Google Shopping and PMax campaign structure by profit

An account ROAS of 3.36 can look healthy while two of its products lose €310 after ads. Campaign structure is how you stop funding them: group products by what they earn, then give each group the budget and target it can afford.

One campaign is the default, and custom labels decide which products each later campaign contains.

Which campaign structure fits Shopping and Performance Max?

The simplest one that supports a different decision. Each campaign is another budget and another set of results to judge, so it has to earn its place.

An asset group in Performance Max organises assets and products but has no budget of its own; a Standard Shopping product group is not a campaign either. If two groups need different budgets or targets, they need different campaigns.

Which products does the account average hide?

The ones whose ROAS clears the account target but not their own break-even, 1 ÷ contribution margin. Six products earn €11,100 of revenue excl. VAT on €3,300 of ads, a ROAS of 3.36:

Show the dataHide the data

The account's profit hides two products spending above contribution

Product	Contribution before ads	Ad spend
Trail Runner 2	€1,350	€1,000
Winter Boot	€960	€800
Everyday Sock	€240	€300
Rain Jacket	€750	€600
Gym Bag	€250	€500
Water Bottle	€175	€100

Everyday Sock beats Rain Jacket on ROAS and still loses money: its 20% margin needs a ROAS of 5, Rain Jacket’s 50% only 2. The break-even ROAS calculator gives each product’s floor.

Efficiency and volume are different questions too. Water Bottle runs at a POAS of 1.75 and contributes €75; Trail Runner 2 runs at 1.35 and contributes €350.

Use one basis throughout: revenue and margin both excl. VAT, and advertising left out of the margin you divide by (POAS vs ROAS).

How do you plan custom labels for the split?

Give each of the five slots one meaning and one writer. Google allows custom_label_0 to custom_label_4, one value each per product (custom label):

Which products a stock rule should flag is a per-product call: plan replenishment around profit and stock cover before you label anything.

Keep each meaning in its own slot rather than one long combined string. The six Product Metrics segments use ad clicks and return (ROAS, or POAS with margins), so a high-margin product isn’t automatically a Star.

A supplemental source can add a label to an existing product, matched on its ID (supplemental sources):

id	custom_label_0
DEMO-001	stars
DEMO-002	question_marks
DEMO-003	dogs

The free campaign label planner holds the six-product example, a five-slot plan and the formulas.

How do Standard Shopping and Performance Max differ when you split?

They select products differently and own budgets the same way:

Performance Max doesn’t automatically outrank Standard Shopping for the same products: overlapping campaigns compete on Ad Rank (campaign interaction). Check which campaigns actually include each product, “everything else” groups included; Performance Max for ecommerce covers exclusions.

When should you include, exclude, split or wait?

When the evidence supports that move, and not before:

Before excluding an entry product, check where its visitors go: Product Journeys shows purchases of other products after a first click. Keep thin-data products together until there is enough to separate them.

Log every change with its date, products, old and new values and reason, and judge it after the conversion delay. A higher POAS with much less contribution may miss the goal.

Final thoughts

A structure is worth its maintenance only when each part makes a different decision; most accounts need fewer campaigns and better labels.

Fill in the label planner with your own products, find the ones spending above their contribution, and give them their own decision before you add a single campaign.