Customer Acquisition Cost: What It Costs to Win One Buyer

What does this figure rest on before a single rupee is divided?

Three things sit underneath the figure. Two of them decide how much the answer can be trusted, and the third decides the limit of what can be done with it, so all three are worth having in view before any arithmetic starts.

The first is that an acquisition cost is an average of a spend over a count, and neither the spend nor the count is lying there waiting to be read off a statement. Somebody decided which rupees belonged to winning. Somebody decided what counted as a buyer won, and over how long. Both decisions were made before the division happened, and the figure carries them out into the world afterwards without mentioning either one. The silence is not a defect in the measure, and it is not grounds for distrusting it. The silence is simply the reason the first sensible question about anybody's acquisition cost is what went into it.

The second is a mismatch of shapes. A spend to win a buyer is paid once, in a lump, usually before that buyer has bought anything at all. The return arrives a slice at a time, year after year. A lump cannot be set against a stream in rupees without first deciding how many years to count, and how many years to count is a view about the future. Set them against each other in time instead and the answer needs nothing about next year at all. Rs 6,000/- paid once against Rs 2,000/- arriving every year is a duration, and a duration can be checked by anybody holding the same two numbers.

The third is a ruling that runs through the whole subject and is applied rather than argued again: what a buyer is worth across a whole lifetime is a valuation, and quality gets judged before any price is put on anything. The ruling does the heavy lifting when the most requested number in the entire subject, the ratio of acquisition cost to lifetime value, is named and turned down.

The business throughout is Setu Bazaar, and every rupee below belongs to it. Its trading year is published under business fundamentals and models. Setu Bazaar kept Rs 20,00,00,000/- of revenue, the slice it held back for itself as its take rateThe slice of an order a marketplace holds back for itself, everything else going to whoever supplied the goods. The slice is the marketplace's revenue; the goods themselves never belonged to it. on the orders placed across it. The marketplace served 50,000 buyers. Each of them left behind Rs 2,000/- of contributionThe part of what a buyer brought in across a year that is still there after every cost which moved because that buyer existed has been taken out. across the year. The costs that no single buyer moved came to Rs 12,50,00,000/- of fixed costCost that stays put whether one more buyer arrives or one fewer does. Rent, salaries and the machinery of running the place are the usual occupants., and the year closed at minus Rs 2,50,00,000/-. Winning one buyer cost Rs 6,000/-.

What is a customer acquisition cost, and why is it a division?

Take everything a business spent on winning buyers across a stated stretch of time. Divide it by the number of buyers it won in that same stretch. The answer is the acquisition cost, and for Setu Bazaar it is Rs 6,000/-. A division is the whole mechanism. A figure this heavily used tends to acquire an air of machinery it does not have, and the plain statement is the cure.

A snack stall outside an office gate runs the same division without ever using the phrase. The stall pays Rs 2,000/- to get handbills printed and pushed under windscreens in the week before a festival. By the end of the week it has ten people coming back who were not coming back before. Two thousand over ten is two hundred a head, worked out on the back of a bill book, and the stall holder knows exactly what it means. The stall holder has no second person quietly disagreeing about whether the boy who distributed the handbills counted as part of the spend, or whether the two who came twice and vanished count as regulars. Setu Bazaar has both of those people, and that is the only real difference.

The top of the fraction is the numeratorThe number sitting above the line in a fraction: the amount being divided up.. The spend sits on top. The bottom is the denominatorThe number sitting below the line: the count the top is being shared out across. Change it and the answer moves even when the top has not.. A headcount sits below. The figure is an average of a spend over a count, and nobody ever paid exactly Rs 6,000/- to win exactly one buyer. Some buyers walked in for nothing at all, having heard the name from somebody. Others took months of effort and cost several times the average. The Rs 6,000/- describes the pile, not any single person in it. An average height or an average bill describes its pile in the same way.

Why can two honest teams report different acquisition costs for the same business?

Because building this figure requires half a dozen decisions and none of them has a single right answer. Nobody has to lie, nobody has to slip a decimal, and nobody has to restate anything. Two competent people can start from the same set of statements, work carefully, and arrive at two different numbers.

Start at the top of the fraction. Does the salary of a staff member paid to win buyers belong in the spend, or does only money paid outside the business count? Does a discount handed to somebody making a first order count as spend to win them, or as revenue that was never earned in the first place? Does the cost of the team that answers a brand new buyer's opening question belong to winning that buyer, or to serving them? Every one of those is arguable in both directions by somebody sensible, and every one of them moves the answer.

Now the bottom. Almost nobody argues about the bottom, and almost everybody should. Is a person who bought once, in a single week, and was never seen again a buyer who was won? Over what stretch is the winning counted, given that money spent in one month goes on winning people who arrive in the next one? Stretch the period and slow arrivals fall inside it. Keep it tight and they fall outside, still paid for and not yet counted.

Notice what that table does not have in it, and notice it deliberately. A made up rival figure would be a number standing in for an argument. The direction each decision pushes is knowable and is printed above. The distance each one pushes depends on how many people that business employs on winning, how generous its first order discounts are and how many of its arrivals came back. Nobody standing outside the business can see any of that, so the distance is not knowable and cannot be.

The figure has no single right construction, so ask what went into it before comparing it with anything.

Where does the spend sit, and what must it never be subtracted from?

Three correct figures can be added wrongly, and the mistake produces a confident answer rather than an obvious error, so it is worth slowing down for.

The published year, rebuilt from its parts, shows every join. Fifty thousand buyers, each leaving behind Rs 2,000/- of contribution, is Rs 10,00,00,000/-. Against that stands Rs 12,50,00,000/- of cost that did not move whether any one buyer turned up or not. Ten crore less twelve and a half crore is minus two and a half crore, and minus Rs 2,50,00,000/- is exactly the published result.

Now the fact that makes the rest of the arithmetic behave. The money Setu Bazaar spends on winning buyers sits inside that Rs 12,50,00,000/-, and it is the largest single line in it. The winning spend is not a separate pot standing outside the accounts waiting to be deducted, and it is not inside the Rs 10,00,00,000/- of contribution either.

Contribution is measured per buyer served and acquisition is paid per buyer won, so they are counted on different heads and cannot be netted. The Rs 2,000/- describes somebody who was already there and bought things across the year. The Rs 6,000/- was paid for somebody who was not there yet. Buyers served and buyers won are two different populations counted on two different rules, and nettingSetting one amount against another and reporting only the difference, instead of reporting both. Netting is only legitimate when the two amounts are measured on the same thing. them against each other treats them as though they described the same people.

Here is the error in full. Seen once in full, it is recognisable the next time it arrives dressed up in a spreadsheet. A buyer contributes Rs 2,000/- a year. Winning that buyer cost Rs 6,000/-. Subtract, and each buyer apparently loses Rs 4,000/-. Multiplied across 50,000 buyers, that gives a loss of twenty crore. Twenty crore is eight times the loss the business actually reported and roughly the size of its entire revenue. The Rs 6,000/- has already been counted once, inside the Rs 12,50,00,000/- that produced the minus Rs 2,50,00,000/-, so the arithmetic is flawless and the answer is nonsense. Taken out a second time, it is a cost the business has paid twice.

A household does this correctly every day without thinking about it. The year's school fees came out of savings in June. Nobody deducts them again from July's salary and concludes that the household is short. The money left once. Stating where a cost sits is part of stating the cost.

How long does it take to get the Rs 6,000/- back?

Three years. Rs 6,000/- was paid once to win a buyer, that buyer leaves behind Rs 2,000/- of contribution every year, and six thousand over two thousand is 3.00. A cost paid once and a return that arrives annually can only be set against each other in time, and the answer is a duration rather than a value.

A duration is a genuinely useful thing to hold, and the reason is what it can be laid beside. Three years sits next to a loan term, a lease, a supplier contract, the length of a season, or the notice period on the office the business rents. All of those are lengths of time too, so the comparison needs no translation and no assumption. If the money that paid for winning those buyers was borrowed on two year terms, the mismatch is visible immediately and nobody had to forecast anything to see it.

The duration does carry one assumption of its own. The Rs 2,000/- is this year's contribution, and holding it flat across three years is an assumption, not a measurement. Contribution could rise as a buyer settles in or fall as prices tighten. The 3.00 years is what the published figures give if that one number stays where it is.

Whether the average buyer actually stays three years is a question about how many buyers remain each year. Staying runs on an entirely separate set of measurements, and it is weighed in the notes on customer loyalty. The payback answers a narrower question and answers it completely: how many years pass before the outlay is back, provided the buyer is still there.

Why is the cost of winning a buyer not divided by what a buyer is worth?

The ratio in question deserves to be named rather than quietly skipped: the acquisition cost to lifetime value ratio, the most commonly quoted number in this entire subject.

The reason is not that the ratio is difficult. The ratio is easy, and the ease is most of the problem. The reason is what each of the two measures requires before it can be built.

A retention rateThe share of the buyers present at the start of a year who are still buying at the end of it. The rate describes a period that has closed, and is often carried forward into ones that have not. tells how many buyers stay. Setu Bazaar's is published, and it is published as an assumption. A discount rateThe rate at which a rupee expected in some future year is written down to what it is worth today. A rupee later is not a rupee now. tells what a rupee arriving in year four is worth today, and nobody in this case ever measured one. A ratio is only as measured as its weakest term, and one of these two terms is a forecast wearing a decimal point.

Customer Acquisition vs Customer Retention: what does each rupee actually buy?

Set the two spends side by side and be precise about what is being compared. The two get spoken of as if they were alternatives on a menu, and they are not. Money paid to win buys somebody who has never bought anything from this business. Money paid to keep buys another year from somebody who already has. Winning buys a buyer who is not there yet, and keeping buys another year from one who is, so the two are not interchangeable rupees.

One figure makes that comparison real rather than abstract. Setu Bazaar loses 10,000 buyers a year, and replacing them at Rs 6,000/- each costs Rs 6,00,00,000/-. Against Rs 20,00,00,000/- of revenue, that is 30.00 per cent of everything the marketplace earns, spent every year on standing still. Replacement is the largest single line inside the Rs 12,50,00,000/- of cost that did not move with any one buyer. Whatever else is arguable about the construction of an acquisition cost, that is a very large number to be spending on replacing what the business already had.

Winning can be priced for Setu Bazaar and keeping cannot, so the two cannot be ranked. Nothing published anywhere in this case says what Setu Bazaar spends to hold on to a buyer. There is no figure to set against the Rs 6,00,00,000/-, and inventing one to complete a tidy comparison would be worse than leaving the space blank. A comparison with one side missing is not a comparison, and saying so is the answer rather than a gap in it.

The missing side matters because the single most repeated sentence in this whole subject is that keeping a buyer is cheaper than winning one. The sentence may well be true of many businesses. Nothing in this case measures it. The notes on customer loyalty measure what keeping a buyer is actually worth, how many leave, and what standing still costs, and the questions belong there.

What should be asked before using somebody else's acquisition cost?

Four questions, in this order, and they are the whole of the practitioner's method. An analyst reading two sets of accounts, a lender sizing up how much of a borrower's cost base is discretionary, or an operator holding a number somebody else built all need the same four answers before the figure is safe to touch.

The figure is almost never wrong and the comparison almost always is. The rule is worth carrying away above everything else. Nobody miscalculates a division. People set one carefully built figure beside another carefully built figure and forget that the two were built to different rules.

A useful contrast sits elsewhere in this case. Anjani Stationers Private Limited, an invented stationer, wins a school by somebody getting in a van and going to see it, and no acquisition cost has ever been computed for the business at all. The absence is not a failing. Most businesses look like Anjani Stationers, and the absence is a reminder that the figure is an instrument for a particular shape of business rather than a universal disclosure.

What goes wrong when two of these figures are set side by side?