Why Your Results Feel... Mushy
You've seen the ROAS numbers. They look fine. Maybe even good. But something feels off.
Conversions are up, but profit margins are tighter. Your best-sellers are getting traffic, but so is everything else. The algorithm is "optimising" but you cannot explain what it is actually doing.
Welcome to the reality of Performance Max for SKU-heavy catalogues.
The Flattening Problem
Performance Max was built for simplicity. One campaign. All channels. Let Google figure it out.
For brands with 50 SKUs, this works reasonably well. For brands with 5,000? It creates a fundamental problem: intent flattening.
Here is what happens:
-
High-intent searches get blended with low-intent discovery. Someone searching for "size 10 red leather ankle boots" gets served alongside someone who clicked a Display ad whilst reading the news.
-
Your best products subsidise your worst. PMax optimises for volume, not margin. It will happily spend on products that convert easily but contribute nothing to profit.
-
Nuance disappears. Product-level bidding strategies, audience layering, negative keyword sculpting—all the controls that let you express commercial judgment—are abstracted away.
The Average Product Problem
Google's algorithm is optimised for one thing: conversions within your target ROAS.
This sounds sensible until you realise what it means in practice. Products with:
- Higher margins but lower conversion rates get deprioritised
- Seasonal timing requirements get ignored
- Complex consideration cycles get abandoned early
The algorithm rewards average products that convert predictably. It punishes products that require patience, timing, or nuance.
For a SKU-heavy brand with genuine product differentiation, this is catastrophic.
What the Dashboard Won't Tell You
Pull your Performance Max audit data and look at asset group performance. You will likely find:
- 80% of spend going to 20% of products
- Those products are often your lowest margin items
- Your premium range is barely visible
- Search terms (where visible) show alarming intent dilution
The dashboard shows green. The P&L tells a different story.
Controlling the Inputs
Performance Max is not inherently broken. But treating it as a black box is.
The brands getting results with PMax for large catalogues share common traits:
Feed architecture matters more than campaign settings. Your product feed is your primary control lever. Custom labels, supplemental feeds, and strategic product grouping let you impose structure on the algorithm.
Asset group segmentation by commercial intent. Not by product category. Group products by margin tier, purchase cycle length, or strategic priority.
Negative keyword layering (yes, it's possible). Account-level negatives and placement exclusions give you back some control over where ads appear.
Profit-based success metrics. If you are still optimising for ROAS, you are optimising for the wrong thing. POAS changes the conversation entirely.
The Real Question
The issue is not whether Performance Max works. It is whether it works for your specific commercial model.
If you sell commoditised products with consistent margins and simple purchase paths, PMax probably works fine.
If you sell differentiated products with variable margins, seasonal dynamics, and considered purchases, you need to control the inputs—or accept that Google will optimise for what it wants, not what you need.
What to Do Next
Before accepting that "this is just how it works," get clarity on what is actually happening in your account:
- Map spend to margin at the SKU level
- Audit search term visibility (what you can see)
- Review asset group structure against commercial priorities
- Calculate true POAS, not just ROAS
If the numbers confirm what you suspected, book a strategy call to discuss whether restructuring makes sense for your situation.
This article is part of our series on ecommerce Google Ads optimisation. For a complete audit of your Performance Max campaigns, see our PMax audit service.