A low ticket price tells you where to look. It does not tell you whether an item should advertise.

One cheap product can absorb clicks without leaving enough margin to pay for them. Another can appear in an attributed basket with three profitable add-ons. Excluding both at the same price threshold treats two different commercial roles as if they were the same.

Paid eligibility should follow expected order contribution and the job the product performs in the catalogue.

Give every advertised product a job

Start by assigning the item to one of three roles.

A profitable lead product is associated with attributed orders whose expected contribution can repay acquisition cost. Keep it eligible and judge the attributed basket alongside a controlled test.

A discovery product has a deliberate subsidy. It may introduce a category, recruit a new customer or create cross-sell. Give that subsidy its own budget, decision date and downstream measure. Revenue ROAS alone cannot show whether the role is working.

A value-destroying product cannot cover media plus the contribution the business needs to retain, and has no evidenced strategic role that justifies the gap. Reduce its exposure or remove it from paid inventory.

The role needs an owner. Otherwise every weak item becomes “discovery” and keeps spending indefinitely.

Calculate the ceiling finance can defend

The useful starting point is contribution before advertising:

gross merchandise revenue - discounts - expected returns and refunds - COGS - payment fees - variable fulfilment - shipping subsidy - other variable order costs

Then set the media limit:

allowable acquisition cost = expected contribution before advertising - required post-ad contribution

There is no universal list of costs or acceptable margin. A retailer that absorbs delivery and returns has different economics from one that charges shipping and rarely receives products back. Finance should approve the definition, the return assumptions and the contribution the order must leave behind.

This calculation also explains why rules such as “pause after spending twice the product price” are unreliable. Price says little about COGS, fulfilment, returns or the rest of the basket.

Read the basket behind the click

Google’s conversions with cart data can distinguish the product shown in the ad from the products purchased later. With complete item data and cost_of_goods_sold, the reports include lead and cross-sell revenue or gross profit.

That distinction matters. If a shopper clicks an ad for an inexpensive refill and buys the refill plus a higher-contribution device, evaluating only the advertised SKU hides the full order attributed to the ad interaction. If the shopper buys only the refill, the same report exposes the weak economics.

Google’s gross profit is still an intermediate measure. It subtracts the submitted COGS from revenue, but it does not separately deduct payment fees, pick and pack, shipping subsidy or expected returns. Unless those costs are already included in the retailer’s COGS convention, join them from order and finance data before calling the result contribution.

Attributed cross-sell also needs restraint. The report shows which basket followed an ad interaction. It does not prove the ad caused every additional item. When the decision is material, run a controlled product-eligibility experiment with a contemporaneous holdout, or use another defensible incrementality design, instead of treating the attribution column as causal evidence.

Turn the decision into feed control

Once the economics are clear, encode the decision in the feed. Google supports up to five custom labels for Shopping ads. Reserve one for a governed state such as:

  • paid_core
  • paid_discovery_capped
  • organic_only
  • margin_data_missing

Use those labels to build Shopping product groups or Performance Max listing groups. Performance Max listing groups default to all products, so an unclassified item can quietly enter an included remainder and become eligible to receive spend. Exclude that remainder until it has an approved state.

Keep the structure proportionate. Products need separate campaigns when they require a different budget or target. A distinct test window may also justify separation when it cannot be governed cleanly inside the existing campaign. They do not need one campaign per SKU. The existing guide to Performance Max controls covers the wider campaign choices.

Labels also decay. Supplier costs, promotions, prices and shipping terms move. Refresh the economic state from a controlled source and audit the items sitting in margin_data_missing rather than letting yesterday’s high-margin label steer tomorrow’s spend.

Revenue ROAS will keep chasing revenue

Target ROAS uses the conversion values supplied through conversion tracking. If a purchase sends revenue, bidding is guided by revenue. Adding COGS for reporting does not silently change that objective.

Profit-weighted bidding requires an intentional value design and a validated transition. Until that exists, separate products that need different economics and use the contribution view to judge the output. A blended campaign ROAS can look healthy while cheap, low-contribution orders consume the budget.

Stop paying without hiding the product

An item that fails the paid test does not have to disappear from every Google surface. Exclude it from the relevant listing group first. If the policy should apply across Shopping ads, Merchant Center’s excluded_destination attribute can set Shopping_ads without excluding Free_listings, provided free listings are enabled and the item remains eligible for that destination.

That is the final discipline: protect paid budget while preserving useful catalogue discovery. Use attributed order contribution to decide which products deserve a controlled eligibility test. Continue funding a strategic role only when the test and its downstream measure support it.