Performance Max is Google’s favorite campaign type for one reason, it gives Google maximum control.
That can be a win. It can also be an expensive way to buy your own brand traffic, serve ads for irrelevant queries, and call it “automation.”
If you are running PMax the default way, one campaign, one asset group, a handful of headlines, and “let the algorithm learn,” you are not doing strategy. You are renting Google’s guesswork.
This guide is how I set up Performance Max campaigns when revenue matters. It is written for ecommerce owners and marketing leads who want outcomes, not dashboards.
For the broader ecommerce paid strategy layer, start with my pillar: Google Ads for Ecommerce. If you want this built and managed, the work lives on my Google Ads management page.
What Performance Max Actually Is (and Why Google Pushes It)
Performance Max is a goal-based campaign type that can serve across Google inventory, including Shopping, Search, YouTube, Display, Discover, Gmail, and Maps. You provide assets, audiences, and optionally product feeds. Google decides where to spend.
Google pushes PMax because it simplifies account structure and increases inventory access. It also removes visibility in places where advertisers used to enforce discipline.
That means your job changes.
You are not “optimizing keywords” inside PMax. You are shaping inputs and guardrails so Google’s system can find conversions without spending money on nonsense.
If you want the official framing, start with Google’s Performance Max documentation:
https://support.google.com/google-ads/answer/10724817
Read it, then ignore the parts that imply Google’s defaults are built for your business. They are built for Google’s scale.
The First Question, Should You Run PMax at All?
PMax works best when:
- you have enough conversion volume for smart bidding to stabilize
- your product feed is clean and complete
- your catalog has enough inventory and margin that you can tolerate testing
- you have solid tracking, ideally with real purchase value, not fake micro conversions
PMax is a bad idea when:
- conversions are scarce and noisy
- tracking is broken or inflated
- product data is thin
- the business is still figuring out product market fit
If you are not sure, do not start by launching PMax. Start by auditing tracking and product feed quality. Then decide.
The Most Common PMax Failure Mode, “It Works” but It Isn’t Profitable
Here is what “working PMax” often looks like in a real account:
- ROAS looks acceptable at the campaign level
- a huge share of conversions are branded searches
- spend leaks into low-intent placements
- product mix is wrong, it sells low-margin items because they convert easily
- the business thinks it scaled, but net profit did not move
PMax can absolutely drive revenue. The problem is that many accounts call revenue “success” while ignoring margin, customer quality, and incrementality.
If you sell a product with thin margins, you do not have room for sloppy distribution.
Asset Groups, How to Structure Them So They Mean Something
Asset groups are not ad groups. Treating them like ad groups is how people create chaos.
I structure asset groups around one thing:
One commercial intent cluster.
Examples that make sense:
- One asset group per top category (not per SKU)
- One asset group per use case (for example “gift,” “travel,” “office”)
- One asset group per price tier when tiers behave differently
Examples that usually do not make sense:
- One asset group per product, unless the product is a true hero and you have volume
- One asset group for the entire catalog, unless the catalog is tiny
The goal is to keep the system’s learning clean. If one asset group mixes very different products, prices, and conversion rates, your bidding signal becomes noisy.
The Minimal Asset Group Build I Use
For each asset group, I want:
- 1 to 2 strong landing pages (category or curated collection pages)
- a feed segment that matches the intent, using listing group controls
- creative assets that speak to that intent, not generic “shop now” lines
- audience signals that match the buyer profile
If you do not have category pages that can convert, fix that first. Ecommerce paid media does not exist in isolation, it needs landing pages that do a job. My ecommerce SEO service is the organic side of that, but the landing page rules apply to paid too.
Listing Groups, the Control Layer Most Accounts Ignore
If you run PMax for ecommerce and you do not manage listing groups, you are giving Google permission to sell whatever is easiest.
That is rarely what you want.
Listing groups let you segment products by:
- category
- brand
- custom labels
- price range
- best sellers vs long tail
- margin tiers (if you label them)
The practical move is to add custom labels in Merchant Center or the feed so you can segment by:
- margin tier
- inventory status
- seasonal priority
- new arrivals
- hero products
Then you allocate budget accordingly.
Google’s Merchant Center feed quality matters more than most advertisers want to admit. Feed fields, titles, GTINs, product types, and attributes are not “SEO chores,” they are matching signals for Shopping. Google’s official product data specs are the canonical reference:
https://support.google.com/merchants/answer/7052112
If your feed is thin, PMax cannot be precise. It will still spend. It will just spend dumb.
The One-Campaign Trap, Why “Simpler” Usually Means “Messier”
Google’s default advice tends to push you toward fewer campaigns. In practice, one PMax campaign that covers everything often creates two problems:
- You lose control of product mix, because the system drifts toward whatever converts easiest.
- You lose clarity in measurement, because you cannot separate categories, margins, or intent clusters cleanly.
If your catalog is tiny, one campaign can work.
If your catalog is real, apparel, home, beauty, electronics, a one-campaign structure usually becomes a budget blender. It mixes high-margin winners with low-margin volume, and it spends wherever it can find conversions fastest.
The fix is not “make ten campaigns for fun.” The fix is to split where the business logic actually changes:
- split hero categories from long-tail inventory
- split high-margin from low-margin
- split seasonal pushes from evergreen categories
Then you use listing groups and custom labels so you can enforce that split.
If you want a simple starting point, build one PMax campaign for your highest-value category cluster, get that stable, then add the next cluster. Do not launch five at once and pretend you can read the data cleanly a week later.
Audience Signals, What They Are and What They Are Not
Audience signals are not targeting in the old sense. They are guidance.
The right way to think about it:
- You are giving the system a starting point.
- The system can expand beyond it.
That is why audience signals can help early learning, but they are not a replacement for structure, feed quality, and conversion integrity.
Audience Signals That Actually Help
I prefer:
- Customer Match lists (past purchasers, high LTV segments)
- Remarketing lists segmented by intent (viewed product, added to cart, started checkout)
- Custom segments built from competitor and category search behavior
I keep it simple. If you throw 20 signals into one asset group, you are not being smart. You are being noisy.
Search Themes, What They Do and How I Use Them
Search themes are Google’s way of letting you nudge the system toward a query family.
They can help. They are not a guarantee.
I use search themes when:
- the asset group represents a clear intent cluster
- the landing page and feed segment match that intent
- I want to speed up discovery of the right queries
I do not use search themes as a band-aid for weak feed data or generic landing pages. The system will still drift.
Creative Assets, How I Avoid the “Generic PMax Look”
Most PMax creative looks like it was assembled by a template generator. That is a problem because it creates two failures:
- users ignore it, because it looks like every other ad
- the system struggles to learn which angle actually drives qualified purchases
My rule is simple, each asset group gets creative that matches the category intent.
For a premium product category, I want:
- one clear product image with clean context
- one lifestyle image that makes the use case obvious
- one short video, even if it is basic, that shows product scale and real texture
And I want copy that does one job: qualify the right buyer and repel the wrong one.
If your offer is premium, do not write copy that tries to attract bargain hunters. You will pay for clicks that do not convert and then blame the algorithm.
If you need the wider paid strategy view, including how I map offer, margins, and landing pages, that is in Google Ads for Ecommerce.
Brand Control, Stop Paying for Your Own Name
This is where PMax gets political.
If you let PMax run unchecked, it often cannibalizes branded search. That can inflate ROAS and make the campaign look like a hero, while incremental value is lower than reported.
What I do instead:
- Run a dedicated branded Search campaign and protect it.
- Use brand exclusions where available and appropriate.
- Monitor search terms and placement behavior, even if reporting is limited.
Google’s documentation on brand exclusions and account-level controls changes over time. Start with the official PMax help pages and verify what your account supports:
https://support.google.com/google-ads/answer/10724817
If you are measuring success, separate brand and non-brand impact. Otherwise you will “scale” by spending more money to capture traffic you already owned.
Negative Keywords, the Lever Everyone Wants and Google Limits
PMax historically limited negative keyword control compared to Search campaigns. That is by design.
Still, you do not run PMax without a negatives strategy.
Here is what I do:
- Build a negative list from obvious irrelevance (jobs, free, cheap, DIY, used, templates, and any industry-specific garbage)
- Watch for query families that burn spend without revenue
- Use account-level negative keyword lists where possible
Google’s negative keyword documentation is the baseline reference:
https://support.google.com/google-ads/answer/2453972
If your business is sensitive to irrelevance, for example luxury, B2B, regulated categories, you need this discipline. PMax will not magically infer your standards.
Incrementality, the Only Test That Stops Self-Congratulation
If you have enough volume, you should run at least one incrementality check.
You do not need a complicated experiment to get value. You need one clean question:
Is PMax driving new revenue, or is it mostly capturing demand you already had?
The simplest approach I use with ecommerce clients is a controlled adjustment:
- keep branded Search protected
- reduce PMax budget for a short window while monitoring non-brand revenue and product mix
- then restore budget and compare changes, focusing on non-brand and new customer signals
This is not perfect science. It is still better than trusting one blended ROAS number.
If the business is smaller and cannot run tests without pain, the fallback is to tighten brand control, tighten listing groups, and measure category-level outcomes instead of the campaign average.
Budget and Bidding, How I Avoid “ROAS Theater”
Smart bidding is only as good as your conversion signal.
If your conversion tracking is wrong, PMax will optimize toward the wrong thing at scale. That is the fastest way to burn money.
Before I touch bidding, I confirm:
- purchase value is accurate
- refunds are not counted as purchases
- conversion duplication is not inflating volume
- primary conversion actions are not polluted with micro events
Then I pick a bidding strategy that matches maturity:
- If the account has stable purchase volume, Target ROAS can work.
- If data is still stabilizing, Maximize Conversion Value can be safer early.
I do not set an aggressive target on day one and then complain the system cannot hit it. That is not how learning works.
If you want a clean baseline for how Google frames smart bidding, read the official docs:
https://support.google.com/google-ads/answer/2471185
Landing Pages, PMax Is Only as Good as Where It Sends People
If your landing pages are weak, PMax cannot rescue you.
For ecommerce, I prefer landing pages that do one of these:
- a category hub that can rank and convert
- a curated collection that maps to a use case
- a hero product page that is built to convert
If your product pages are thin and generic, fix that before you scale paid spend. I covered the product layer in my product page SEO guide. Organic and paid are different channels, but they share a core truth, bad pages do not convert.
Reporting, What You Can and Cannot Trust in PMax
PMax reporting is limited. That is not a bug. It is the product.
So I focus on what is reliable:
- overall revenue and margin impact
- product mix
- new customer share (when available)
- branded vs non-branded split (with separate campaigns and query analysis)
- incrementality tests when the business is large enough
What I do not do is obsess over tiny creative-level changes without first controlling product segmentation and brand cannibalization. Creative matters, but it is not the first fix in most PMax accounts.
New Customer Acquisition, Useful When You Read It Correctly
Some accounts use “new customer acquisition” settings inside PMax. It can be useful, but only if you define “new” correctly and you do not confuse a reporting label with incremental growth.
I treat it as a directional signal:
- If new customer share rises while non-brand revenue stays healthy, the campaign is likely expanding reach.
- If new customer share rises but profit drops, you are probably buying low-quality first purchases that never repeat.
If the business has repeat purchase economics, new customer settings can be a lever. If it is a one-time purchase product, you should be careful. New customers do not matter if the margin does not survive the acquisition cost.
How I Audit a Performance Max Campaign (My Practical Checklist)
If I take over a PMax account, I start with a short checklist. It surfaces the real problems fast.
1. Conversion Integrity
- Are purchases tracked correctly?
- Is purchase value accurate?
- Are there duplicate conversion actions?
- Is the primary conversion action the one that matters?
If this is wrong, nothing else matters.
2. Brand Cannibalization Risk
- Is there a dedicated branded Search campaign?
- Is PMax capturing branded conversions?
- Are brand terms separated in reporting?
If brand is mixed in, reported ROAS can lie.
3. Feed Quality and Segmentation
- Are product titles descriptive and accurate?
- Are GTINs present where applicable?
- Are custom labels used for margin and priority?
- Are listing groups controlled, or is everything lumped together?
If the feed is thin, the system cannot be precise.
4. Asset Group Structure
- Do asset groups map to clear intent clusters?
- Are landing pages aligned with those clusters?
- Are audience signals coherent, not random?
If structure is vague, the system learns slowly and drifts.
5. Waste Signals
- Spend concentration on low-margin products
- High spend with weak profit impact
- Low-intent query families showing up in insights
Then I tighten controls. Not by “optimizing” endlessly, but by fixing the inputs.
If this is what you want a professional to run, it lives on my Google Ads management page.
The SEO Feedback Loop, Why PMax Gets Better When Your Organic Data Is Clean
PMax improves when your site is structured well.
Why:
- category pages clarify product taxonomy
- product pages have better structured data and clearer use cases
- internal linking routes users and clarifies intent
- conversion paths are cleaner
This is why I like combining paid and organic thinking. You can use organic query data to shape PMax search themes and landing pages. You can use paid search term signals to identify which category pages deserve stronger SEO investment.
If you are treating paid and organic as separate planets, you are leaving money on the table.
Useful References
FAQ
What are Performance Max campaigns and how do they work?
Performance Max campaigns are goal-based Google Ads campaigns that can serve across multiple Google channels using assets, audience signals, and product feeds. Google allocates spend automatically based on conversion goals. The practical way to make them work is to control inputs, feed quality, conversion tracking, and brand cannibalization, not to expect “automation” to do strategy.
How do I structure asset groups in Performance Max?
Structure asset groups around one commercial intent cluster, usually a top category, a use case, or a price tier. Keep landing pages and feed segments aligned to that cluster. If you mix unrelated products in one asset group, learning gets noisy and spend drifts toward what is easiest to sell, not what is best for your business.
How do I stop Performance Max from spending on my brand terms?
Protect branded search with a dedicated Search campaign, then use brand exclusions and account-level controls where available. Also separate branded and non-branded measurement so ROAS does not get inflated by demand you already own. If your account cannot apply the control you want, the fallback is campaign structure and measurement discipline.
Do Performance Max campaigns need negative keywords?
Yes. Even with limited control compared to Search, you still need a negatives strategy for obvious irrelevance and query families that burn spend. Build a base negative list, then expand it based on spend and conversion quality. If the business is premium or regulated, negatives and brand control are non-negotiable.
Why is my PMax ROAS high but profit not improving?
Because ROAS can be inflated by brand cannibalization, low-margin product mix, or conversion tracking issues. Audit conversion integrity, separate brand and non-brand, and segment products by margin using custom labels and listing groups. Then measure profit impact, not just revenue.
If you recognize these problems, the next step is not tweaking headlines. It is building a Performance Max system with clean inputs, feed discipline, and measurement you can trust. That is what I do on my Google Ads management page.
About the Author
Luciano Bonanno is an independent SEO and Growth Consultant with 18 years of experience. Founder of SameAPI and DeLeak.co. Book a strategy call →