From spend to margin: the chain
Every paid channel runs the same sequence. Spend buys impressions; some impressions become clicks; some clicks become results at a conversion rate (CVRShare of visits or clicks that convert.); each result carries revenue; revenue less the cost of goods, fulfilment, payment fees and returns leaves contribution margin; and contribution margin less the spend that produced it is what the programme actually earned. The KPI tree framework is the general form of this decomposition. Applied to paid acquisition it shows that cost per acquisition (CPACost of each conversion.) and return on ad spend (ROASRevenue generated per unit of ad spend.) are summaries of the chain rather than independent levers.
| Stage | What it measures | Who mostly controls it | What moves it |
|---|---|---|---|
| Spend to impressions | Price of attention | The auction | Competition, audience, placement, season |
| Impressions to clicks | Whether the ad earns a visit | Creative and targeting | Message, format, relevance |
| Clicks to results | Whether the visit converts | Landing page and offer | Price, proof, friction, speed |
| Results to revenue | Value of each result | Merchandising and sales | Basket size, lead quality, close rate |
| Revenue to margin | What the business keeps | Operations and finance | Cost of goods, returns, fees, discounts |
| Margin less spend | What the programme earned | Everyone above | All of the above |
Reading the chain this way changes what an underperforming campaign means. A rising CPA can come from a more expensive auction, weaker creative, a slower page or a worse offer, and each has a different owner. A ROAS that looks healthy can still lose money if the products it sells carry thin margins or high return rates. The chain also explains why the platform cannot set the target for you: it sees the first three stages and nothing of the last three. On Meta, the first stage is set by an auction that charges mostly per impression, and the guide to what drives the cost of Meta ads follows the chain from cost per thousand impressions to cost per result.
Targets from unit economics, not from platforms
The target begins at the bottom of the chain. Take what one result is worth in contribution margin, decide how much of that margin the business is prepared to spend acquiring it, and the target CPA follows. For revenue targets, divide the order value by the margin per order to find the break-even ROAS: the return at which spend exactly consumes the margin. Below it every sale loses money; above it each sale contributes something. The Break-even ROASMinimum ROAS to avoid a loss. is the floor, not the target, because attribution error, refunds and overheads the calculation ignored all eat into the gap.
Lead generation adds one step. A lead is not revenue, so its worth is the value of a closed deal multiplied by the share of leads that close, adjusted for the time and cost of the sales effort. That is why a cost per lead target set without the sales team's close rate is a guess, and why lead targets are best set per lead type rather than per form.
Platforms will offer a target of their own. Google's and Meta's help centres document bidding strategies that aim at a cost or return the advertiser supplies, and the interfaces often propose one based on recent performance. A recent average describes what happened; it says nothing about what the business can afford. Use the break-even calculator, the CPA calculator and the ROAS calculator to derive the number from margin, then hand that number to the platform.
Blended vs channel vs campaign views
There are three ways to look at acquisition cost, and they answer different questions. The blended view divides all acquisition spend by all new customers recorded in the business's own systems. That is customer acquisition cost (CACCost to acquire one new customer.), it has no attribution overlap, and it is the figure that has to work for the business to be healthy. Reported to a board, the blended figure sits beside gross margin and contribution after marketing in a marketing P&L, which shows whether it does. The channel view divides each channel's spend by the results that channel's platform or the analytics tool credits to it. The campaign view does the same one level down.
Channel and campaign figures are where decisions are made, because they are the only views granular enough to act on. They are also systematically optimistic. Each platform counts conversions it touched under its own attribution window, and several platforms touch the same customer, so the sum of channel-reported results is nearly always larger than the number of customers who arrived. The practical answer is to set the blended target from the economics, set channel targets that are tighter than the blended one by the amount of overlap you have observed, and reconcile the two every month. When channel figures all look on target and the blended figure does not, the channels are over-claiming, and the next step is a holdout test rather than another bid change; the iROAS versus ROAS comparison explains how a test result and a platform figure divide the work between budgeting and bidding. The budget allocation guide in the strategy section covers how those views feed the annual split across all marketing.
The comparison of CPA and CAC sets the campaign figure and the business figure side by side and shows how to bridge them each month.
Average vs marginal returns
An average ROAS tells you how the whole budget performed. It does not tell you what the next unit of budget will buy, and that is the only question a budget decision asks. Returns in paid media diminish: the first spend in any campaign reaches the most responsive people at the lowest cost, and each further increment reaches people who are less likely to act, at higher prices, with more frequency on the same individuals. The average falls slowly while the marginal returnExtra revenue returned by the last increment of ad spend. falls fast.
The consequence catches out most teams at least once. A campaign averaging comfortably above break-even can have its last tranche of spend running well below it, so cutting that tranche would raise profit without visibly hurting the average. The reverse is equally true: a campaign with a mediocre average may still have cheap incremental results available because it is under-funded. Average figures should therefore never decide a budget change on their own.
Estimating the marginal return does not need a model. Compare periods where spend changed and little else did, run a deliberate budget step on a stable campaign and watch the incremental results, or use the platform's own budget simulators as a rough guide while treating their output as a hypothesis. The test that settles the question is in the testing and scaling guide, which uses the marginal cost as its main scaling guardrail.
Payback and cash constraints
A target can be economically correct and still unaffordable. If a customer is worth more over a year than on the first order, the business may rationally accept a first-order CPA above break-evenHighest cost per acquisition at which a sale still covers its variable costs., recovering the difference from repeat purchases. The lifetime value (LTVTotal value a customer brings over time.) that justifies this belongs to the growth economics section, and the LTV and CAC guide covers how to estimate it without flattering the result. What belongs here is the cash consequence: every customer acquired above first-order break-even is a loan the business makes to itself, repaid only as the customer returns.
Payback period is the time it takes for cumulative contribution margin from a cohort to cover what was spent acquiring it. It decides how fast a business can grow on its own cash. A short payback lets spend recycle quickly; a long one means growth consumes cash faster than it returns it, and the limit becomes the bank balance rather than the auction. Three rules keep this honest:
- Use observed retention. Base the payback on cohorts you have actually measured, by channel where possible, because customers acquired through heavy discounting or remarketing often repeat less.
- Use margin, not revenue. A cohort that has repaid its acquisition cost in revenue may be far from repaying it in contribution.
- Agree the payback window with finance. It is a funding decision as much as a marketing one.
Setting and revisiting targets
Targets are not set once. Prices change, costs move, retention data accumulates and channels saturate. A sensible rhythm is to re-derive the targets whenever an input to the chain changes materially and to review them formally each quarter, alongside the reconciliation of channel figures to the blended view. The worked example below runs the whole chain once, from margin to channel targets and back.
An online tea retailer derives its targets from margin
Kettle & Leaf sells loose-leaf tea online at an average first order of £48. After cost of goods, packaging, delivery subsidy, payment fees and a 4% return allowance, contribution margin per first order is £26, so break-even CPA on the first order is £26 and break-even ROAS is 48 ÷ 26, about 1.85. Observed cohorts show customers place 2.1 further orders in the first twelve months at £21 margin each, so twelve-month margin per customer is roughly £70. Finance agrees a six-month payback window; cohort data shows £44 margin per customer by month six. The team sets a blended CAC target of £38, leaving room for error inside the payback limit. Channel reporting has historically summed to 1.3 times actual new customers, so channel CPA targets are set at £29. Three months later a supplier price rise cuts first-order margin to £22, and every target is re-derived before the next budget review.
The calculator below lets you change price, margin and spend and see how the return and break-even point move.
Return on spend calculator
Once the targets exist, the next question is how finely campaigns must be split to manage against them, which is the subject of campaign structure across channels. The fundamentals hub places economics alongside structure and tracking.
If the cost of winning a customer keeps drifting upward across channels, the rising acquisition cost diagnosis shows which component moved: saturation, mix, conversion, definition, auction prices or the close rate.
Where the reported return looks healthy but margin does not follow, the diagnosis of high ROAS without profit separates a target set without the margin from revenue counted before returns and sales the ads did not cause.
Common mistakes
- Copying the platform's suggested target. It reflects recent performance, not what the business can afford; derive the target from margin and give it to the platform.
- Using revenue where margin belongs. A ROAS target based on revenue alone ignores cost of goods and returns, so a campaign can hit target and lose money.
- Treating break-even as the goal. Break-even is the floor; with attribution error and overheads, spend at the line is usually spend below it.
- Judging budget changes on averages. The average hides the marginal return, which is what the next increment of budget actually buys.
- Assuming lifetime value instead of measuring it. Accepting first-order losses on an assumed repeat rate is the fastest way to grow into a cash problem.
- Never reconciling channel figures to blended CAC. Without the reconciliation, over-claiming goes unnoticed and every channel looks on target while the business is not.
- Setting targets once a year. Prices, costs and retention move; targets that do not move with them drift away from the economics they were meant to protect.
