Performance Max
Segment Performance Max by margin, not by category
Categories are an order for people. The algorithm needs an order by profitability. What a campaign structure looks like that sends budget to where it actually sticks.
Published on · 7 min read
- 1One asset group per shop category mirrors your catalogue, not your business. Inside one category, products with 8 and with 62 percent margin sit side by side, and Google treats them the same.
- 2Performance Max looks for the easiest conversion. Without steering, budget ends up with thin-margin bestsellers and with products that would have been bought anyway.
- 3Custom labels bring margin, stock coverage and return rate into the feed. One campaign per margin tier with its own target, that is all it takes.
- 4The effect is unspectacular and immediately visible: budget moves to where the most sticks per euro of media spend.
The usual structure for Performance Max looks the same in almost every account: one campaign, inside it one asset group per shop category. Tops, trousers, shoes, accessories. It is intuitive because it rebuilds the shop navigation, and for steering it is mostly worthless. A category is an order for people browsing the shop. The algorithm does not browse. It needs an order by what a sale earns you.
Performance Max is built to find as many conversions as possible with as few constraints as possible. That is what makes it good. But it also means that inside an asset group it always takes the path of least resistance: the product that sells easiest, the user who wanted to buy anyway. If a bestseller with 8 percent margin and a new product with 62 percent sit in the same group, the bestseller gets the budget. Not because it pays off. Because it converts.
Why categories steer nothing
Take any category from your shop and write the margin after returns next to every product. The result looks similar in almost every range: a third of the products carry the profit, a third tag along, a third cost money. The algorithm sees none of it.
- Above 40% margin, carry the profit25 %
- 20 to 40% margin, tag along45 %
- Below 20% margin, lose money after ad costs30 %
Schematic split. In fashion and supplements the red group is often larger, because returns and discounts eat the margin there.
With one target ROAS across the whole category you lay a compromise over all three groups. For the green group it is too cautious, you leave reach on the table that would have paid off. For the red group it is too loose, you pay for sales you earn nothing on. And the reporting shows an average that hides both.
The structure that steers something
The alternative needs no new technology, only different information in the feed. Through custom labels every product gets a margin tier, plus stock coverage and return rate. From these labels come campaigns that each get a target fitting the goods.
| Campaign | Margin tier | Target | Budget logic |
|---|---|---|---|
| PMax A | above 50% margin | Low target ROAS, aggressive | Gets every budget the margin allows |
| PMax B | 35 to 50% | Medium target ROAS | Growth with a clear limit |
| PMax C | 20 to 35% | High target ROAS | Only where it pays off individually |
| PMax D | below 20% | Very high target ROAS or cap | Stock clearance only |
| No budget | negative after returns | Excluded from the feed | Not advertised at all |
The tier boundaries are examples. They follow your margin distribution, not a template.
- 1
Pull margin and return rate per product from the ERP
Gross margin after cost of goods, adjusted for the average return rate of the product group. Not perfect, but updated daily. A margin maintained once a year steers nothing.
- 2
Assign custom labels in the feed
Label 0 for the margin tier, label 1 for stock coverage in weeks, label 2 for the return rate in three bands. The assignment happens in the shop or the feed tool, not through rules in Merchant Center.
- 3
Create one campaign per tier
With its own budget and its own target ROAS derived from the margin. Inside the campaign the asset groups may happily be ordered by product type again, so the ads fit the range.
- 4
Exclude brand searches
Through the brand exclusions in every campaign. Otherwise the tier A campaign with its low target collects your brand searchers first, and the effect of the segmentation disappears behind revenue you would have had anyway.
- 5
Do not touch anything for four weeks
Every campaign learns afresh. During this time the distribution fluctuates, individual tiers get too much or too little. Whoever moves targets now extends the learning phase. After four weeks it shows where the budget really goes.
What happens afterwards
The effect is unspectacular and immediately visible. Budget moves to where the most sticks per euro of media spend. The tier A campaign spends more than expected, because its low target allows reach nobody had released before. The tier D campaign spends almost nothing, and that is exactly right.
- Tier A, above 50% margin42%
- Tier B, 35 to 50%31%
- Tier C, 20 to 35%18%
- Tier D, below 20%9%
Schematic split after the learning phase. Before, the red group from the split above often got more budget than the green one, because it converts more easily.
Structure by catalogue
- One campaign, asset groups per shop category
- One target ROAS across all products
- Budget lands with thin-margin bestsellers
- Reporting shows an average that explains nothing
Structure by margin
- One campaign per margin tier
- Target per campaign, derived from margin
- Budget lands with products you earn on
- Reporting shows per tier what it brings after ad costs
What you need first
The structure stands and falls with the labels, and the labels stand and fall with the data from the ERP. Whoever has no margin per product starts with a margin per product group. Whoever has no return rate takes the one of the category. Both beat a category structure that knows nothing about profitability.
A campaign structure is not a mirror of your range. It is a control instrument.
If your Performance Max is ordered by category today: export product performance for the last 90 days, add the margin per product and sort by media spend. If a product with single-digit margin sits in the top twenty rows, you know where part of your budget is going right now. And you have found the first product for tier D.