Google Performance Max is not set and forget. Here are the 6 settings that actually determine whether it works.
Performance Max has been positioned by Google as a campaign type that removes the complexity of manual campaign management — give it your assets, give it your budget, let the AI do the work.
The reality is more nuanced. Performance Max is extremely capable when configured correctly and genuinely wasteful when it is not. The brands who are frustrated with PMax have almost always made the same configuration mistakes. The brands who are happy with it have the same configuration best practices in common.
Six settings. None of them hidden. All of them frequently missed.
SETTING 1: BRAND EXCLUSIONS
This is not a PMax optimisation. It is a requirement. Without a brand exclusion list applied to your PMax campaign, it will serve ads on branded search queries — people searching for your brand name directly — and claim those conversions as PMax revenue.
The consequence: PMax reports an inflated ROAS. You appear to have a high-performing campaign. The performance is partly fake — it is crediting itself for customers who were going to find you through branded search regardless of the ad.
How to configure: Google Ads > Tools > Shared Library > Brand Lists > Create > Add your brand name and variations > Apply to PMax campaign as Brand Exclusion.
After applying: Your PMax reported ROAS will decline. This is correct. You are now seeing what PMax is actually contributing rather than what it is claiming.
SETTING 2: AUDIENCE SIGNALS (NOT TARGETING — SIGNALS)
PMax allows you to add audience signals — starting points that tell the algorithm what kind of people to prioritise while learning. This is not audience targeting. The algorithm is not restricted to these audiences. It uses them as a reference point while exploring.
Without signals, PMax starts cold. It has no information about who your buyers are and has to explore broadly to find them, which is expensive and slow.
Two signals to always provide:
First: your customer email list. Upload a CSV of past purchasers (minimum 100 emails, ideally 500+). This gives PMax a starting model of what your buyers look like.
Second: website visitors from the last 90 days. Create a remarketing audience in Google Ads from all website visitors. Apply as an additional signal.
These two signals consistently reduce the time PMax takes to perform efficiently and improve the quality of the audiences it eventually finds.
SETTING 3: ASSET GROUP STRUCTURE BY PRODUCT CATEGORY
Many brands set up PMax with a single asset group containing all products. This produces averaged performance — the algorithm optimises toward whatever subset of products converts best within the mixed pool, often at the expense of your other products.
Better structure: one asset group per major product category. Each asset group has its own headlines, descriptions, images, and video tailored to that category. Products are assigned specifically to the relevant asset group.
This gives you cleaner data (you can see which category is performing well and which is not), more relevant creative for each category, and more control over how budget is allocated across your range.
SETTING 4: VIDEO ASSETS
Google strongly prefers PMax campaigns with video assets over those without. Campaigns without uploaded video will have video auto-generated by Google — and auto-generated videos are typically poor quality, frequently embarrassing, and almost always off-brand.
You do not need a high-production brand video. A 15–30 second product demonstration filmed on a phone, with text overlays and your product clearly visible, is sufficient. It will dramatically outperform Google’s auto-generated alternative.
If you genuinely have no video: monitor what Google has auto-generated and replace it as soon as possible. Check it regularly — Google can regenerate video assets after changes to your asset group.
SETTING 5: CONVERSION VALUE RULES FOR MARGIN-ADJUSTED BIDDING
PMax optimises toward “maximise conversion value” by default — weighted equally across all products. If your product range has different margin levels (40% margin on one product, 65% margin on another), you are letting the algorithm treat a low-margin sale the same as a high-margin one.
Conversion value rules allow you to apply a multiplier to conversions from specific products or categories. Set a higher multiplier for high-margin products (e.g., 1.3×) and a lower multiplier or none for low-margin products.
The result: PMax progressively shifts its targeting toward the buyer profiles that purchase your most profitable products. Revenue may stay constant or even decline slightly while gross profit improves.
SETTING 6: TARGET ROAS (SET CORRECTLY)
PMax should not run on “Maximise conversion value” indefinitely. Once you have 50+ monthly conversions, switching to Target ROAS gives the algorithm a profitability constraint.
The target to set is your minimum viable ROAS — not an aspirational figure, not the ROAS you want to achieve, but the minimum level at which the channel is profitable.
Minimum viable ROAS = 1 ÷ (1 − gross margin %)
Example: 55% gross margin → minimum ROAS = 1 ÷ 0.45 = 2.22× (222%)
Set your Target ROAS at 50–100% above this minimum (to ensure profitability after fixed costs). Start conservatively — a Target ROAS set too high will restrict delivery. Increase it by 10% every two weeks until you find the level that balances efficiency with sufficient impression volume.