Contribution Margin Is the Number That Decides Whether a Business Actually Works

The Question 717 Billion Dollars of Revenue Cannot Answer

Amazon posted revenue of $716.9 billion in 2025 up 12.4 percent from the previous year. It's one of the largest numbers in the history of commerce and by itself it says almost nothing about whether the business is healthy. Revenue measures size not quality. A company can grow revenue for a decade while losing money on each transaction and many have done so. The discipline that cuts through the illusion is unit economics: taking a single unit ofbusiness an order a customer a subscription a server instance and compare the revenue it generates with the costs it directly causes. What remains is the truth about whether making the business bigger creates value or destroys it

This is the lens that operations and finance teams apply to almost all decisions because aggregate financial statements are constructed to obscure per-unit reality. A company with record revenues and growing losses and a company with record revenues and growing profits can appear almost identical at the summary level. At the unit level they are opposites and the entire job of a business unit financial analyst is to see the difference before anyone else

Contribution Margin, Defined Properly

The critical number is contribution margin: revenue per unit minus variable costs that scale directly with each additional unit. For a retailer those variable costs are the cost of goods payment processing shipping and returns. For a software company they are hosting support and the marginal cost of servicing one more account. Contribution margin deliberately excludes fixed costs research and development headquarters and brand spend because they don't move when one more unit is sold. What's left is the money each unitcontributes to cover that fixed base and eventually to obtain profits

A Worked Example: One Order, Line by Line

Definitions only get you so far. Build the thing once and it's no longer abstract. Below is a single retail order with illustrative figures chosen to be more realistic than precise

The average order value is $35.00. Now subtract everything that would not exist if this order had not been placed

Two things come out of that table immediately and neither is visible in an income statement

First the business works. The contribution margin is positive at 4.79 which means that each additional order helps amortize the fixed base. Growth is more friend than foe. That is the most important fact about this business and it took six lines to establish it

Second look at operating leverage. Suppose the company regionalizes inventory so that packages travel shorter distances and shipping drops 30 cents from 4.20 to 3.90. Contribution margin increases from 4.79 to 5.09 a 6.3 percent improvement. Operating profit per order increases from 0.99 to 1.29 a 30 percent improvement

Thirty cents. A rounding error on a $35 order moved profits by almost a third because the savings falls entirely on the thin remainder at the bottom. Multiply that by billions of orders a year and you'll understand why a company can pursue enormous growth and at the same time behave as if cents were a moral issue. With enough repetition the number per unit is not a detail of the strategy. It is the strategy

The Tell Hidden Inside Amazon's 2025 Results

The clearest evidence of why this matters is found in data from Amazon's own segment. In 2025 Amazon Web Services generated $128.7 billion in revenue an increase of 19.7 percent and about $45 billion in operating income about a 35 percent margin and about 57 percent of Amazon's total $80 billion in operating income generated with less than a fifth of the company's revenue.company. At its side Amazon's advertising business a nearly pure-margin unit built into the retail platform grew 22 percent to $68.6 billion. Two units cloud and advertising account for a wildly disproportionate share of the company's profits

Cost to Serve: The Retail Unit

In the retail operation the relevant unit is the order and the governing discipline is the cost of serving the cost of picking packing shipping returns and customer service associated with a single package. Because that unit is repeated billions of times a year improving the contribution per order by even a few cents is worth more than almost any flashy initiative. This is why a company pursuing aggressive growth can at the same time be ruthless with cents: when the unit is repeatedat that scale number per unit is the strategy.Amazon's continued re-engineering of its fulfillment network regionalizing inventory automating warehouses and shortening delivery distances is in financial terms a multi-year campaign to increase contribution margin per order by fractions of a dollar

CAC, LTV, and the Payback Period

For recurring revenue companies two more numbers complete the picture. Customer acquisition cost or CAC is the sales and marketing spend required to gain a customer. Lifetime value or LTV is the total contribution margin the customer produces before churning. Analysts look at the relationship between LTV and CAC;A general rule of thumb is that durable businesses run above three times and the CAC payback period the number of months of contribution margin required to recoup the acquisition cost. A subscription like Amazon Prime is a textbook case: The membership fee barely covers its own shipping costs but the lifetime value comes from the repeat orders advertising exposure and ecosystem lock-in that each member generates for years. Considered as a standalone product Prime may seem marginal; Considered as an engine ofcustomer acquisition for the entire platform the unit economics are extraordinary

It's worth looking at the arithmetic once. Let's say a member places 28 orders a year with the 4.79 contribution from the table above which is about $134 and generates another $25 of advertising exposure across the platform. Let's say the net membership fee of the incremental shipping subsidy is negative 40.The annual contribution is 134 plus 25 minus 40 which is $119. Against a CAC of 60 the payback is 60 divided by 119 times 12 which is roughly equal to six months and a member who stays five years produces an LTV to CAC ratio close to ten times

Notice that the membership fee line is negative and doesn't matter at all. Viewed as a product the subscription loses money. Viewed as the unit that changes purchasing behavior it is one of the best acquisition investments in the history of retail. Choosing the wrong unit would have produced exactly the wrong answer

Case Study: MoviePass and Negative Contribution Margin at Scale

If you want to see what the legend above describes that scale digs the hole the fastest MoviePass conducted the experiment in public

In August 2017 MoviePass lowered its subscription price to $9.95 per month for what was effectively unlimited movie attendance one movie per day. The company paid theaters almost full retail price for most of those tickets roughly between $9 and $12 each depending on the market

Read it as a statement of unit economics. Revenue per subscriber per month was 9.95. The variable cost was about the price of a ticket. A subscriber who went to the movies twice a month produced a contribution margin of about negative ten dollars and an avid moviegoer who went weekly produced negative thirty or less

The contribution margin was structurally negative and prices had been set specifically to attract the most frequent users. Subscribers grew from about 20,000 to more than three million in about a year which at the time was considered spectacular growth and in fact was the fastest possible path to insolvency. Each new customer made the company worse

The end came as arithmetic demands. The parent company Helios and Matheson posted huge losses attempted a series of increasingly desperate restrictions on the service and MoviePass closed in September 2019. The parent company filed for bankruptcy in January 2020

What makes this the definitive case is that no operational excellence could have saved it. There was no efficiency to find no scaling curve to lower no fixed cost base to eventually cover. The unit itself lost money so the only strategy that improved the bottom line was to sell less. That's the distinction that the legend of contribution margin describes and MoviePass is what it sounds like with three million customers attached

Where Unit Economics Lie

The discipline has pitfalls and a careful analyst tests each one. The most important is allocation: which costs count as variable is partly a judgment call and a flattering definition can make a marginal unit look profitable on paper. Contribution margin also deliberately ignores fixed costs which are entirely real;A company can post great unit economics and still never sell enough volume to cover its fixed base. And combined averages are misleading when new and mature customers are mixed which is why a cohort view which tracks each group of customers separately over time almost always tells a more honest story than a single company-wide number. The analyst who reports a clean unit margin without questioning these assumptions is the one who is surprised later

Where the Framework Itself Breaks

Those are pitfalls in applying the tool. There are also situations where the tool is wrong and it's worth naming them because unit economics has become something of an article of faith

Amazon would have failed its own test. For years the retail business deliberately ran at or below breakeven per order and a strict assessment of unit economics in 2002 would have said it to stop growing. The judgment Amazon made was that the unit would improve with scale and that customer habit would worsen neither of which appears in any contribution margin table. Applied rigidly the framework kills exactly the companies that need patience

The first unit economies are not representative when network effects exist. The first thousand passengers on a ride-hailing platform have terrible unit economics because there are no drivers and the first thousand drivers have terrible unit economics because there are no passengers. Measuring the unit at that point measures the absence of the network rather than the quality of the business. It also means that every failing market can claim the same excuse which is why this argument is true and is endlessly abused

The framework has a systematic bias toward attributable costs. Companies whose value comes from branding research or a decade of accumulated engineering look worse on a per-unit table than companies whose costs are all in the shipping line because the framework eliminates exactly the spending that built the moat. A company can substantially improve its contribution margin by ceasing investment and the table will applaud

Choosing the drive is the analysis and is usually done in five seconds. Prime measured as a product loses money. Measured as an acquisition channel it is extraordinary. No one checks that choice and determines the answer more than any input to the model

My view is that unit economics is the best first question available and a bad last one and that the analysts who get into trouble are the ones who treat a positive contribution margin as a verdict rather than an initial condition

How I Actually Build a Unit Model

When I sit down with a business I haven't looked at before the sequence I follow is deliberately slow at first

I spend the first leg just choosing the unit and I usually write down two or three candidates and model the cheapest version of each. The order the customer and the cohort give different answers for a retailer and the disagreement between them is more informative than any of them alone

I then build the table from the example above with each line I can defend and put a question mark next to any costs I had to assign instead of tracking. If the contribution margin changes sign depending on how I treat a line with question marks I don't have an answer yet I assume it has a number on it

Next I run the sensitivity before running the base case conclusion. The thirty cent shipping example is the point: I want to know which input moves the bottom line the most because that input is the one the operational team should be working on and is almost never the one they are working on

Then I look for cohorts. A mixed number of a growing customer base are flattered by whichever group turns out to be the largest and separating them is where the real bad news usually lies

Finally I ask the MoviePass question out loud: If this business doubled tomorrow without any other changes would it be better or worse? If I can't answer that in one sentence of my own model the model is not finished

This is how I would approach it. It is a description of the method not investment advice about Amazon or anyone else

Why This Is the Core of Operating Finance

Unit economics is not an academic exercise it is the daily work of business unit finance. The task is to translate a strategic question (should we launch this expand into that region eliminate this program) into the right unit build a model around its drivers and tell the operator which lever actually moves the contribution? When a finance team is asked to quantify a new business idea or find a cost-reduction opportunity what they are really doing is finding and defending unit economics. Learn toThink in per-unit terms and the rest of corporate finance will stop being intimidating and start being readable: You can look at any business no matter how big its revenues and ask the only question that matters: Does the next unit make this better or worse?

The Bottom Line

Amazon's $716.9 billion in revenue in 2025 says nothing about whether the business works. The segment data does: AWS generated about $45 billion of the company's $80 billion in operating income on less than a fifth of the revenue and a combined margin of 11.2 percent is an average that ranks above two businesses that have almost nothing in common

Create an order and you can see all the discipline. Thirty-five dollars in revenue 4.79 in contribution 0.99 in operating profit and a savings of thirty cents on shipping that increases the profit per order by thirty percent. That's why cents are the strategy for scale

MoviePass is the other extreme. Nine ninety-five a month versus a ticket that costs between nine and twelve three million subscribers acquired and a bankruptcy filing because a negative contribution margin means that growth is the problem and not the solution. Ask any business no matter how big its revenue if the next unit will make it better or worse. Everything else is comments