ROAS

Divide the revenue attributed to advertising by what the advertising cost. €20,000 of revenue on €5,000 of spend is a ROAS of 4.0, sometimes written 400%. That much is arithmetic and nobody argues about it.

Both inputs are less solid than they look. Spend is usually the cleanest figure available, though it still excludes agency fees, creative production and the tooling billed to the account, so "ad spend" in the denominator is narrower than what the client actually spent on advertising. If you want a number that describes the whole marketing investment, the term for it is ROMI, and it will be considerably lower.

Revenue is where the real ambiguity sits. The platform is reporting revenue it decided to credit to itself, under rules it set, over a window it chose. None of that is visible in the number, which is why two dashboards can disagree by a factor of two about the same week and both be internally consistent.

A conversion window is the period after seeing or clicking an ad during which a purchase still gets credited to it. Meta has historically defaulted to a 7-day click and 1-day view window; Google Ads defaults vary by conversion action and can run to 30 days or beyond. Widen the window and ROAS rises. Narrow it and ROAS falls. The campaign did not change.

View-through conversions are the sharpest version of this. Crediting a sale to someone who saw an ad and did not click it is a defensible modelling choice and an indefensible claim of causation, and it is the single largest contributor to the gap between platform-reported revenue and the client’s order total.

Then there is double counting. Each platform can only see the paths that included itself, so a customer who saw a Meta ad and later clicked a Google one is claimed in full by both. Summing platform-reported revenue across channels produces a number larger than the business actually took, and a client who spots that stops trusting everything around it.

The practical consequence: ROAS is a reasonable tool for comparing two campaigns measured the same way, and a poor tool for stating what the advertising earned. Those are different jobs and the same number gets used for both.

A ROAS target set without reference to the client’s margin is a guess. The break-even point, where the revenue exactly covers the cost of goods and the advertising, is the reciprocal of gross margin.

At a 50% gross margin, break-even ROAS is 2.0: every €1 of spend must return €2 of revenue for the advertising to pay for itself. At 25% margin it is 4.0. At 70% it is about 1.43. Which means a target of 3.0 is comfortable for a software business and catastrophic for a grocer, and the same number describes both situations.

This is why "industry benchmark ROAS" is close to meaningless. The benchmark that matters is the client’s own break-even, and it takes one question at kickoff to establish it. It is also the question that most often reveals the client does not know their own blended margin, which is worth finding out early rather than during a renewal.

Profit on ad spend, POAS, closes the gap properly by swapping revenue for gross profit, so the target becomes simply "above 1". It is the better decision metric and much harder to get, because it needs margin data per product that many clients cannot produce and some will not share.

ROAS and volume pull against each other, and the tension is structural rather than a sign of poor management. The cheapest demand gets captured first: people already searching for the brand, people who abandoned a basket last week, people on a remarketing list. Serve those and ROAS looks excellent on a small budget.

Push for growth and you are necessarily buying demand that is further from the purchase, which converts worse and costs more. ROAS falls. If ROAS is the only target, the rational response is to cut back to the efficient core, and an account managed that way gets steadily smaller while reporting steadily better.

The clients who get hurt by this are the ones growing. An e-commerce business that needs new customers cannot get them from a remarketing pool, and an agency defending its ROAS number is, in effect, refusing to spend on the only thing that would grow the business.

The way out is to report prospecting and remarketing separately with different expectations for each, and to talk about marginal return rather than average. The question worth answering is not "what is our ROAS" but "what would the next €10,000 return", and those have very different answers.

The argument, when it comes, is almost never about the advertising. It is about a client comparing a platform number to their own accounting and concluding they have been misled, which is a trust problem, not an analytics one, and it is far easier to prevent than to repair.

Three things prevent it. Report platform ROAS next to the client’s own revenue figure, from the start and without being asked. Label the attribution model and the conversion window on the report itself, so a change in either is visibly a change in measurement. And show prospecting and remarketing as separate lines, so nobody has to work out on their own that the headline was mostly re-reaching existing customers.

The agencies that get into trouble here are usually the ones that reported a single blended figure for six months because it looked good, and then had to explain it on the day it stopped looking good.