take out the three lines that move. Payment handling takes Rs 700/-. Delivery and returns support takes Rs 800/-. Buyer servicing takes Rs 500/-. The three lines come to Rs 2,000/-, and Rs 2,000/- of contribution is what is left.
| The build, one buyer, one period | Where the line was found | Amount |
|---|---|---|
| Revenue a buyer | Income statement top line, divided by the buyer count | Rs 4,000/- |
| Less payment handling | Cost schedule, and it moves with a buyer | Rs 700/- |
| Less delivery and returns support | Cost schedule, and it moves with a buyer | Rs 800/- |
| Less buyer servicing | Cost schedule, and it moves with a buyer | Rs 500/- |
| The three lines together | Added, not disclosed as a total | Rs 2,000/- |
| Contribution a buyer | Computed here, printed nowhere | Rs 2,000/- |
Field note, and it is the sharpest one here. The three lines are found in the cost schedule, but the schedule does not say they belong here. The volume test decided that, and somebody equally careful could have decided it otherwise. The split between a cost that moves with a buyer and a cost that does not is an assumption and never a disclosure, so the lines taken are written down the moment they are taken.
There is a second route to the same place. 50,000 buyers multiplied by Rs 2,000/- gives Rs 10,00,00,000/-, the published total contribution for the year. The agreement between the two routes is worth precisely nothing as a test. The same buyer count appears in the division that produced the Rs 2,000/- and in the multiplication that undoes it, so the two cancel and the routes cannot disagree whatever went wrong upstream. The match is an identity, and dressing an identity up as a confirmation is a fault rather than a feature.
Setu Bazaar's contribution a buyer is Rs 2,000/-, and 50,000 buyers times Rs 2,000/- comes to the published Rs 10,00,00,000/- of total contribution. What has that confirmed?
How long does one buyer take to repay what it cost to win?
Rs 6,000/- was spent to win the buyer. Rs 2,000/- of contribution arrives from that buyer in a year. Rs 6,000/- over Rs 2,000/- a year is 3.00 years. A payback question hands back a length of time. A length of time can be set beside a loan term, a lease, or the notice period on a warehouse. None of those comparisons would work if the answer came back as a rupee amount.
Field note, and this is the placement rule that decides everything downstream. The Rs 6,000/- is found inside the Rs 12,50,00,000/- of fixed cost, where it is the largest single item. The Rs 6,000/- is not inside the Rs 10,00,00,000/- of contribution, and it never was. Contribution counts a buyer that was served across the period. The Rs 6,000/- is paid a single time, at the front, for a buyer that was won. Three correct figures will be added together wrongly unless somebody says where each one sits.
A household knows this without being taught it. Getting a water connection put in costs something once. The water itself costs something every month. Both amounts are real and both were paid, and a single monthly figure built by adding them describes no month that ever happened.
Setu Bazaar's cost of winning a buyer is Rs 6,000/-. Where does that spend sit?
How many buyers does it take to cover the fixed base?
Another single division, and the answer comes back in heads rather than in rupees. Rs 12,50,00,000/- of fixed costCost the period carries whatever the level of activity, so it does not rise when one more buyer arrives and does not fall when one leaves. over Rs 2,000/- of contribution a buyer is 62,500 buyers. At that count, total contribution and the fixed base are equal.
Field note. The fixed base is found in the cost schedule, as the lines the volume test rejected, and the Rs 2,000/- comes straight out of the build above. The second half of that sentence matters more than it looks. The division inherits every assumption made when the cost lines were classified, so a reader who disagrees with one line in the schedule disagrees with the 62,500 as well.
Then the rounding rule, stated openly and settled before any untidy division arrives. A business gets no cover at all from part of a buyer and stays short until the whole buyer turns up, so a break even count that comes out with a fraction attached is taken up to the next whole buyer. On these figures the rule sits there and does nothing. Rs 12,50,00,000/- divided by Rs 2,000/- lands on 62,500 with no remainder whatever. Nothing was rounded here. The rule is stated because the panel above returns fractions the moment an input changes, and it is real there.
Setu Bazaar's fixed base is Rs 12,50,00,000/- and each buyer contributes Rs 2,000/- a year. Without scrolling back to the panel, how many buyers does the business need before contribution and fixed cost are equal?
The published rule is that a break even buyer count landing on a fraction is taken up to the next whole buyer. How does that rule apply to Setu Bazaar's 62,500?
Where does each cost actually sit?
Placement is where the arithmetic is got wrong in practice, so lay all of it out once. There are three positions and every rupee Setu Bazaar spent occupies exactly one of them. Above the contribution line sit the costs measured on one buyer served: Rs 2,000/- a buyer, or Rs 10,00,00,000/- once the whole year is in view. Below the line sits the fixed base of Rs 12,50,00,000/-, carried by the period whatever happens. Inside that base, not beside it, sits the spend to win buyers. A cost appears exactly once in this arithmetic, and the only question worth asking about any line is which side of the contribution line it falls on.
The fixed base is not divided across the 50,000 buyers to produce a profit for each of them. Spreading it that way is overhead absorptionSpreading costs that no single unit moves across the units anyway, using a chosen rule, so that each unit carries a share of them., it answers a different question, and why every buyer can be profitable while the business as a whole loses money is worked through in these notes under Unit Economics: Profitability at the Level of One Customer. The published result for the year, a loss of Rs 2,50,00,000/-, is quoted once so that the arithmetic can be seen landing where the business itself reported: Rs 10,00,00,000/- of contribution set beside a fixed base of Rs 12,50,00,000/-.
Why can a rupee, a duration and a count not be combined into one score?
Rs 2,000/- is an amount of money, 3.00 years is a stretch of time, and 62,500 is a number of heads. The three figures came out of the same set of statements and they answer three questions that have nothing in common: what one buyer leaves behind once the costs that follow that buyer around have been paid, how long that buyer has to keep coming back before the money spent winning them is back in the till, and how many buyers there would have to be before the costs nobody's arrival changes are covered. None of the three can be added to, multiplied by or scored against either of the others, and anybody offering a single customer economics score has multiplied a rupee by a year.
The three can also point in different directions at once, and that is not a defect in the arithmetic. A business can hold a healthy Rs 2,000/-, wait a long time to be repaid, and still sit well short of the count it needs. Which of the three a particular reader stops at is a separate question and it is worked through elsewhere in these notes.
An associate combines Rs 2,000/- of contribution, 3.00 years of payback and 62,500 buyers into a single customer economics score. What is wrong with that?
What can this arithmetic not settle on its own?
Customer economics says nothing about what a buyer is worth. A worth placed on a buyer, usually called a lifetime value, is the most quoted number in this whole subject. The panel above builds one the moment it is given what it needs. The inputs it needs are the lesson, rather than the number it returns.
Set the two side by side. The payback needs Rs 6,000/- that was already spent and Rs 2,000/- that was already earned, and both of those sit in a period that has closed and been written down. A worth placed on a buyer needs two further things: how long that buyer will keep coming back, which is measured by following a cohortA group of buyers who all started in the same period, tracked forward together so that later joiners never enter the group and dilute what is being observed. of buyers forward and is covered separately in these notes, and a discount rateThe rate used to convert a rupee expected later into a rupee today. Nobody observes it in the accounts; whoever builds the figure has to choose it. to turn rupees expected later into rupees now. Nobody measured either. Somebody has to pick them, and the panel leaves that choice to whoever is working the figures. A duration needs nothing about the future, and a number that requires a forecast is not evidence.
The panel opens on the only retention rate a closed period supports, and that is no rate at all: every buyer returning, no rupee discounted, and ten periods counted. On those settings the ladder is the closed period written out ten times, Rs 2,000/- in each of them, and the lifetime value reads Rs 20,000/-. With twenty in the horizon instead it reads Rs 40,000/-. Nothing about Setu Bazaar moved. The figure doubled because the number of periods counted doubled, and that is the whole of what a lifetime value resting on an unmeasured retention rate can tell anybody.
Now a real assumption. At a retention rate of eighty, meaning eight buyers in ten come back each period, those same ten periods come to Rs 8,925/-. A buyer who leaves stops mattering long before the horizon does, so twenty periods come to Rs 9,883/- and thirty to Rs 9,986/-. Moving the rate from seventy per cent to ninety takes the figure from Rs 6,478/- to Rs 13,027/-, slightly better than double. Every audited number stays exactly where it was. A lifetime value on this business is a statement about the retention rate somebody chose rather than a statement about the business.
There is a second way to inflate it and the panel will do that one too. Switch the amount carried forward from contribution to revenue and the ladder repeats Rs 4,000/- a period instead of Rs 2,000/-, so the opening settings return Rs 40,000/- rather than Rs 20,000/-. The extra Rs 20,000/- is ten periods of the Rs 2,000/- of costs that move, every rupee of which was paid out and never came back. A lifetime value built on revenue counts rupees the business spent as though it had kept them.
No worth for a buyer is reported here, and no price is put on a name a business has built for itself; both are covered separately. In each case the first question is what the arrangement does: what it costs to win a buyer, how much that buyer leaves behind, how long the money takes to come back. Pricing comes later, in a different kind of work, and by different people.
Why a 3.00 year payback rather than a worth for one buyer?
In what order does a practitioner pin the four inputs down?
Before any of the three answers gets written, four things get settled and written down. The order is not decorative. Each one closes off a way the next reader could get a different number from the same statements.
Four things settled before a single division is done
- Name the period, and check that all four inputs cover it. A revenue figure for a full year set against a buyer count taken at one month end is the quietest error on this list, because both numbers are correct and neither looks wrong.
- The counting rule behind the buyer count is named, along with where it came from. Anyone who counted differently will land somewhere else, and that argument should be about the rule rather than about the arithmetic.
- Name which cost lines passed the volume test and which did not. This is the assumption everything downstream inherits: the contribution, the payback and the break even count all move together if one line changes sides.
- Name where the spend to win a buyer sits, and check it has not also been taken out above the line. Say out loud that it sits inside the fixed base and is counted there in full.
The fourth check is the one people skip, and its failure is the only one that is invisible. A cost counted twice looks exactly like a business with worse economics, and nothing in the arithmetic raises a hand. With all four written down, anybody can rebuild the three figures and anybody can dispute them, which is the only property that makes a number worth quoting at all.
Where does this arithmetic go wrong in the hands of a careful reader?
The same rupees charged twice
An associate is asked to build Setu Bazaar's customer economics from the statements. The associate divides Rs 20,00,00,000/- by 50,000 and gets Rs 4,000/-. Out come the three lines that move: Rs 700/- for payment handling, Rs 800/- for delivery and returns support, Rs 500/- for buyer servicing. Rs 2,000/- is left. Then, wanting to be thorough, the associate takes out the Rs 6,000/- it cost to win the buyer as well, on the entirely reasonable ground that winning the buyer really did cost Rs 6,000/- and the money really did leave the business.
The buyer now shows minus Rs 4,000/-. The note goes upstairs saying every buyer Setu Bazaar has loses money. The true position is the exact opposite, and every single figure the associate used was correct.
Here is what went wrong. Contribution counts a buyer that was served across the period, and the Rs 6,000/- was paid once, at the front, for a buyer that was won. The Rs 6,000/- is already sitting inside the Rs 12,50,00,000/- of fixed cost, counted there in full. Taking it out a second time above the contribution line charges the business for the same rupees twice. The right treatment is the one performed above: the spend stays where it sits and is set against the contribution as a length of time, Rs 6,000/- over Rs 2,000/- a year, or 3.00 years.
The cost of the error is a verdict on the wrong thing. Setu Bazaar's buyers work. Its fixed base is the size of its problem, and the note that went upstairs never mentioned the fixed base at all. The fix is one sentence: a cost is counted once, and the only question is which side of the contribution line it falls on.
An associate subtracts Rs 700/-, Rs 800/-, Rs 500/- and the Rs 6,000/- cost of winning a buyer from Rs 4,000/- of revenue a buyer, reports minus Rs 4,000/- and concludes that every buyer loses money. Every figure used is correct. What went wrong?
What does the Indian setting fix here, and what does it leave open?
Two things, and both are conventions rather than rules. Every amount is written in rupees and grouped the Indian way, so Rs 12,50,00,000/- reads as twelve and a half crore, and Setu Bazaar is written as a private limited company trading in India. Beyond that, the arithmetic is the same wherever the reader is standing. The one place a real rule book does sit behind the arithmetic is the line drawn between a cost that moves with activity and a cost that does not. The line is a cost accounting question before it is an analytical one, and which businesses are required to keep records of that kind changes from time to time. The Institute of Chartered Accountants of India holds the current position on thresholds, rates, classes of company and dates.
Which questions belong to other work?
Customer economics builds three figures for one buyer out of a period that has closed, and it stops there. A worth for a buyer, and a rate for turning later rupees into present ones, both rest on somebody's choice, and the panel above says so on every line. Whether any business is a good one to hold is a separate question. Customer economics stops at how a business runs, and what a business is worth is taken up separately.
How long a buyer stays, and what keeping one is worth, is covered under Customer Loyalty: Retention as an Economic Asset. Building the Rs 6,000/-, and why two honest teams would build it differently, is covered under Customer Acquisition Cost: What It Costs to Win One Buyer. Why every buyer can be profitable while the business as a whole loses money is covered under Unit Economics: Profitability at the Level of One Customer. Which buyer count belongs in a denominator is settled under Average Revenue Per User. Sorting buyers into groups, and judging whether one customer is too large a share of anything, is a different exercise again.
What a business should spend, where it should spend it, and which buyers it should chase are decisions rather than arithmetic, and no division performed on a closed period produces them.
What stands behind the numbers divided here?
| Where it was read | What was read from it | Site | When read |
|---|---|---|---|
| These notes, Business Fundamentals and Models | Setu Bazaar's revenue for the year, its buyer count, the three cost lines that move with a buyer, its fixed base, the price it paid to win a buyer and its result for the year | finmaverick.com | 23 August 2026 |
| These notes, Revenue and Pricing | That the Rs 20,00,00,000/- on the top line is what the marketplace kept for itself and not the value of the goods that crossed it | finmaverick.com | 23 August 2026 |
| These notes, Average Revenue Per User | That a buyer count rests on a counting rule somebody chose, which is why the rule used is written down alongside the answer | finmaverick.com | 23 August 2026 |
| The Institute of Chartered Accountants of India | Where a reader goes for the actual position on classifying and recording costs | icai.org | 23 August 2026 |
Setu Bazaar is invented.
Educational material. Not advice on any investment, tax, budget or market position.