Performance Max Is Only as Smart as the Signals You Give It

Most advertisers set up a Performance Max campaign, add a few headlines, pick a budget, and walk away. Then they wonder why 30% of their spend goes to junk traffic.

The problem isn't Performance Max itself. The problem is that most accounts give Google's AI almost nothing to work with. No audience signals. Barely any search themes. No negative keywords. No first-party data. You're handing the algorithm a blank canvas and expecting a masterpiece.

We've audited thousands of PMax campaigns. 73% have incomplete audience signals. The average account uses 8 out of 50 available search theme slots. And most have zero negative keywords applied.

Here's how to fix that.

Audience Signals Are Not Optional

Audience signals tell Google who your ideal customer looks like. Without them, the algorithm guesses. And it guesses wrong about 30% of the time, based on the accounts we've analyzed.

There are four categories of audience signals, and most advertisers only fill in one or two. You need all four working together.

1. First-Party Data (Your Most Valuable Signal)

This is the data you already own: customer email lists, website visitors, past converters, cart abandoners, high-value purchasers. Upload these as Customer Match lists. Google's AI uses them to find patterns — what these people have in common — and then finds more people who match.

Your customer list needs at least 1,000 active users to be effective. Refresh it every 90 days minimum. Stale lists train the algorithm on outdated patterns.

If you have CRM data with lead quality scores or lifetime value, feed that in too. The difference between optimizing for “someone who filled out a form” versus “someone who became a $10K customer” is the difference between burning money and making it.

2. Custom Intent Audiences

These are audiences built from search terms and URLs that signal purchase intent. Think competitor URLs, high-intent search queries specific to your product, and terms that indicate someone is actively shopping — not just browsing.

A roofing company should target people searching “roof replacement cost” and “roofing contractors near me,” not “how to fix a roof leak yourself.” One signals a buyer. The other signals a DIYer who will never hire you.

3. In-Market Segments

Google's pre-built audiences of people actively researching or comparing products in your category. These are people who have demonstrated recent purchase intent through their browsing and search behavior.

Don't skip these. They're free data that Google has already compiled. Layer them into your asset groups alongside your first-party data.

4. Interest and Demographic Signals

Broader lifestyle and demographic targeting. These are the weakest signals on their own, but they help refine the other three. Use them to exclude audiences that clearly don't fit — if you sell enterprise software, excluding the 18-24 demographic makes sense.

The key principle: one audience concept per asset group. If you wouldn't show every ad in that asset group to every audience signal you've added, split them. Mismatched assets and audiences confuse the algorithm and drag down performance across the board.

Search Themes: You're Leaving 42 Slots Empty

Google gives you 50 search theme slots per asset group. The average advertiser we audit uses fewer than 10. That means you're giving the algorithm almost no guidance on what searches should trigger your ads.

Search themes replaced traditional keyword targeting in PMax. They're not keywords — they're intent signals. Instead of telling Google “bid on [best running shoes],” you're telling it “my business sells running shoes for marathon training.” The algorithm interprets these themes and decides which actual search queries match.

Here's the problem: if you leave 40 slots empty, the algorithm defaults to what it already knows — your branded terms and retargeting pool. You end up paying for clicks you would have gotten anyway, while your competitors who filled all 50 slots capture the high-intent traffic you're missing.

How to Fill Them Effectively

• Start with your website content. What terms describe your products, services, and unique value? These become themes.
• Mine your existing search term data. Look at what's converting in your Search campaigns. Translate those patterns into broader themes for PMax.
• Use your product feed. Product titles, descriptions, and categories are a goldmine of relevant themes that most advertisers never translate into PMax signals.
• Think in intent stages. Some themes should target awareness (“what is [product category]”), some should target consideration (“best [product] for [use case]”), and most should target purchase intent (“buy [product] online”).
• Refresh them regularly. Static themes go stale. The market shifts, seasonal patterns change, and new competitors enter. Review your search themes monthly, at minimum.

The goal is 50 relevant, high-intent themes per asset group. Not 50 variations of the same phrase. Fifty distinct signals that tell the algorithm exactly what kind of traffic you want.

Negative Keywords: The Budget Leak You Can't See

Here's a stat that should make you uncomfortable: the average PMax account we audit wastes 15-40% of total spend on irrelevant search terms. On a $10K/month budget, that's up to $48,000 a year gone.

A dentist paying for “dental school admissions.” A roofer paying for “DIY roof repair.” An attorney paying for “how to represent yourself in court.” These are real examples from real accounts. And the advertisers had no idea because Google doesn't show you most of the search terms triggering your PMax ads.

Account-Level Negatives: Your First Line of Defense

Google now allows up to 1,000 account-level negative keywords that automatically apply to all Search and Shopping inventory across every campaign, including PMax. Start here with the obvious terms that should never trigger your ads:

• Job seekers: “salary,” “hiring,” “careers,” “jobs,” “training,” “certification”
• DIY/free traffic: “free,” “DIY,” “how to,” “tutorial,” “template”
• Irrelevant modifiers: “reddit,” “youtube,” “course,” “wiki”
• Competitor brand terms (if you don't want to bid on them)

Campaign-Level Negatives Go Deeper

Google increased the PMax campaign-level limit to 10,000 negative keywords. Use this for industry-specific exclusions. If you sell premium products, exclude “cheap,” “discount,” “budget.” If you serve a specific geography, exclude other locations that keep triggering your ads.

Important limitation: negative keywords only apply to Search and Shopping inventory. They don't touch Display, YouTube, Gmail, or Discovery placements. That's a separate problem — and one reason you need to monitor your placement reports and exclude junk placements manually.

Timing matters. For new PMax campaigns, let them run 2-4 weeks before adding aggressive negative keyword lists. The algorithm needs time to explore query spaces and learn. During this initial period, add only the most obviously irrelevant terms. After the learning period, do a full search term review and build out your negative keyword lists based on actual data.

Brand Exclusions: Stop Paying Twice for the Same Customer

This one is straightforward, and it's costing you more than you think.

Performance Max aggressively claims credit for branded search terms. Someone searches your company name, clicks your PMax ad instead of your brand search campaign, and PMax reports a conversion with an incredible ROAS. Your reports look great. Your actual customer acquisition is flat.

You're paying premium CPC for people who were already looking for you by name. That's not growth. That's cannibalization.

Set up brand exclusions immediately. Go to your PMax campaign settings, find “Brand exclusions” under additional settings, and add your brand name, product line names, and common misspellings.

Even with brand exclusions applied, the protection isn't perfect. Data across the accounts we manage shows that brand exclusion lists still allow some branded clicks through — particularly on Shopping inventory. But reducing brand cannibalization from 40% of clicks to under 10% is a significant budget recovery.

Also: create a dedicated Brand Search campaign if you don't have one. This lets you control your branded traffic at a fraction of the CPC that PMax charges, and it gives you clean data on actual brand demand versus PMax-attributed conversions.

The Compound Effect

None of these signals work in isolation. A PMax campaign with strong audience signals but no negative keywords still wastes budget on irrelevant traffic. A campaign with 50 search themes but no first-party data still takes longer to learn who your real customers are.

The compounding works like this:

• Audience signals tell the algorithm who to target
• Search themes tell the algorithm what queries to prioritize
• Negative keywords tell the algorithm what to avoid
• Brand exclusions prevent the algorithm from cannibalizing your own traffic

Together, they transform PMax from a black box burning your budget into a channel that actually finds new customers at scale. Separately, each one is a partial fix. Together, they're the difference between a campaign that wastes 35% of spend and one that wastes under 10%.