ROAS Calculator

What this tool does

CM8-369 takes the return-on-ad-spend figure your advertising platform reports and turns it into one your accounts would recognise. It applies the corrections that can be quantified — margin, refunds, fees and measured incrementality — and compares the result against the ROAS your margin actually requires you to reach.

Everything runs inside this single file — no account, no upload, no network request of any kind.

Why the reported figure is not the real one

A platform reports revenue it can claim, within its own attribution window, on its own definition. It does not know four things, and every one of them makes its figure too high:

• Your gross margin. Revenue is not profit.
• Your refunds. Revenue that came back is not revenue.
• Your fees. Agency and tool costs never appear in a platform's ROAS.
• Whether the sale would have happened anyway.

None of this is the platform being dishonest — it is measuring what it can see. The mistake is treating what it can see as the whole answer.

Break-even ROAS

This is the number most advertisers cannot state for their own business, and everything else depends on it. At a 60% margin you need about 1.67× to stand still. At 25% you need 4×. At 15% you need nearly 7×, which is why advertising rarely works on thin margins however good the targeting.

The reference table prints break-even at every common margin. It is worth pinning up: it turns "our ROAS is 3" from a fact into a question about which product mix was sold.

The margin correction

Set a default margin in Settings and override it per campaign where the product mix differs. That override matters more than it sounds: discount and clearance campaigns sell a worse mix, so the campaign with the highest reported ROAS often has the lowest margin behind it.

The sample has exactly this. Clearance shopping reports 3.6×, which sounds respectable, and sits below break-even because the stock carries 28% margin rather than 61%.

Refunds

Entered per campaign as a share of attributed revenue, because the variation between channels is large and rarely compared. Prospecting on social routinely refunds at two or three times the rate of branded search, for the obvious reason: it sells to people who were not looking.

Refunds raise break-even as well as reducing revenue, which is why they appear on both sides of the calculation.

Fees

Agency retainers, management percentages and platform tools are real money spent to run the advertising, and no platform includes them. A 15% management fee raises break-even ROAS by 15%, which is usually enough to move a marginal campaign onto the wrong side of the line.

Incrementality, and why the tool will not guess it

The hardest correction, and the largest. Some of the revenue a campaign claims would have arrived without it — most obviously on brand search, where you are paying for people who typed your name, and on retargeting, where you are paying to reach people who already visited.

Incrementality is a measurement, not a setting. The only way to know is to run a test: hold out a region or an audience, turn the campaign off there, and see what happens to revenue.

The tool therefore assumes 100% incremental unless you have tested, which is deliberately the least flattering possible treatment of your evidence. It will not accept an incremental share on a campaign not marked as tested, and it will not accept a tested campaign without the test written down in the notes. An assumed uplift is a way of getting the answer you wanted.

The sample shows why it matters: brand search reports 17× and measures 21% incremental, which brings it to roughly 3.6× — still good, and a fifth of what the platform claimed.

Attribution windows

Recorded per campaign because comparing a 1-day window against a 28-day one is not a comparison. A longer window always reports more revenue for the same advertising.

View-through attribution — crediting a campaign when somebody saw an advert and did not click — is the most generous of all, and a campaign using it that has never been tested is flagged on that basis alone. It may still be working. Nothing in the reported figure tells you either way.

Profit ROAS

Profit ROAS is scaled so that 1.0 means the advertising exactly paid for itself in gross profit. Above 1.0 it contributed; below it, every additional unit spent made things worse. That is a cleaner reading than a reported multiple, which means something different for every business.

Multiples against money

Do not manage by the multiple alone. A campaign at 8× on a small budget contributes less than one at 1.4× on a large one, and cutting the second to protect an average is how advertisers reduce profit while improving their reporting.

The profit chart is in money for that reason. Use the multiple to decide whether a campaign should exist, and the money to decide what to do next.

What this does not do

• It cannot measure incrementality. It records what your test measured, and refuses to substitute a guess.
• It stops at the first purchase window the platform reports. A customer who buys again next year is not in here — that is a lifetime value question.
• It cannot see one campaign helping another. Prospecting that generates the retargeting audience looks worse here than it is, and retargeting looks better.
• It has no view about diminishing returns. A campaign at 4× may still be at 2× after you double it.
• It assumes attributed revenue is at least approximately real. Where tracking is broken, no correction here fixes it.

Printing and sharing

The Report tab prints the tiles, charts and both tables with a title block you fill in. The workings table walks from reported ROAS to profit in one line per campaign, which is the fastest way to settle a disagreement with whoever is quoting the platform figure.

Saving your work

Campaigns are held in this browser, on this computer, and stay there between visits. Use the backup button to write a JSON file you control.

Accuracy & disclaimer

Every figure here is one you entered, including the reported revenue, which came from a system with its own definitions. The tool applies the stated corrections exactly and takes no view on whether a campaign should run.