Unit Economics: Profitability at the Level of One Customer

Unit economics asks one question: does a single customer bring in more than that customer costs to serve? Take the revenue from one customer, subtract only the costs that move when that customer arrives or leaves, and the remainder is contribution. Contribution is not profit. Profit waits until the costs that no customer moves are covered, and that gap is why a business full of profitable customers can still lose money.

Three things already established carry everything below. A business model states what a company does and who hands it money. The form of the charge, whether that is a fee on every transaction or a flat subscription, fixes how much revenue one customer brings. And the business life cycle raised a question it left open: when a young company reports a loss, is that loss the sound of something being built or the sound of something failing? The question turns out to be arithmetic rather than judgement.

Setu Bazaar, an invented business, is a two-sided marketplaceA business that has to attract two different sets of users who need each other, such as sellers and buyers, rather than one set of customers. How such a business behaves is covered under platform businesses. where fifty thousand buyers each transact Rs 1,00,000/- of goods a year, and where the marketplace keeps 4.00 per cent of what passes through. Anjani Stationers Private Limited, another invented business, manufactures paper and stationery and sells it to a few dozen business buyers. The company reports revenue of Rs 2,70,00,000/- and an operating result of Rs 41,50,000/-.

What is the unit, and why must it be named before anything is computed?

Start with a tea stall on a street corner. The owner wants to know whether he is making money, and there are at least three honest versions of that question. Per cup? Most people buy two, so perhaps per customer? He pays the same rent whether twelve people come or two hundred, so perhaps per day? Each version gives a different number and none of them is wrong. The mistake would be reporting one of them without saying which.

Changing the unit changes every number in the analysis, so an analysis that does not say what its unit is cannot be checked by anyone. The unit is not a technicality. An unstated unit is the most common reason two people looking at the same business argue about figures that were never comparable in the first place.

Watch what happens to Setu Bazaar when only the unit changes and nothing else does. One buyer brings Rs 4,000/- of revenue in the year. The same buyer places roughly twenty orders, so one order brings Rs 200/-. Measured against the goods rather than the marketplace, every Rs 100/- of gross merchandise valueThe total value of goods that pass through a marketplace before the marketplace's own share is separated out. Why that total is not revenue is covered under take rate. brings Rs 4/-. And measured against revenue itself, every rupee of revenue brings 50 paise. Same year, same business, same underlying facts, four different sentences.

The buyer is the unit used from here on, and the reason is worth stating rather than assuming. The cost of winning a customer attaches to the customer, not to the order. A marketplace pays once to bring somebody in and then hopes they order twenty times. With the order as the unit, the acquisition cost has nowhere sensible to sit. The most important question about winning a customer then quietly disappears.

How does one customer's revenue become that customer's contribution?

Contribution is revenue from the unit less the costs that vary with it. The word doing all the work is vary. The test is not whether a cost feels related to the customer, or whether an accountant filed it under a customer heading. The test is whether the cost would move if that one customer had never existed.

A household that takes in a paying guest gets this immediately. The guest pays Rs 8,000/- a month. The extra rice, the extra electricity and the extra gas cylinder move because there is one more person in the house. The rent does not move. The property tax does not move. So the contribution from the guest is Rs 8,000/- less the food and the power, and that number is real and useful. The rent is still standing there waiting to be paid, so that number is not the household's profit.

Setu Bazaar's build is the same shape with bigger numbers. A buyer transacts Rs 1,00,000/- of goods across the year, and the marketplace keeps 4.00 per cent, so revenue from that buyer is Rs 4,000/-. Three costs move with the buyer: payment handling of Rs 700/-, delivery and returns support of Rs 800/-, and buyer servicing of Rs 500/-. The three moving costs add to Rs 2,000/-. Contribution is the Rs 2,000/- that remains, a contribution margin of 50.00 per cent on the revenue Setu Bazaar keeps.

Contribution is not profit per customer. The difference between the two carries everything that follows. The moment somebody says a customer is profitable, ask which of the two they mean, because the answers can point in opposite directions at the same time.

Both numbers set beside each other look like this. Contribution needs one buyer and nothing else: Rs 4,000/- in, Rs 2,000/- of moving costs out, Rs 2,000/- left. Profit per buyer needs the whole business. Setu Bazaar carries Rs 12,50,00,000/- of costs that no single buyer moves, and spreading that across 50,000 buyers puts Rs 2,500/- on each one. Rs 2,000/- of contribution against a Rs 2,500/- share leaves a shortfall of Rs 500/- per buyer. Multiplied by 50,000 buyers, the shortfall is Rs 2,50,00,000/-, and that is exactly the loss the company reports. Both numbers are honest. The two answer different questions.

How can every single customer be profitable while the business loses money?

The business life cycle left this question open. The answer repays slow reading, and few places in business analysis let an apparent paradox dissolve into plain arithmetic and stay dissolved.

Fifty thousand buyers, each contributing Rs 2,000/-, give Setu Bazaar Rs 10,00,00,000/- of total contribution. Against that stand the costs no buyer moves: platform technology of Rs 4,50,00,000/-, warehousing and logistics of Rs 3,00,00,000/-, brand and category marketing of Rs 3,00,00,000/-, and salaries and offices of Rs 2,00,00,000/-. The four fixed costs add to Rs 12,50,00,000/-. Contribution of Rs 10,00,00,000/- less fixed cost of Rs 12,50,00,000/- leaves the company Rs 2,50,00,000/- short. Setu Bazaar reports a loss, and not one of its fifty thousand buyers is the reason.

The fixed costs sit above the unit line, and no amount of per customer profitability reaches them until there are enough customers. That sentence is the whole mechanism. A buyer can only ever contribute upward into the block; a buyer cannot make the block smaller.

Now the payoff, and it is what the business life cycle pointed towards. Two companies can report an identical loss and be in completely different conditions. In the first, the unit works and the count is short: every customer covers their own costs and then some, and there are simply not enough of them yet. In the second, the unit itself does not work: each customer costs more to serve than they bring, so every new customer makes the loss larger. A loss is the price of building something only when the unit itself already pays; where the unit does not pay, growth is not a plan, it is the problem. The statement shows one loss. Only the unit arithmetic separates the two conditions. Nobody who reads statements for a living stops at the loss line.

How many customers does it take before the two lines meet?

Once the fixed cost and the contribution from one unit are known, the number of units it takes to cover the fixed cost is a division, and the answer is a count of customers. Rs 12,50,00,000/- divided by Rs 2,000/- gives 62,500 buyers. At 62,500 buyers total contribution and fixed cost are equal, and the company reports neither a profit nor a loss.

The answer is a number of customers, and a number of customers is a far more useful thing to carry around than a percentage. Ask an operator how far off the business is and a percentage tells them nothing they can act on; 12,500 more buyers is a target somebody can be given.

Two details matter. First, the shortfall stated as a rate. Setu Bazaar has 50,000 buyers and needs 62,500, a gap of 12,500. Twelve thousand five hundred out of the fifty thousand it has is 25.00 per cent, so it needs 25.00 per cent more buyers than it currently has. The same gap divided by 62,500 instead gives 20.00 per cent, a true statement about a different question, namely what share of the break even count is still missing. Both figures are correct and they are not interchangeable, so the denominator has to be stated alongside the figure. Second, rounding. Half a customer covers nothing, so a break even count that lands on a fraction is rounded up to the next whole customer. The business stays short until the whole one arrives.

Why is payback a separate question from contribution?

A shopkeeper pays Rs 6,000/- for a painted signboard. The board brings in a little more trade, worth Rs 2,000/- a year after the extra stock it needs. Everybody agrees the board pays for itself. The shopkeeper still has to find Rs 6,000/- today and wait three years to get it back, and if the rent is due next month, an argument about whether the board pays for itself is not the argument that matters.

Customers behave exactly like that signboard. Setu Bazaar spends Rs 6,000/- to win a buyer: the discount on the first order, the advertisement that reached them, the referral paid to whoever brought them in. The acquisition money leaves once, in full, at the beginning. The Rs 2,000/- of contribution arrives once a year, in slices. Rs 6,000/- divided by Rs 2,000/- a year is 3.00 years, and three years is the payback.

The cost and the return are on different clocks. Cash does not wait for arithmetic, so a business can be sound on every customer and still run out of money. This is the reason a business that is winning customers fast can be in more trouble than one that is winning them slowly. Each new customer is a fresh Rs 6,000/- out of the bank today against a promise of Rs 2,000/- a year for three years. Grow quickly enough and a company can bankrupt itself one profitable customer at a time.

Two working notes belong beside that figure. Rs 6,000/- is itself a blended averageOne figure covering groups that behave differently, such as a single cost per customer covering both the cheap ones and the expensive ones, which hides the spread inside it.: some buyers walked in and cost nothing, some cost far more than Rs 6,000/-, and the single figure hides that spread. Deciding which spending counts as acquisition at all is an attributionThe rule that decides which marketing effort gets credited with bringing a customer in, when several of them touched the same person before anything was bought. question with no single right answer, so two honest teams can report different acquisition costs for the same business. Ask what went into the figure before comparing it with anything.

What does the whole build look like when it is laid out in one column?

An outside reader can check an analysis only by recomputing it, and every component below can be recomputed from the one above it.

Set the same three questions against Anjani Stationers and the shape of the answers changes completely. Anjani reports revenue of Rs 2,70,00,000/- and an operating result of Rs 41,50,000/-, an operating margin of 15.37 per cent, and it sells to a few dozen business buyers. Because the exact count of those buyers is not published, the honest thing to report is a range rather than a figure: at two dozen buyers, revenue per buyer is Rs 11,25,000/-, and at five dozen it is Rs 4,50,000/-. Even the low end is more than a hundred and twelve times Setu Bazaar's Rs 4,000/-. Anjani's unit and its total are both positive, so they point the same way. Setu Bazaar's unit and its total point in opposite directions. Neither is the better business on this evidence, and a single closed year of these three quantities settles nothing at all about either of them. They are two different shapes, and how selling to firms differs from selling to people is covered on its own.

What will unit economics not settle?

Everything so far is built from numbers that already exist: a revenue line, a set of costs, a customer count, an acquisition spend. Nothing has required a view about the future. The very next question everybody asks is what a customer is worth, and that question breaks the restraint.

A lifetime value is a valuation, and lifetime values, discount rates and retention assumptions are covered separately. The line falls exactly there for a reason. Payback needs two numbers that have already happened: what winning a customer cost, and what a customer contributed in a year. One divided by the other gives a duration. A lifetime value needs two things that have not happened and that nobody has measured. The first is how many years the customer will stay, a churn rateThe share of customers who stop buying over a stated period. The rate is measured after the event and is not the same thing as a forecast of who will leave next year. assumption dressed as a fact. The second is what a rupee arriving in a future year is worth today, and answering that means choosing a discount rateThe rate used to restate a future rupee as an amount today. Selecting one is a valuation step, and valuation is covered separately.. Choosing a discount rate is valuation. How a business works is settled before anybody puts a price on it, and putting a price on it is covered separately.

Three more things unit economics will not settle, and each is worth being blunt about. Unit economics looks backward at a period that has closed, so it does not establish whether customers will stay. Nor does it establish whether the fixed costs are the right size. Rs 12,50,00,000/- of platform and warehouse might be exactly right for a business heading to 62,500 buyers and absurd for one that will never get there. And it does not establish whether the business is worth owning, a question no arithmetic about one customer can reach. Unit economics does establish, precisely and checkably, whether one customer pays, how many customers it takes, and how long each one needs to return what it cost.

One more warning about how the numbers are grouped. A single contribution figure across every customer is a cohortA set of customers grouped by when they arrived, so the ones who joined in April are followed separately from the ones who joined in September. question waiting to be asked. Buyers who joined three years ago may behave nothing like buyers who joined last month, and one average across both can conceal a unit that is quietly getting worse. Similarly, the way a shared cost is pushed down onto units through overhead absorptionA cost accounting step that spreads a shared cost across units of output using a chosen rule, so each unit ends up carrying a slice of something no single unit caused. can make a cost look variable when the underlying spending never moved at all. Both are reasons to ask how a per customer figure was built before arguing about what it means.

How to Analyse a Company's Unit Economics: which question comes first?

Somebody who does this for a living asks three questions and asks them in a fixed order, because each one is only worth asking once the answer to the one before it is in hand. Is the unit profitable? Then how many units cover the costs no unit moves? Then how long does each unit take to repay what it cost to win?

Different readers stop at different questions, and the reason is worth knowing. A lender cares most about the third, because a payback of three years against a loan due in eighteen months is a mismatch no contribution figure repairs. An equity analyst cares most about the second, since the gap between 50,000 and 62,500 buyers is the entire distance between a loss and a profit, and it is a distance somebody can be held to. A strategy team inside the business cares most about the first, because everything downstream is arithmetic once the unit is fixed, and no amount of growth repairs a unit that does not pay.

A household does the same thing without the vocabulary. Somebody deciding whether to buy a second sewing machine and hire a helper asks whether the helper produces more than they cost, which is question one. Then how many orders a month cover the rent on the extra room, which is question two. Then how long the machine takes to pay for itself, which is question three. The answers can easily be yes, forty orders, and four years, at which point the decision turns on whether the household can survive four years rather than on whether the helper is profitable.