There's a common pattern in eCommerce Google Ads that almost always has the same root cause. The Shopping campaign is performing well — strong ROAS, consistent revenue, happy account managers. But when you try to scale spend, ROAS collapses. Pull back, and it recovers. The account feels stuck at a ceiling it can't break through.
The instinctive diagnosis is a bidding issue, or a campaign structure issue, or a budget cap issue. Usually, it's none of those. It's a margin mix problem.
What's actually happening
When your Shopping campaign runs at a target ROAS — either through a smart bidding strategy or manual bidding against a benchmark — the algorithm learns which products and queries convert efficiently at that target. At moderate spend levels, it concentrates budget on your best-performing products: typically those with clear intent, strong creative, and sufficient margin to hit the target.
When you increase budget, the algorithm has to work harder to find volume. It starts reaching beyond your best performers into the rest of your catalogue — products with lower conversion rates, higher costs, different margin profiles. Those products can still hit your ROAS target in terms of revenue per ad dollar, but if their margins are lower, the real-world profitability of those sales is much worse.
Diagnosing the problem
The first thing to do is pull a product-level performance report from your Shopping campaign and cross-reference it with your margin data. Sort by spend and look at what's happening at the edges — the products Google is spending on when budget increases. Are they your highest-margin products, or are they lower-ticket, higher-volume items with thinner margins?
Then look at your average order value as spend increases. If AOV drops as you scale, that's a signal that the product mix is shifting toward lower-value items. If your average margin per order drops alongside it, you've confirmed the diagnosis.
The structural fix: segmentation by margin
The most effective long-term fix is to stop managing your Shopping catalogue as a single undifferentiated pool and start segmenting by margin tier. The mechanics vary depending on whether you're using Standard Shopping campaigns or PMax, but the principle is the same:
- Tag your products in the feed with custom labels that reflect margin tiers — high, medium, low
- Create separate campaigns or asset groups for each tier
- Set different ROAS targets for each tier based on the actual economics (high-margin products can tolerate lower ROAS; low-margin products need higher ROAS to be worth advertising at all)
- Allocate budget accordingly — more toward tiers where the economics support it
This requires knowing your margins by product, which can feel like a finance project. It is, to some extent. But most brands have a rough sense of which product categories are high-margin and which aren't — and even rough segmentation produces dramatically better budget decisions than treating everything the same.
What to do with your lowest-margin products
Once you've segmented, you'll often find a tier of products that can't be advertised profitably at any realistic ROAS target. These products might still have a role in your catalogue — as upsells, as bundle components, as loss leaders that drive repeat purchase — but paying Google to acquire customers for them at scale is usually a money-losing proposition.
The right answer is usually to exclude them from your paid traffic campaigns, or to only advertise them when bundled with higher-margin items. Let organic and email handle the demand capture for these SKUs.
The scaling path
Once you've addressed margin mix, scaling Shopping profitably becomes a different kind of problem. Instead of hitting a ceiling at an aggregate ROAS target, you're managing multiple tiers with different performance expectations — and you can grow spend in the high-margin tiers without being dragged down by the economics of your lower-margin catalogue.
The ROAS ceiling you thought was a structural limit turns out to be self-imposed. It was just an artefact of treating all your products as equivalent when they weren't.