Setting different target ROAS for different margin groups

Margin groups are the middle way, as long as no margin is in the conversion value

If you still report revenue as the conversion value to Google Ads, a target ROAS steers every product as if it had the same margin. Margin groups correct this roughly: products with a similar contribution margin rate are put together, and each group gets its own target ROAS. This is less precise than margin as conversion value, but it can be done without touching tracking. Common mistakes are too many groups and wrongly calculated targets. Two simple calculations avoid both.

The target per group comes from the profit that should be left

The lower limit for each group is its break-even ROAS: 1 divided by the contribution margin before ad spend as a percentage of revenue, including all costs that arise with each order (Lyonic). How to determine that rate cleanly is covered in calculating break-even ROAS. At a 40% contribution margin the break-even is 2.5, at 25% it is 4, at 15% it is 6.67.

At break-even you earn nothing. The target must therefore sit above it, and many people reach for a flat markup such as break-even times 1.25. This factor is not derived from anything, and it has a side effect: at a factor of 1.25, exactly 20% of the contribution margin is always left after ad spend. The group with the lowest margin therefore automatically delivers the lowest profit per euro of revenue, even though it often carries the same share of warehouse, service and fixed costs.

It is better to calculate backwards from the profit each group should still yield after ad spend:

Target ROAS = 1 ÷ (contribution margin rate − profit to keep, as a percentage of revenue)

A worked example with round, assumed figures: a shop wants to keep 10% of revenue as profit from every group.

The right-hand column shows why this calculation is more honest. For group C the result is a target of 20, which hardly any campaign reaches in practice. According to Google, a target ROAS set too high can reduce the traffic your ads receive (Google Ads Help). That puts the real decision openly on the table: either you accept less profit for group C, say 5% and therefore a target of 10, because these products bring new customers or baskets. Or you accept that they will hardly be served any more. The factor of 1.25 makes this decision silently for you, with 3% left over.

If the group is meant to bring in new customers, that belongs in a calculation of its own, see which target ROAS to set. For goods that serve only existing customers, I advise setting the profit kept the same for each group and justifying exceptions deliberately.

How many groups your data volume supports

The number of groups does not follow from the range but from the conversions. For target ROAS in Search and Shopping campaigns, Google names at least 15 conversions in the last 30 days as a precondition. For a meaningful evaluation of target ROAS, Google recommends around a month with at least 50 conversions (Google Ads Help, Smart Bidding).

From that I derive a simple upper limit:

Number of groups ≤ conversions of the affected campaigns per 30 days ÷ 50

A shop with 160 Shopping conversions a month therefore supports three groups, one with 70 only one. What counts is the distribution: count the conversions per planned group, not on average. High-margin niche products often sell little, and a group with 12 conversions a month is one you steer practically blind.

The reason for the 50 lies less with the algorithm than with you. According to Google, Smart Bidding can optimise using data from all campaigns in the account, even for new campaigns without data of their own. Learning is therefore not limited to the individual group. Evaluation is. Whether group B hits its target ROAS of 6.67 cannot be told apart from chance at 15 conversions a month. A group whose result you cannot check is not steering but an assumption. More on data volume in how many conversions target ROAS needs.

In most shops I consider three groups the sensible upper limit, even if the data would support more. If two groups have targets close together, say 5.5 and 6, separating them achieves little and costs data. Then merge them.

How the groups get into the campaigns

The group belongs to the product as an attribute, so in the feed. Merchant Center offers five custom labels, custom_label_0 to 4, with up to 1,000 unique values per label. Google names margin explicitly as an example for such a label, and the labels as filters for product groups in reports and bidding, though only for Performance Max, Shopping and Demand Gen (Merchant Center Help). Which other labels need space is covered in custom labels: which five fields are really worth it.

I would name the label value after the contribution margin range, such as "db_35_45", not "high" or "medium". That keeps it clear which rate stands behind the target when prices or purchasing change. The label has to come from the shop or inventory system and move with price changes. Otherwise a product on sale slips into a group that is too generous, still carrying its old label.

Each group then gets its own campaign, filtered to its label value. Common specialist articles recommend this as well (Klickkonzept), and Google's bidding logic leaves hardly any other route:

• Performance Max: According to Google, portfolio bid strategies are not available for Performance Max (Google Ads Help). Every group with its own target needs its own campaign.
• Shopping: A portfolio would only have one target for all groups anyway. So here too, one campaign per group.
• Targets per ad group: Possible, but not recommended by Google because they can restrict Smart Bidding. For margin groups I advise against it.

Every additional campaign needs its own budget. Give the group with the highest profit per euro of ad spend room first, before you scale up the weakest group.

When a group has too little data

A group below the data threshold is merged with the neighbouring group instead of being run with a target of its own. You calculate the target for the new group from the contribution margin rate weighted by revenue. Worked example: group B at 25% makes 70% of revenue, group C at 15% the rest. The blended rate is 22%, and with 10% profit kept that gives a target of 1 ÷ 0.12 = 8.33.

This calculation has a limit: Smart Bidding does not know your margin. Within the merged campaign, revenue can shift towards the low-margin products if they sell more easily. The weighted rate then no longer holds. I would recalculate the campaign's actual contribution margin rate monthly and adjust the target as soon as it deviates noticeably.

With data that is too thin, patience in evaluating also helps. After a target change, according to Google you should wait 1 to 2 conversion cycles until the new target is reached. If you readjust margin groups weekly, you are mostly measuring your own restlessness. How much and how often to adjust is described in changing target ROAS.

When margin groups are not worth it

If your range already sits in a narrow margin band, say everything between 28 and 34% contribution margin, the targets are so close together that splitting only spreads data thin. One target ROAS for everything is then enough.

If the conversions support only one group, splitting is also premature. The better lever is then to pass the margin directly as the conversion value: the bid then sees the contribution margin of each individual order, without you spreading data across several campaigns.

And if margins per product fluctuate strongly, through frequent promotions or changing purchase prices, any grouping goes out of date quickly. Margin groups are a tool for a stable range with clearly separated margin levels. Feature information as of: November 2026.