What cost per acquisition measures, and what it is not
CPA is the price you paid for one conversion. Divide the spend by the number of conversions and you have it: $5,000 across 50 new customers is a $100 CPA. Because it is denominated in dollars per customer rather than as a ratio, it compares directly against something concrete — the profit a customer produces — which is why it survives as the working metric in lead generation, B2B and any business with a long sales cycle.
Three related terms get used interchangeably and should not be. Marketing's own standards body, the Marketing Accountability Standards Board, maintains the Common Language Marketing Dictionary precisely to stop that drift; when a definition is contested inside your company, settle it there before you argue about the number. Cost per lead divides the same spend by enquiries rather than closed customers, so it is always the smaller number; the gap between the two is your close rate. Customer acquisition cost is broader than CPA — it includes sales salaries, tooling and overhead, not just media, which is why a $100 CPA can sit alongside a $350 CAC in the same business. And in affiliate marketing, cost per action means a payment model rather than a measurement, where the publisher is paid only when a defined action fires.
The number that makes CPA useful is not the CPA itself but its ceiling. Your maximum allowable CPA is the contribution one order generates: revenue times gross margin, minus the variable costs of serving that order. Spend less than that per conversion and you make money; spend more and you are buying customers at a loss, which is defensible only when repeat purchases pay you back later.
The formulas, and how the funnel links them
CPA = spend ÷ conversions. Simple, provided you are strict about what counts. Include every acquisition cost you want the ceiling to cover — media alone for a bid-level CPA, or media plus agency fees, creative production and lead-gen tooling for a figure you can show a finance team.
CPA = cost per lead ÷ close rate. The funnel version, and the more actionable one. If leads cost $20 and your sales team closes 20% of them, each customer costs $100. Two independent levers sit in that equation, and they are usually owned by different people: marketing moves the cost per lead, sales moves the close rate. Lifting the close rate from 20% to 25% cuts CPA from $100 to $80 without buying a single extra lead.
CPA = CPC ÷ conversion rate. The same identity one step earlier in the funnel. A $2.00 click at a 4% conversion rate is a $50 CPA. This is the form you use to turn a CPA target into a bid — see the cost per click calculator for the maximum-bid version.
Maximum allowable CPA = AOV × gross margin − other variable costs. Take a $400 order at a 55% gross margin: $220 of gross profit. Subtract $30 of shipping, processing and fulfilment and you have $190 of contribution. That $190 is the entire budget available to win the order. Bid to $190 and you break even; bid to $100 and you keep $90.
Notice that fixed overhead is deliberately absent. Rent, salaries and software do not change when you sell one more unit, so they do not belong in a marginal decision about whether to buy one more customer. They matter enormously for whether the business is profitable overall — just not for this calculation.
Worked example: $5,000 of lead-gen spend on a $400 product
A month of paid search and paid social costs $5,000 and produces 250 enquiries, of which the sales team closes 50. Average order value is $400 at a 55% gross margin, and each order costs another $30 in shipping, card fees and onboarding.
- Cost per lead: 5,000 ÷ 250 = $20.00.
- Close rate: 50 ÷ 250 = 0.20 = 20%.
- CPA: 5,000 ÷ 50 = $100.00. Cross-check through the funnel: 20.00 ÷ 0.20 = $100.00.
- Gross profit per order: 400 × 0.55 = $220.00.
- Contribution per order: 220 − 30 = $190.00. This is the maximum allowable CPA.
- Profit per conversion: 190 − 100 = $90.00.
- Total contribution: 50 × 190 = $9,500, minus $5,000 of spend = $4,500.
- Break-even conversions: 5,000 ÷ 190 = 26.3. You needed 27 sales to cover the spend and you got 50.
Now test the two levers against each other. Suppose the sales team lifts the close rate to 25%: the same 250 leads produce 62.5 customers, CPA falls to $80, and contribution rises to 62.5 × 190 − 5,000 = $6,875. Alternatively, marketing cuts cost per lead from $20 to $16 while holding the close rate: the same $5,000 buys 312.5 leads, 62.5 customers, and the identical $6,875. A 20% improvement in either lever is worth the same $2,375 — useful to know before you decide which team gets the investment.
How to read your CPA: three thresholds
The contribution threshold. Is CPA below the maximum allowable CPA? This is the test the calculator runs, and it is the right one for a business selling a one-off purchase. Anything above the ceiling loses money on every order.
The lifetime-value threshold. If customers buy again, the first order is not the whole revenue and the ceiling rises. A subscription business or a consumable brand can rationally pay a CPA several times a first order's contribution, provided the repeat behaviour is proven from cohort data rather than assumed. The usual discipline is a payback period: many operators want acquisition cost repaid inside 12 months of contribution. Work it out with the CAC payback period calculator and check the ratio with the LTV to CAC ratio calculator; a lifetime value at least three times acquisition cost is the widely used rule of thumb, though it is a heuristic and not a law.
The volume threshold. Marginal CPA rises as you scale, always. The cheapest, most intent-driven audience is bought first, and the next increment of budget reaches people who are less ready. So the CPA that matters for a spending decision is not your average but the CPA of the next dollar. If average CPA is $100 and your ceiling is $190, you can keep spending until the marginal CPA reaches $190 — which will usually be at a much higher spend than the average figure suggests.
Watch the sample size too. CPA on 12 conversions is noise; two lucky sales move it 20%. Below roughly 30 conversions in a period, treat the figure as directional and judge on a longer window.
Maximum allowable CPA by order value and gross margin
| Average order value | 20% margin | 30% margin | 40% margin | 50% margin | 60% margin | 70% margin |
|---|---|---|---|---|---|---|
| $50 | $10 | $15 | $20 | $25 | $30 | $35 |
| $100 | $20 | $30 | $40 | $50 | $60 | $70 |
| $200 | $40 | $60 | $80 | $100 | $120 | $140 |
| $400 | $80 | $120 | $160 | $200 | $240 | $280 |
| $800 | $160 | $240 | $320 | $400 | $480 | $560 |
| $1,500 | $300 | $450 | $600 | $750 | $900 | $1,050 |
These are single-order ceilings. If customers reliably repeat, multiply by the number of orders you can defend from cohort data — not from optimism.
CPA and CAC are not the same number
CPA divides media spend by conversions. CAC divides all acquisition cost — media, agency retainers, creative production, sales salaries and commission, CRM and enrichment tooling — by new customers. In a business with an inside sales team the gap is wide: add $12,000 of sales salary and $500 of CRM to the $5,000 media budget in the worked example and the same 50 customers cost $350 each all-in, against a $100 media CPA.
Use CPA to set bids and judge channels, because it moves with the levers a media buyer controls. Use CAC to judge whether the business model works, because it includes the costs a channel report never shows. Reporting one and calling it the other is how a company convinces itself that unprofitable growth is profitable.
Mistakes that make a CPA look better than it is
- Counting leads as acquisitions. A form fill is not a customer. If your CPA is really a cost per lead, the ceiling has to be the lead's expected value — contribution per order multiplied by the close rate — not the full order contribution.
- Leaving agency and creative cost out of the numerator. A 15% retainer plus production makes a reported $100 CPA closer to $120 all-in. Decide which convention you use and label it.
- Using revenue instead of contribution for the ceiling. The most common and most expensive error. A $400 order supports $190 of acquisition cost, not $400.
- Ignoring refunds and cancellations. A conversion that refunds still cost you the acquisition. Deduct your refund rate from the conversion count or from the contribution before setting the ceiling.
- Averaging CPA across new and returning customers. Retargeting and email pick up buyers you already own. Segment new-customer CPA from blended CPA or you will over-credit the cheapest channels.
- Comparing CPA across attribution models. A platform-reported CPA on a 7-day click window and a last-non-direct CPA from your analytics are different numbers about the same spend. Never mix them in one table.
- Setting a target CPA from history rather than margin. “Last year we paid $85” is not a constraint, it is a habit. The constraint is contribution per order.
Where CPA fits among the other efficiency metrics
Use CPA when order values are similar across conversions, when you are buying leads rather than transactions, or when the sales cycle is long enough that revenue arrives well after the click. It is the natural bid unit for B2B, services, education and insurance.
Use ROAS or ACoS when order values vary widely, because a fixed dollar CPA over-rewards cheap orders and starves expensive ones. A catalogue spanning $20 accessories and $900 systems needs a revenue-weighted target: the ROAS calculator covers the break-even version, and the Amazon ACoS calculator the marketplace form.
Use ROMI to report the same result upward as a percentage return on the marketing investment — the marketing ROI calculator does that, including the incremental version that strips out baseline sales.
Underneath all of them sits the same input: contribution per order. Build it properly with the contribution margin calculator, and if the number comes out too thin to support any realistic acquisition cost, the answer is not better bidding — it is a higher price, a bigger basket or a cheaper product. Check what the price change would take with the average order value calculator.