Which target ROAS to set: from your margin, not from a benchmark

Target ROAS answers three questions, and each has its own number

In my view, the field in the account answers three separate questions at once. If one of them is missing, the target is either unprofitable, unreachable or holds the campaign back for no reason.

Google itself names business goals and historical ROAS performance as the basis. The recommendation in the account covers only the second row: it is based on the actual ROAS of recent weeks. Whether that value makes you money, it does not know. You therefore calculate the first and third rows yourself.

The floor: 1 divided by what may remain of the margin

Break-even ROAS is 1 divided by the margin before ad spend. If you want to earn something beyond that, subtract the desired contribution first: 1 ÷ (margin minus desired contribution). At an assumed margin of 40%, that gives:

There is no universally valid mark-up on break-even; the gap is a commercial decision: how much should advertising contribute to fixed costs and profit? No formula shows what that contribution costs in volume; only the campaign does.

The margin has to match the value definition in the account. If the conversion value is reported gross but the margin is calculated net, the floor no longer holds. If you report the value before returns, returns and their costs belong in the margin, not just as a deduction at the return rate. If you report the margin itself as the value, break-even is 1 by definition. Which costs belong in it is covered in the article on break-even ROAS.

Why industry benchmarks have no place in the derivation

A benchmark cannot answer any of the three questions, because it knows neither your margin nor your campaign. The same 400% leaves a 25% contribution at a 50% margin, but only 15% at a 40% margin. Even a benchmark overview for Facebook Ads concedes that its global average treats all revenue equally, regardless of margin, shipping costs or order value.

On top of that, in my view, there is the measurement basis. ROAS divides your own conversion value by your costs. Whether others report gross or net, count shipping, and how much brand traffic is in the value, you do not know. Using such a figure as an input means tying your bidding to someone else's arithmetic.

Floor against historical ROAS

You find the reference value in the account: Conv. value/cost multiplied by 100 gives the percentage. Leave out the most recent period still affected by conversion delay, otherwise the campaign looks worse than it is. Then put the two numbers side by side.

Historical ROAS above the floor. I would start close to the historical value, which also fits Google's advice to set the target at or below historical ROAS. From there, lower it in steps as long as the check in the next section comes out positive.

Historical ROAS below the floor. Here margin and feasibility contradict each other. Simply raising the target to the floor is risky according to Google: a target that is too high can limit the amount of traffic, and in Shopping campaigns targets set without considering historical performance can mean that the campaign may not serve at all. The business problem then remains, just with less revenue. I see three options:

• Change the basis: the range in the campaign, the margin or the definition of the conversion value.
• Accept the loss deliberately, for example for new customers, with a fixed budget cap. When that works out is covered under advertising below break-even.
• Withdraw budget.

Why ROAS 4 can still cost money

Target ROAS is an average, and that is where the popular argument that the campaign is well above break-even falls apart. The target is lowered so that the campaign can enter more auctions and generate more volume. Whether that pays off is decided solely by the additional money.

Worked example with invented, round numbers: margin 40%, so break-even is 2.5. The target is lowered, and costs rise by €500 in both cases.

In case A everything looks healthy: more revenue, ROAS 4, well above break-even. Yet €100 less is left over, because the last €500 brought only €1,000 in value. The check is therefore: additional value divided by additional cost, compared with break-even. If incremental ROAS is above it, carry on; if it is below, go back.

Google defines marginal ROAS as the additional value from additional spend and warns that average values for individual segments can be misleading. Applying the same logic to the target step is my conclusion. The comparison does not belong in the first few days: changes take effect immediately, but according to Google the new target is only reached after 1 to 2 conversion cycles. Beforehand, the bid simulator and the Save as experiment option give you a better basis than gut feeling. How large and how frequent the steps should be is covered under changing target ROAS.

One point remains a judgement call: if incremental ROAS is above break-even but the average is already below your desired value, the contribution rises in euros while its share of revenue falls. That is then a growth decision. It is legitimate, but it should be taken by the management team, not in passing in the account.

New campaign: the formula gives you the cut-off

Without history, feasibility is missing, and a number calculated from the margin does not replace it. I advise against starting a new campaign with the formula value, because then nobody knows whether it is failing because of the target or because of the offer.

For new Performance Max campaigns, target ROAS is selected when a recommendation is available, and you can uncheck the box. This is how I would proceed:

1. Start without a target if possible, and cap spend through the budget beforehand. The trade-off is covered under Maximise conversion value with or without a target.
2. Report values for at least 4 weeks, as Google recommends before setting a target.
3. Compare ROAS, excluding the delay period, with the floor, and then place the campaign in one of the two cases above.

Set the formula number in advance and write it down: it tells you when, after the four weeks, you stop or deliberately continue the campaign at a loss. The 15 conversions in 30 days that Google requires for Search and Shopping campaigns (as of November 2026) are an entry hurdle, not a sign that the target can already be steered reliably. For stable target ROAS in Performance Max, I work with around 60 conversions a month per campaign.

Seasonality distorts incremental ROAS

Incremental ROAS from a before-and-after comparison is distorted as soon as seasonality, promotions or conversion delay come into play. So compare periods whose delay has run out, or use an experiment in the account. If the step falls into a discount period, the result says little about normal trading.

Reported value is also not the same as revenue genuinely caused by the ads. With brand and existing customer traffic, a campaign can show high values, part of which would have come without the ad. Then the real ROAS is below the reported one, and every calculation here comes out too favourable.

Repeat purchases are left out of the example. If your business depends on them, the first purchase may sit below the floor, provided you can prove the follow-on value rather than just assume it. And a single margin for a range with very different margins produces an average that is right for none of the products. In that case, do the calculation per group, as described under target ROAS per margin group.