Maximum CAC, payback period and contribution margin: the complete system

There are nine numbers that get used interchangeably and are not interchangeable at all. Current CAC, maximum CAC, contribution margin, payback period, average ROAS, marginal ROAS, customer value, revenue and profit. Mixing two of them up is not a vocabulary slip: it is what lets a green advertising dashboard sit next to a bank balance that keeps falling.

This guide puts them in order and shows how they chain together. There is no target figure at the end. There is a method for working out yours.

1. [The nine concepts, one at a time](#the-nine)
2. [The mistake everything else follows from](#the-mistake)
3. [The formulas](#formulas)
4. [A chained example](#worked-example)
5. [Why average ROAS misleads](#average-roas)
6. [The table you need to look at](#marginal-table)
7. [Payback, and what it cannot see](#payback)
8. [Common mistakes](#mistakes)
9. [Warning signs](#warning-signs)
10. [Limits of the model](#limits)
11. [When not to use it](#when-not)
12. [The procedure](#procedure)
13. [Frequently asked questions](#faq)
14. [Sources and methodology](#sources)

The nine concepts, one at a time

Revenue. What you charge. It includes tax if your prices carry it, and that tax is not yours.

Contribution margin. Tax-exclusive revenue minus the variable costs of the sale: product, packaging, shipping you absorb, fees, returns provision. It is the budget everything else comes out of.

Current CAC. What you are paying right now per new customer: ad spend divided by new customers in the same period. A measured fact, not a target.

Maximum CAC. The most you can pay without losing money, given your margin and what that customer will buy afterwards. A calculated ceiling, not an aspiration.

Payback period. How many months it takes to recover that CAC out of the contribution the customer leaves. A cash question, not a profitability one.

Average ROAS. Attributed revenue divided by total spend. It measures the whole.

Marginal ROAS. The increase in revenue divided by the increase in spend between two budget levels. It measures only the last step, and it is a different thing.

Contribution-based customer value. The sum of the contributions a customer leaves over their life, not the sum of what they bill. The honest version of LTV.

Profit. What is left after fixed costs. None of the eight above is profit.

The mistake everything else follows from

The dominant mistake is working out the CAC ceiling on gross revenue.

With the product in the example below, an order of €60.00 leaves €24.00 of contribution, because tax and variable costs come out first. Treat revenue as if it were available instead, and the CAC ceiling comes out at €60.00 rather than €38.39.

That is not a nuance. It is a ceiling inflated by more than half, and with it any campaign looks profitable. Where variable costs exist, and in practice they always do, maximum CAC is not calculated on revenue.

The first part of the calculation has a statutory basis, and it pays to cite the right rule. The tax statute — article 88.Uno of Spain's Ley 37/1992 — requires you to pass the full amount of the tax on to the customer, who is obliged to bear it: that governs the mechanics. The accounting rule — recognition and measurement rule 12 of Spain's Plan General de Contabilidad — is the one that says the VAT you charge does not form part of the revenue of taxed transactions. You collect it and you hand it over. It is not margin, and it is not revenue. Both rules are Spanish: check your own market's equivalents.

Do not conflate three quantities: revenue is what accounting recognizes; the taxable base is what the tax is levied on; cash billed is what reaches the bank with the VAT inside. The CAC ceiling is worked out on the first one.

The formulas

revenue ex-tax = price charged / (1 + rate/100)

contribution margin = revenue ex-tax − variable costs of the sale

contribution ratio = contribution margin / revenue ex-tax

break-even maximum CAC =
    contribution on the first order
  + repeat orders × contribution per repeat order

maximum CAC with a target = break-even maximum CAC − profit target

payback period = CAC / monthly contribution per customer

average ROAS  = total revenue / total spend
marginal ROAS = Δ revenue / Δ spend
break-even ROAS = 1 / contribution ratio

That last line is the most useful of the lot and almost nobody uses it. If forty cents of every euro billed are left to pay for everything, you need two and a half euros of revenue for every euro of advertising just to break even.

A chained example

A hypothetical shop sells a product at €60.00 tax included, at a rate of 21%.

Step 1: the margin. Stripping the tax leaves €49.59. Variable costs add up to €25.59: product €19.00, packaging €0.90, shipping €3.80, payment processing €1.09 — charged on the tax-inclusive amount, which is how it is actually billed — and a returns provision of €0.80.

Contribution is €24.00, which against tax-exclusive revenue is 48.4%.

Step 2: the CAC ceiling. If a customer buys once and then, on average, 0.6 more times, accumulated contribution is €38.39. That is the break-even maximum CAC: €38.39. Ask for €4.00 of profit per customer on top and the ceiling drops to €34.39.

Paying €16.00 per customer, the safety margin is €18.39.

Step 3: payback. At 0.15 orders per customer per month, monthly contribution per customer is €3.60. Recovering the €16.00 takes 4.45 months, or 135 days. Against a target of 6 months, is it met? yes.

Note that payback is computed on tax-exclusive revenue. Use the gross price and monthly contribution comes out inflated, making the period artificially short.

Step 4: break-even ROAS. With that contribution, break-even ROAS is 2.50. Below that figure every euro of advertising loses money, however green the dashboard. The equivalent maximum ACOS is 40.0%.

Why average ROAS misleads

Average ROAS blends the first campaign that worked with the latest one that no longer does. As long as the whole stays above break-even, the dashboard looks fine. And the whole can stay fine long after the last step stopped paying for itself.

Think about how scaling works. You start with the cheapest, warmest audience: the people already looking for you. When that runs out, you pay for audiences that know you less. Every new step returns less than the one before. How much less is an empirical question only your own data answers; nobody can hand you a saturation curve in advance, and anyone who does is making it up.

What can be stated, because it is algebra, is this: a profitable average ROAS can coexist with a ruinous marginal one, and deciding on the average leads you to keep raising a budget that has stopped paying.

The table you need to look at

A scenario with three spend levels and a contribution margin of 40%, which gives a break-even ROAS of 2.50:

Read it by rows, then by steps.

By rows, the third level has an average ROAS of 3.50, above break-even. On a dashboard that is green. And total profit is still positive: €4,800.00.

By steps, the story changes. The last €4,000.00 of spend brought only €6,000.00 of revenue. That is a marginal ROAS of 1.50, below break-even, and it destroys €−1,600.00 of profit.

The result: total profit falls from €6,400.00 to €4,800.00 while revenue rises from €36,000.00 to €42,000.00. You bill more and earn less, and average ROAS does not say so.

The optimum in this scenario sits at the middle level, €8,000.00 of spend. Not because that is a recommendable figure — it depends entirely on these numbers — but because that is where the next step stops paying.

The figures in this table are an arithmetic scenario, chosen so the effect is visible. They are not a market datum and not a forecast. The calculation you need is this same one, with your own last three real spend levels.

Payback, and what it cannot see

Payback answers a cash question: how long until you get back what you fronted. It matters because a campaign that is profitable over twelve months can kill you at three if you have no buffer.

But by construction it ignores two things, and they should be said:

Customer churn. The formula divides a cost by a periodic contribution. It contains no churn term at all. If half your customers do not reach month three, the calculated payback is optimiztic and the formula never notices.

The time value of money. There is no discount rate either. A euro recovered a year from now counts the same as one recovered tomorrow.

That is why payback is used as a cash constraint, not as a profitability measure. Profitability is what maximum CAC is for; cash is what payback is for; scaling is what the marginal figure is for. Three questions, three numbers.

And a warning about the benchmarks in circulation: there is no "good" payback for an ecommerce business. The payback periods quoted as standard come from subscription software, not retail, and they do not even agree with each other. Your benchmark is your own cash position.

Common mistakes

Working out maximum CAC on revenue. Already said. It is the mistake that costs the most.

Using the blended CAC for the whole account. Brand CAC and prospecting CAC have nothing to do with each other, and blending them hides exactly the step that is failing.

Counting returning customers as new ones. It inflates the denominator and lowers CAC artificially.

Basing LTV on revenue. A customer who bills a lot at zero margin is not worth a lot.

Comparing ROAS across two periods with different campaign mixes. That is not the same question.

Deciding scale on average ROAS. That is half of what this guide is about.

Borrowing a payback benchmark from another sector. It does not apply, and nobody says so.

Warning signs

• Revenue rises and profit does not. Look at the marginal step.
• Blended CAC climbing month after month while average ROAS holds: you are offsetting bad steps with old good ones.
• Payback approaching the length of your collection cycle: any delay leaves you short of cash.
• A maximum CAC calculated six months ago. The margin has moved.
• A ROAS target inherited from an agency without the arithmetic behind it.

Limits of the model

Attribution. All of this rests on being able to assign revenue to spend. With several channels, different attribution windows and purchases that start in one place and finish in another, that assignment is approximate. These numbers are only as good as your attribution.

Repeat orders are a forecast. Maximum CAC uses how often the customer will come back. That comes from the past and is not guaranteed. Inflate it and you inflate the ceiling.

Margin is not constant. It moves with the product, the promotion and the channel. A maximum CAC built on the average margin is too generous on thin products and too tight on fat ones.

The marginal figure needs comparable steps. Comparing two spend levels separated by a change of creative, season or product does not measure saturation: it measures something else.

None of this predicts. The marginal table describes what already happened. The next step has to be measured, not deduced.

When not to use it

• With very few orders: average contribution means nothing and CAC swings with every sale.
• On a launch with no repeat history: the CAC ceiling would rest on an assumption rather than a measurement.
• When the real problem is the cost structure. If break-even sits far above your order count, no campaign optimization fixes it.

The procedure

1. Work out the contribution margin of an average order, ex-tax.
2. Measure how many orders a customer places in twelve months, from real data.
3. Compute break-even maximum CAC and subtract the profit you want.
4. Measure your current CAC per campaign, keeping brand and prospecting apart.
5. Compute the payback period and compare it with your cash cycle.
6. Compute break-even ROAS as 1 divided by your contribution ratio.
7. Take your last three real spend levels and build the marginal table.
8. Find the first step whose marginal ROAS falls below break-even.
9. Decide on the step, not on the average.
10. Redo the whole thing when margin, price or channel mix changes.

Frequently asked questions

Sources and methodology

Sources

• Accounting rule. Spain's Plan General de Contabilidad, Royal Decree 1514/2007, recognition and measurement rule 12, consolidated text in the Boletín Oficial del Estado (BOE-A-2007-19884). It states in terms that VAT charged to customers does not form part of the revenue of the taxed transactions. This is what supports doing the whole calculation on tax-exclusive revenue. Territory: Spain. This is the Spanish accounting framework and it does not carry over to other markets on its own; the same principle exists elsewhere, but you have to open your own market's rules. In force; consolidated text consulted 9 September 2026.
• Regulator's interpretation. Ruling 5 in BOICAC 96, issued by Spain's Accounting and Audit Institute (ICAC). It confirms that taxes the company must charge on to third parties, and amounts received on behalf of third parties, do not form part of revenue. Territory: Spain. Consulted 9 September 2026.
• Tax statute. Ley 37/1992, the Spanish VAT Act, consolidated text in the Boletín Oficial del Estado (BOE-A-1992-28740). Article 88.Uno. Used only for the obligation to pass the tax on, the way it is passed on, and the relationship between the taxable person and the customer. It is not used as the basis for the accounting treatment of revenue — that is what the accounting rule above is for. Territory: Spain. In force; consolidated text consulted 9 September 2026.

What is not cited, and why. There is no "good" payback benchmark here. The research turned up medians that contradict each other — sixteen months in one source and under seven in another — all of them about subscription software, none about ecommerce, and none with a published method. Both were discarded rather than picking whichever suited.

Methodology

The figures in the example are produced by Profyza's calculators — contribution margin, maximum CAC, CAC payback and break-even ROAS — run in sequence with the inputs declared above. The marginal ROAS table is generated by a private, reproducible function from the declared margin and the three declared spend levels; not one figure in that table is typed by hand into the text. Everything is separately checked against an independent calculation.

The data describe a hypothetical shop and the spend scenario is arithmetic: it exists to show the effect, not to predict any campaign's performance.

Currency and market: euros, European Union. The statute cited is Spanish.

Last reviewed: 9 September 2026.

Author: Profyza editorial team.

How the figures were produced. Each one is generated by the Profyza calculation engine for that step, using the inputs declared above. Every figure is also cross-checked against an independent arithmetic check — also automated — that recomputes it without calling the engine, which is what catches the text and the calculation drifting apart. It is not a human review, and this piece has not been reviewed by an economist, an accountant or a licensed adviser. If your situation needs that, get it.

Methodology: how all of this is calculated is set out on the methodology page.

Automated mathematical verification: Profyza calculation engines. This is not a human or professional review.