Fixed Costs vs Variable Costs: The Test Is Volume

A cost is variable if it moves with volume and fixed if it does not. Volume is the whole test. The test is about behaviour, not about what the cost buys or how large it is. The distinction decides what happens when volume moves: variable costs follow it down, fixed costs stay exactly where they were. The distinction holds only over a range.

Both of these have already been met separately. A variable cost turned up when revenue was pulled apart into what a business keeps out of one sale, and a fixed cost turned up on the other side of that same subtraction, sitting underneath and waiting to be covered. Neither idea is new here. The line between them is new, and it is thinner and more argued over than it looks. The line also has a specific job.

The job is this. Two businesses can spend the identical amount of money in a year and be in completely different positions the moment volume moves, and nothing on the face of their accounts will say so. The split into fixed and variable is what tells them apart, and it is almost never printed. The split has to be built from scratch, out of a test that takes one question.

What is a variable cost, and what is a fixed cost?

Take them one at a time, fully, before putting them next to each other. A variable cost is a cost whose total rises and falls with the number of units the business makes or serves. Each unit brings its own share of that cost with it. The paper in a notebook is the clean case. Four thousand notebooks take paper for four thousand. Two thousand take paper for two thousand. The cost per notebook stays roughly where it was and the total moves. Packing material, delivery charges paid per parcel, a commission paid per sale and the cost of revenueThe costs a business treats as directly attached to delivering what it sold. Which costs belong inside that line is set out under gross margin. of a marketplace order all behave the same way.

A fixed cost was committed before the volume was known, so its total stays where it is over a period regardless of how many units the business makes or serves. The rent on a workshop is the clean case here. Once a lease is signed for a year, the landlord is owed the same amount whether the workshop runs one shift or three, and whether everything it makes is sold or none of it. A supervisor's salary, an insurance premium, an audit fee, the depreciationThe slice of an asset's cost charged against one year's profit, spreading a purchase made once across the years the asset gets used. on a machine bought two years ago: all of them arrive at the same size whatever the month looked like.

Notice what has just been described and what has not. Neither definition mentions the amount. Neither mentions what the money buys. Neither mentions whether it is paid monthly or once a year. Both definitions describe how the total responds to a change in volume, and nothing else. The omission is deliberate. Every one of those other properties feels like it ought to matter, and that feeling is where most of the trouble starts.

What is the test, and what is not the test?

The test is one question with one variable in it: if the number of units changed, would the total on this invoice change with it? A yes makes the cost variable. A no makes it fixed. There is nothing else to apply and no second criterion hiding behind it. Every difficulty below comes out of that single question, including the parts where it gets uncomfortable.

Now the three things that are not the test. Each is a real reading that people make, and each gives the wrong answer. Size is not the test. A variable cost can be the largest line in the accounts and a fixed cost can be a rounding item, and the reverse is just as common. Purpose is not the test. Two costs can both be spent on people, or both on machinery, and sit on opposite sides. An overheadAn older accounting word for costs not traced to a single unit of output, such as rent, supervision and insurance. Traceability is not behaviour. is not automatically fixed, because that word describes whether a cost can be traced to one unit, which is a different question from whether it moves. How often it is paid is not the test. A monthly bill is not more variable than an annual one; a lease paid every month is as fixed as a lease paid once.

Anjani Stationers Private Limited, an invented notebook manufacturer, and Setu Bazaar, an invented marketplace, between them occupy all four cells that come from crossing those two questions. A two-way test crossed with a two-way test gives four cells, not three. Once somebody has found a large fixed cost and a small fixed cost and a large variable cost and a small variable cost, the idea that size decides anything is gone for good.

What happens to each kind of cost when volume moves?

Here is what the whole separation is actually for, and it is the only reason anybody bothers to build it. Setu Bazaar serves 50,000 buyers a year. The marketplace spends Rs 10,00,00,000/- that moves with the buyer count, at Rs 2,000/- a buyer, and Rs 12,50,00,000/- that does not move at all. Now halve the buyers to 25,000 and follow each line separately.

25,000 buyers at Rs 2,000/- each is Rs 5,00,00,000/-, so the variable cost halves to Rs 5,00,00,000/-. The fixed cost does not move by a rupee. Nothing in it was ever attached to a buyer, so it stands at Rs 12,50,00,000/-, exactly as it was. Total spending falls from Rs 22,50,00,000/- to Rs 17,50,00,000/-. The fall of Rs 5,00,00,000/- against 25,000 fewer buyers came entirely out of one of the two lines and not at all out of the other.

The whole of operating risk lives in that one fact. A business whose costs are mostly variable shrinks when its volume shrinks and largely survives the experience. A business whose costs are mostly fixed keeps paying while the money coming in falls away. A half-empty aircraft, a half-empty cinema and a half-empty hotel are all uncomfortable things to be running for exactly that reason. The same arithmetic is visible at household scale. If a household earning one salary loses half of it, the groceries bill can be halved by next week. The rent cannot, the school fee cannot, and the loan instalment certainly cannot. Half the income met an entirely unchanged set of obligations, and that is the same arithmetic in a smaller room.

Why is a per-unit fixed cost misleading?

The per-buyer fixed cost is the sharpest point in the whole subject, and it is worth slowing down for. Setu Bazaar's Rs 12,50,00,000/- of fixed cost divided by its 50,000 buyers gives Rs 2,500/- a buyer. The Rs 2,500/- is arithmetically correct. The figure is also printed in management packs everywhere, sits in a column next to the Rs 2,000/- of variable cost per buyer, and looks exactly like it belongs there. It does not.

A per-unit fixed cost is the fixed total divided by however many units there happen to be, and it moves whenever they do. Watch it happen. At 50,000 buyers the fixed cost is Rs 2,500/- a buyer. At 25,000 buyers that identical, untouched Rs 12,50,00,000/- reads as Rs 5,000/- a buyer, double the earlier figure, and not one rupee of the cost itself has been disturbed. No contract was renegotiated, no salary rose, no machine was bought. The only thing that moved was the number underneath the division line.

Put the two per-buyer figures side by side and the difference in kind is obvious. The Rs 2,000/- is a property of a buyer: it exists because that buyer exists, it arrived with them and it leaves with them. The Rs 2,500/- was never a property of a buyer at all, and calling it a cost per buyer gives it a status the arithmetic never granted it. The Rs 2,500/- is an average of something that has nothing to do with buyers, wearing a per-buyer label because a division was performed.

What is a step cost, and which of the two is it?

Now the question the test cannot answer on its own, and the honest thing is to meet it head on rather than pretend the two boxes are enough. A great many real costs sit flat across a stretch of volume and then jump. One warehouse handles orders up to a point and then a second warehouse is needed. One shift covers output up to a point and then a second shift opens. One supervisor manages a floor up to a point and then another is hired. Between the jumps the cost behaves like rent. At the jump it behaves like nothing rent ever does.

A step cost is fixed within its range and variable across ranges, and both halves of that sentence are needed. Ask the volume test at 45,000 buyers and the answer is a confident no, this does not move. Ask it again at 61,000 buyers and the answer is that it moved by Rs 5,00,00,000/- in one jump. Both answers are correct, and they are answers to the same question asked at different places. An account that forces every cost into one of two boxes has told a tidy lie, and it is a lie that shows up as a forecast that was right until the month the business grew.

The step is a capacityThe most a business can make or serve with the people, machines and space it already has standing. How capacity gets planned and added is covered under operating model and supply chain. boundary, which is why the jumps are usually large rather than gentle: nobody can rent a third of a warehouse or hire a quarter of a supervisor. The capacity boundary also explains a pattern worth recognising. A business running at low utilisationHow much of what a business already pays for is actually in use, usually written as a share of capacity. can grow for a while at almost no extra fixed cost, and then absorbs a whole step at once.

Over what range does the distinction hold?

Step costs lead straight to the qualification that makes this whole subject honest instead of merely tidy. Every statement that a cost is fixed carries an unstated range with it, and the statement is only as good as the range. Setu Bazaar has actually operated between roughly 40,000 and 60,000 buyers, and the arrangement it signed applies across that stretch, so its Rs 12,50,00,000/- is fixed there. Calling it fixed at 5,000 buyers instead describes a business that would have shut most of what it pays for. Calling it fixed at 5,00,000 buyers describes a business that would have taken on several more of everything.

Fixed means fixed over a stated range and means nothing without one. This is not a caveat added at the end for safety. The range is part of the claim, in the same way that a speed is meaningless without a direction. A cost line that is genuinely flat across the volumes anybody has seen is evidence about those volumes and about nothing else, and the moment a forecast leaves that stretch it has left the evidence behind and is running on assumption.

The practical version of this is a discipline that applies to any model handed over. Ask for the highest and lowest volume the business has actually run at, then ask how the model behaves outside that pair. If a fixed cost line marches flat through a volume three times the largest the business has ever handled, somebody has extended an observation into territory where it was never tested, and the number that looks most solid on the sheet is the one carrying the most assumption.

Is either kind of cost the better one to have?

No. The temptation to pick a winner here is strong in both directions, so the reason is worth working through rather than asserting. Take Setu Bazaar as it stands, spending Rs 2,000/- a buyer that moves and Rs 12,50,00,000/- that does not, and set it beside an invented alternative arrangement that spends Rs 2,500/- a buyer and carries Rs 10,00,00,000/- of fixed cost. At 50,000 buyers the two spend exactly the same Rs 22,50,00,000/-, and against revenue of Rs 4,000/- a buyer both land on the identical result of minus Rs 2,50,00,000/-. Indistinguishable at that one volume.

Move the volume and they separate immediately. At 25,000 buyers the heavier fixed block of the arrangement as it stands cannot shrink, so that arrangement is Rs 1,25,00,000/- worse off. At 75,000 buyers its lighter variable cost lets more of every extra rupee of revenue stay in the business, so it is Rs 1,25,00,000/- better off. The same structure that protects on the way down limits on the way up. Neither arrangement is flawed. The trade has that shape.

So the honest statement is a comparison on a stated criterion and a refusal to rank overall. On resilience to a fall in volume, more variable is better. On what a business keeps from a rise in volume, more fixed is better. Which of those two criteria matters more depends on what volume is going to do, and that is a fact about the future that no cost structure contains. An account that named one arrangement as the one to prefer would be smuggling in a forecast and calling it arithmetic. There is one more thing worth noticing and worth labelling correctly: the two arrangements have different break-even counts, 62,500 buyers for the first and about 66,667 for the second. Neither count is an independent confirmation of anything. Both are the same two cost equations rearranged, so they could never have disagreed with the chart, and what they add is a single number for where each line crosses zero.

What question sorts a cost faster than any list of categories?

Practitioners do not carry a taxonomy around. Practitioners carry one question and apply it to whatever invoice is in front of them: if volume halved next month, would this invoice be smaller? Yes puts the cost on the variable side. No puts it on the fixed side. One answer turns up more often than either, and a two-box list has no room for it: not yet. A step cost announces itself in precisely that way.

How do the two businesses look once every component is printed?

Setu Bazaar first, at both volumes, with every component printed so either per-buyer figure can be rebuilt from scratch. Buyers pay Setu Bazaar across a gross flowThe full value of goods changing hands across a marketplace, which is not the marketplace's own revenue. of Rs 5,00,00,00,000/- and the marketplace keeps a take rateThe slice of the money crossing a marketplace that the marketplace keeps, written as a percentage of that flow. of 4.00 per cent, giving revenue of Rs 20,00,00,000/- from 50,000 buyers, or Rs 4,000/- each. Its contributionWhat one sale leaves behind after the costs that exist only because that sale happened. is Rs 2,000/- a buyer, so its variable cost is the other Rs 2,000/-.

The third row and the fourth row read against each other one more time carry the entire subject in two lines. The variable figure a buyer is Rs 2,000/- in both columns and the total behind it halved. The fixed total is Rs 12,50,00,000/- in both columns and the figure a buyer doubled. One number describes a buyer and the other describes an arithmetic operation.

Now Anjani Stationers Private Limited, whose figures are carried forward unchanged from where its statements were built. Revenue for the year was Rs 2,70,00,000/- and operating profitWhat is left of revenue after every operating cost and before interest and tax. Also written as EBIT. was Rs 41,50,000/-, so Rs 2,28,50,000/- was spent. Here are the four lines it was spent on, and the volume test applied to each.

The two businesses carry the distinction in opposite proportions. A notebook needs paper and paper is bought a sheet at a time, so Anjani Stationers spends 64.99 per cent of its money on things that move with volume. A marketplace mostly pays for technology and support that sits there whether an order arrives or not, so Setu Bazaar spends Rs 10,00,00,000/- of Rs 22,50,00,000/-, or 44.44 per cent. Neither proportion is a verdict on either business; it is a description of what each of them would still be paying if next year were half the size.

One honest note about that Anjani table. A tidy classification is exactly the trap here. Sorting an employee benefits expense wholly into the fixed column is a teaching treatment rather than a fact. A factory that runs overtime in a busy month pays more in that month, and a workshop that keeps its staff through a quiet one pays the same. The four lines are printed separately above rather than collapsed into two for that reason. A reader who disagrees with where a line was put can move it and rebuild the total, and that is the only defence against a classification nobody can check.

What goes wrong when a per-buyer fixed cost enters a decision about one more buyer?