CM1 vs CM2 vs CM3 vs CM4: Contribution Margin Breakdown, Unit Economics, and E-commerce Profitability

Why Split Margin Into Layers At All

A single profit figure tells you whether a business is working. It does not tell you where it is working or where it is failing, and those are the questions that determine what to do next.

Consider two businesses with identical overall losses. In the first, the product costs more to make than customers pay for it — every order deepens the hole, and no amount of operational efficiency or marketing discipline fixes it. In the second, the product is profitable but customer acquisition costs exceed what customers are worth. These require completely different responses: the first needs repricing or a supply chain change, the second needs a marketing efficiency programme or a retention improvement. A consolidated profit and loss statement cannot distinguish them; contribution margin layers can.

The layering works because costs enter at different points in the value chain, and each point is owned by a different function. Product costs sit with sourcing and pricing. Fulfilment sits with operations. Acquisition sits with marketing. Support and platform sit with the wider business. Isolating the margin impact at each stage tells you which function owns the problem, which is what makes the framework operationally useful rather than merely descriptive.

This is also why the framework became standard in marketplaces, direct-to-consumer commerce and delivery businesses. In all three, an order can pass through several cost layers that are individually small and collectively decisive, and where a leadership team can genuinely disagree about which layer is the problem. Putting numbers on each layer converts that disagreement into a measurement.

The Definitions Are Yours, Which Is the Danger

Contribution margin layers are management metrics, not accounting standards. No regulator defines them and no auditor checks them. That flexibility is useful and it is the source of nearly every problem teams have with these numbers.

The practical consequence is that CM1 at one company is CM2 at another, and CM3 in a board deck may not be the CM3 in the finance model. When someone benchmarks against a competitor's reported contribution margin, they are almost certainly comparing two differently constructed figures. The comparison feels rigorous and is not.

Internally, the failure is slower and more damaging. Without a written definition, each layer accumulates interpretation. Someone includes payment processing in CM1 because it feels like a product cost; someone else puts it in CM2 because it feels transactional. Both are defensible, and once both exist in different reports the numbers stop being comparable across time as well as across teams.

The fix is unglamorous and effective: write down exactly which cost lines belong in which layer, publish it, and treat changes to it as a versioned event rather than a quiet correction. When a definition does change — and it will, as the business adds cost lines — restate history so the trend remains readable. A margin series where the definition shifted silently in month seven is worse than no series, because it invites conclusions about a change in the business that was really a change in the maths.

One specific discipline worth adopting: keep discounts and returns out of the layers entirely by handling them in net revenue. Both are frequently smuggled into a lower layer where they inflate the numbers above, and both are large enough in most consumer businesses to change the conclusion. If CM1 looks strong while discounting is heavy, that is usually where to look first.

Where to Cut the Numbers

A company-level contribution margin is an average, and averages in commerce hide more than they reveal because the underlying economics vary enormously across a catalogue and a customer base.

By category or SKU is the first and most valuable cut. It is common for a minority of products to carry the entire margin while a long tail runs at or below break-even, sustained by an assumption that they drive traffic or basket size. That assumption is testable and frequently false. Products that lose money and do not measurably increase basket value are simply losing money.

By acquisition channel exposes something the marketing dashboard cannot. Channels differ not only in cost per customer but in what those customers buy, how much they discount, and how often they return items. A channel with attractive acquisition cost that delivers customers concentrated in low-margin categories with high return rates can be the worst channel in the portfolio at CM3, while looking excellent on cost per order.

By geography or delivery zone matters wherever fulfilment cost varies with distance. Businesses that price uniformly across a wide service area are cross-subsidising remote customers with nearby ones, which is a legitimate strategic choice and a poor accidental one. Seeing it in the numbers converts it into a decision.

By order value band reveals fixed-per-order costs. Delivery, packaging and payment processing do not scale down with basket size, so small orders can be structurally unprofitable while the blended figure looks acceptable. This is the analysis behind minimum order values and tiered delivery charges, and it is straightforward to run once the layers exist.

The practical recommendation is to build the contribution margin model at the order line level with these dimensions attached, rather than as a summary calculation. Aggregating up is trivial; disaggregating a summary is impossible, and the questions that matter almost always require a cut nobody anticipated.

What Each Layer Tells You to Do

The value of the framework is diagnostic. Each layer, read against the one before it, points at a specific set of interventions.

A weak CM1 means the product economics do not work. The available responses are raising price, reducing product cost through sourcing or specification, or changing mix toward better-margin items. Nothing downstream can rescue it — a business with negative CM1 becomes less viable as it grows, which is the most dangerous position on this list because growth feels like progress.

A large drop from CM1 to CM2 points at operations: fulfilment cost per order, packaging, delivery, warehousing. These are frequently the most tractable costs in the stack because they respond to process change, better carrier terms, order consolidation and packaging redesign. A business losing most of its product margin between CM1 and CM2 has an operational problem masquerading as a pricing problem.

A large drop from CM2 to CM3 points at acquisition. This is the most common shape in venture-funded consumer businesses, and it is where the framework earns its reputation. The responses are improving marketing efficiency, shifting mix toward cheaper channels, or improving repeat rate so acquisition cost amortises across more orders. Note the third option carefully — it does not reduce CM3 on the first order at all, which is why first-order CM3 is a misleading metric for any business with genuine repeat purchase.

A weak CM4 with healthy CM3 points at overhead: support cost per order, platform and payment infrastructure, allocated headcount. These are usually the least responsive to short-term action and the most responsive to scale, which is why CM4 tends to improve with volume in a way the earlier layers do not.

The Errors That Make Contribution Margin Look Better Than It Is

Contribution margin figures tend to drift optimistic over time, and they do so through a small number of recurring mistakes rather than through deliberate misrepresentation. Each one is easy to make and difficult to spot afterwards.

Calculating on gross rather than net revenue. Discounts, promotional codes, cancellations and returns all reduce what the business actually received. Starting the calculation from list price rather than realised price inflates every layer beneath it, and in a promotion-heavy category the difference is substantial rather than marginal.

Excluding costs that genuinely vary. Payment processing fees, packaging materials, transaction taxes, marketplace commissions and reverse logistics all scale with volume and all belong in the variable stack. They are individually small enough to be overlooked and collectively large enough to change the conclusion. The test is simple: if selling one more unit increases the cost, it is variable.

Using blended averages across a mixed catalogue. A single contribution margin figure across a product range with genuinely different economics tells you very little and can conceal that a substantial share of volume is loss-making. A blended CM2 of a healthy-looking figure can easily contain one category comfortably positive and another deeply negative. Segment by category, by price band and by channel before believing an aggregate.

Allocating fixed costs into variable layers. The reverse error, and it makes margins look worse rather than better, but it distorts decisions just as effectively. Warehouse rent does not vary with the next order; warehouse labour partly does. Mixing them means the model tells you that reducing volume improves unit economics, which is true arithmetically and false operationally.

A useful periodic check is to reconcile the sum of contribution margin across all orders in a month against the corresponding lines in the profit and loss statement. They will not match exactly, because the layers exclude fixed costs by design, but the variable components should tie. Where they do not, something is being counted in one place and not the other, and finding it is usually a day's work that prevents a quarter of misdirected decisions.

Contribution Margin Across a Customer Lifetime

The single most common analytical error with these metrics is calculating them on an order basis and drawing conclusions about a customer basis. For any business with repeat purchase, the two diverge sharply.

Acquisition cost is incurred once, on the first order. Every subsequent order from that customer carries no acquisition cost at all. A per-order CM3 that spreads acquisition across all orders equally understates the first and overstates the rest; a per-order CM3 that loads all acquisition onto the first order makes the first order look catastrophic and every later one look excellent. Neither is wrong, and neither answers the question that matters.

The question that matters is cumulative contribution margin per customer cohort against the cost to acquire that cohort. Plot it by month since acquisition and you get a payback curve, which tells you two things a single-period figure cannot: whether the business ever recovers its acquisition cost, and how long that takes. A business with negative first-order CM3 and a twelve-week payback is healthy; one with the same first-order figure and no repeat purchase is not, and they are indistinguishable at the order level.

This connects the framework to cash rather than just profitability. Payback period determines how much working capital growth consumes, which is frequently the binding constraint on how fast a business can scale regardless of how good the unit economics eventually look. Two businesses with identical mature contribution margin and different payback periods have materially different funding requirements.

A final caution on cohort work: use contribution margin rather than revenue on the return side. A payback curve built on gross revenue crosses the acquisition cost line far earlier than one built on contribution margin, and only the second represents money the business can actually keep. The revenue version is a considerably more encouraging chart and it is not the one to run the business on.