Contribution Margin

How do you calculate contribution margin?#

1. Name the unit and the decision. Order, transaction, customer-month, location-year, deployment, cohort, or channel. A percentage without a decision unit is an invitation to classify costs however flatters you.
2. Fix the revenue basis so it matches the cost period. If the decision is a cash decision, keep a separate cash-contribution schedule — recognized revenue and cash collection diverge, sometimes by a full contract year.
3. Stack the cost layers as a waterfall, so each subtotal is inspectable:

  revenue
− cost of revenue                                  → gross profit
− variable selling, commissions, and partner share
− variable customer operations and success
− transaction-specific refunds, fraud, and incentives
                                                   → contribution profit

4. Show the excluded costs immediately below the subtotal. Step-fixed capacity, R&D, G&A, financing, and tax should be visible, not implied. A waterfall that ends at contribution profit with nothing beneath it is a marketing document.
5. Separate variable, step-fixed, and sunk. Variable moves with the unit. Step-fixed moves after a threshold. Sunk is already spent and is irrelevant to a forward marginal decision — though it still matters to whether the business as a whole works.
6. Stress the boundary at scale. Test discounting, supplier price changes, support intensity, fraud, returns, utilization, and the next capacity block. Historical average contribution is not next-unit contribution.

Worked example: the cohort that stops contributing#

A marketplace cohort generates $200,000 of monthly platform revenue.

Gross profit

$200,000 − ($20,000 + $80,000) = $100,000
gross margin = $100,000 / $200,000 = 50.0%

Contribution profit

$200,000 − $100,000 − $30,000 − $10,000 = $60,000
contribution margin = $60,000 / $200,000 = 30.0%

Both are correct under their stated boundaries, and the 20-point gap is the point of the exercise. Half of every dollar survives delivery; only three-tenths survives the cost of getting and keeping the customer.

Now add a step-fixed cost#

Volume growth requires a second support pod costing $50,000 per month, triggered once cohort revenue exceeds $240,000. Assume the pre-step cost ratios are unchanged.

Growing from $240,000 to $250,000 added $10,000 of revenue and destroyed $47,000 of contribution profit. The extra $50,000 of revenue that management expected to yield $15,000 of contribution ($50,000 × 30%) instead tripped a capacity block.

How much revenue refills the pod?

incremental revenue required = $50,000 / 30% = $166,667
revenue at which contribution returns to the pre-step $72,000
    = ($72,000 + $50,000) / 30% = $406,667

So the cohort must reach roughly $406,667 of monthly revenue — about 70% more than the $240,000 trigger point — before contribution dollars are merely back where they started.

Two caveats the arithmetic hides. First, the marginal contribution margin on that incremental revenue is still 30%; it is the average that collapses and then recovers. Reporting the average around a capacity step will mislead anyone who does not know the step is there. Second, "the pre-step cost ratios are unchanged" is a strong assumption — the volume that triggered a new pod usually also changes support intensity, discounting, and mix. Model the step and the ratio change together, or the answer is precise and wrong.

What are the common mistakes?#

• Publishing a percentage with no boundary. Contribution margin without the unit, horizon, and included-cost list is not a metric; it is a claim.
• Calling it gross margin, or comparing it to someone else's. Different boundary, different question, no comparability. Upstart's 56% and a marketplace's 30% are not on the same scale.
• Treating salaried labor as permanently fixed. Over a 12-month horizon, support and success headcount is step-fixed and economically incremental. "Fixed" is a statement about the horizon, not the cost.
• Ignoring costs outside the immediate transaction. Refunds, downstream support, partner revenue share, and retention work are often required to preserve the revenue you just counted.
• Excluding normal recurring cash expenses to make the number look better. This is the specific behavior the SEC's non-GAAP interpretations target, and it is also how founders talk themselves into scaling a losing motion.

When does contribution margin break?#

When attribution requires arbitrary allocation. If you cannot trace a cost to the unit without inventing an allocation key, including it makes the number less reliable, not more. Show it as a separate layer instead of burying it in the subtotal.

When behavior changes with scale. The assumption that cost ratios hold is exactly what fails at the moments you most need the metric — during a discount push, a new channel, or a capacity expansion.

When the window is too short. Refunds, chargebacks, support tails, and churn all arrive after the sale. A one-month contribution figure for a business with a 90-day return cycle is a forecast, not a measurement.

When it is mistaken for profitability. Product, engineering, G&A, financing, taxes, and capacity investment still need funding. Upstart's own disclosure makes this limitation explicit, and the 2025 gap between a 56% contribution margin and $53.6 million of net income shows why it matters.

When you need one number. You will not get one. Build several labelled layers — order contribution, cohort contribution, channel contribution — rather than hunting for a single universal percentage.