Performance Max for eCommerce: The Campaign Type Nobody Understands
Google's Performance Max campaigns promise to do everything. But most eCommerce brands are running them wrong and bleeding money. Here's how PMax actually works and how to make it profitable.

Mark Cijo
Founder, GOSH Digital
Performance Max for eCommerce: The Campaign Type Nobody Understands
Performance Max is the campaign type that Google wants you to use for everything. They push it constantly. Every Google rep will tell you to consolidate into PMax. And the pitch sounds incredible — one campaign that runs across Search, Shopping, Display, YouTube, Discovery, Gmail, and Maps simultaneously, powered by machine learning that optimizes everything automatically.
Sounds too good to be true, right?
It is and it isn't. PMax can be extremely effective for eCommerce. But the way most brands (and agencies) set it up is fundamentally flawed. They turn it on with default settings, feed it some assets, and hope the algorithm figures it out. Then they wonder why their ROAS tanked and their brand search costs doubled.
I have spent the last three years dissecting PMax campaigns across dozens of eCommerce accounts. Let me tell you what is actually going on under the hood and how to make it work.
What PMax Actually Does (Versus What Google Says It Does)
Google's pitch: "PMax uses the power of Google's machine learning to automatically find the best customers across all of Google's channels and serve them the best creative at the right time."
What it actually does: PMax takes your product feed, your creative assets, and your audience signals, and then distributes spend across Google's entire network to hit whatever target you set (usually target ROAS or maximize conversion value).
The machine learning is real. It does optimize. But here is what Google doesn't emphasize:
PMax cannibalizes your brand search. This is the dirty secret. A significant portion of PMax "conversions" come from brand search queries — people who were going to buy from you anyway. PMax swoops in, serves them a Shopping ad for their brand search, and claims the sale. Your standard brand search campaign lose volume, and PMax looks like a hero. But it didn't actually find a new customer. It intercepted an existing one.
PMax loves low-hanging fruit. The algorithm optimizes for the easiest conversions first. That means retargeting recent visitors, capturing branded queries, and finding people who are already at the bottom of the funnel. This makes early results look amazing — but it is not incremental growth.
PMax is a black box. You cannot see which search terms triggered your ads (except through a limited search terms report in insights). You cannot see how much is going to Shopping vs. YouTube vs. Display. You cannot exclude specific placements. Google controls the levers, and you have to trust the machine.
How to Set Up PMax So It Actually Works
Despite these issues, PMax can be powerful. The key is structure and strategy.
Step 1: Protect Your Brand Search
Before you launch PMax, make sure you have a standard brand search campaign running with an exact match on your brand name and close variants. Set a higher priority or bid for this campaign.
Then in PMax, add your brand name as a negative keyword at the campaign level. Google finally allows this (it didn't used to). This prevents PMax from eating your brand search traffic and forces it to find new customers instead.
Why this matters: We have seen PMax campaigns where 40-60% of "conversions" were actually brand searches. After adding brand exclusions, reported PMax ROAS drops, but actual new customer acquisition improves because the campaign is forced to work harder.
Step 2: Structure by Product Margin, Not Product Category
Most brands create PMax asset groups by product category. All skincare in one group. All supplements in another. All apparel in a third.
This is wrong. PMax doesn't care about your categories. It cares about which products convert most easily and profitably. If you have a $20 product with 70% margins and a $100 product with 30% margins, PMax will push the $20 product because it converts more often — even though the $100 product is more profitable for you.
Better approach: Group products by margin tier.
- High-margin products (50%+ margin): Set an aggressive target ROAS
- Medium-margin products (30-50% margin): Set a moderate target ROAS
- Low-margin products (under 30%): Set a conservative target ROAS or exclude them
This way, PMax optimizes within each margin tier appropriately instead of defaulting to whatever converts easiest.
Step 3: Feed Quality Is Everything
Your product feed is the backbone of PMax. If your feed is garbage, PMax is garbage. This is the single most impactful thing you can control.
Titles: Your product titles should include the primary keyword, brand name, and key attributes. "Men's Running Shoes" is bad. "Nike Air Zoom Pegasus 41 Men's Running Shoes - Black/White - Size 10" is what actually matches search queries.
Descriptions: Include detailed, keyword-rich descriptions. PMax uses these to match search intent. Generic descriptions mean generic targeting.
Images: Use high-quality, white-background product images as the primary image. Include lifestyle images in the additional images fields. PMax uses these across Shopping, Display, and Discovery.
Custom labels: Use custom labels to tag products by margin, best-seller status, seasonality, or any other attribute you want to use for campaign segmentation. This is how you group products by margin tier (Step 2).
Pricing and availability: Keep inventory status accurate and updated. PMax penalizes feeds with frequent disapprovals or out-of-stock items.
Step 4: Give the Algorithm Real Audience Signals
PMax "audience signals" are not targeting. They are suggestions. You tell PMax "these are the types of people who buy from us" and the algorithm uses that as a starting point, then expands beyond it.
What to include:
- Customer match lists: Upload your email list of past purchasers. This is the strongest signal.
- Website visitor lists: People who visited your site but didn't buy.
- In-market audiences: People Google identifies as actively shopping for products like yours.
- Custom segments: People who search for specific keywords or visit specific URLs.
What NOT to do: Don't leave audience signals empty. Without signals, PMax starts from scratch and spends your money learning. With signals, it has a head start.
Also, don't only use one audience type. Layer them. Customer match plus in-market plus custom segments gives PMax multiple starting points to learn from.
Step 5: Creative Assets Matter More Than You Think
PMax runs across text-based search, Shopping, image-based display, and video (YouTube). Your creative assets need to work across all of these.
Text assets: Write 5+ headlines and 5+ descriptions. Mix benefit-driven (free shipping, 30-day returns) with feature-driven (organic ingredients, handmade). PMax will test combinations automatically.
Image assets: Provide at least 5 landscape and 5 square images. Mix product shots with lifestyle imagery. PMax uses these for Display and Discovery placements.
Video assets: This is where most brands fall short. If you don't provide a video, PMax will auto-generate one from your images. Auto-generated PMax videos are terrible. They look like a bad slideshow with stock music. Create at least one 15-30 second video that shows your product in use.
The PMax Monitoring Framework
Because PMax is a black box, you need to be extra diligent about monitoring. Here is what to check weekly:
Check 1: The Search Terms Insight
Go to PMax campaign, then Insights, then Search term insights. This shows you the categories and themes of search queries that triggered your ads.
Red flag: If your brand name is a top search category, brand exclusions aren't working or weren't set up. Fix this immediately.
Red flag: If search themes are too broad (like "shoes" or "gifts"), your audience signals and feed quality need work. PMax is casting too wide a net.
Check 2: Asset Group Performance
Check which asset groups are spending and converting. If one asset group is eating 80% of the budget, PMax is over-concentrating on easy wins while ignoring other products.
Fix: Split the dominant asset group into smaller groups. Give other product groups their own budgets through separate PMax campaigns if needed.
Check 3: New vs. Returning Customer Split
PMax reports conversions but doesn't clearly separate new customer conversions from returning customer conversions.
Fix: Set up a new customer acquisition goal in Google Ads. This tells PMax you value new customers more than returning customers. You can even set a different CPA or value for new vs. returning.
Check 4: Placement Breakdown
Go to Insights and check where your PMax spend is going — Shopping vs. Search vs. Display vs. YouTube vs. Discovery.
Red flag: If Display is eating a huge portion of spend with low conversion rates, PMax is chasing cheap impressions on the Display network. This is common in the early weeks as the algorithm "learns." If it persists beyond 4-6 weeks, your audience signals need improvement.
PMax Mistakes That Burn Money
Running PMax as your only campaign. PMax should work alongside standard Shopping campaigns, brand search campaigns, and potentially standard search campaigns for non-brand terms. It is an addition to your strategy, not a replacement for your entire Google Ads account.
Setting target ROAS too aggressively from the start. If you set a 500% target ROAS from day one, PMax restricts itself to only the most conservative, easiest conversions. Start with a lower target (200-300%), let the algorithm learn for 4-6 weeks, then gradually increase.
Not giving it enough budget. PMax needs data to learn. If your budget is too small, the algorithm never gets enough conversion data to optimize effectively. Google recommends at least 15 conversions per week for PMax to function well.
Ignoring the learning period. PMax takes 4-6 weeks to learn. Making major changes during this period (budget changes, target changes, asset changes) resets the learning. Be patient and resist the urge to tinker during the first month.
Using PMax for a tiny product catalog. If you have 10 products, PMax doesn't have enough to work with. It works best with larger catalogs (50+ products) where the algorithm can identify patterns and find the best product-audience matches.
The Bottom Line
PMax is not the magic button Google sells it as. But it's also not the wasteful mess that PMax skeptics claim. It's a powerful tool that requires strategic setup, constant monitoring, and a clear understanding of what it is (and isn't) doing.
The brands winning with PMax are the ones who protect their brand search, structure by profitability, invest in feed quality, provide strong audience signals, and monitor relentlessly. Everyone else is just handing Google money and hoping for the best.
If you're running PMax and not sure if it's actually finding new customers or just claiming credit for existing ones, that's worth figuring out.

Written by Mark Cijo
Founder of GOSH Digital. Klaviyo Gold Partner. Helping eCommerce brands grow revenue through data-driven marketing.
Book a free strategy call →