Performance Max is now the default recommendation for almost every campaign type in Google Ads. Shopping, lead gen, local, app, Google pushes everything into PMax. And yet, most practitioners I talk to run it with a mix of resignation and frustration. The interface gives you almost nothing to work with, the reporting is deliberately opaque, and diagnosing problems is almost impossible. This guide is about the levers that actually exist, and how to use them.
Why PMax Feels Like a Black Box
This is not accidental. Performance Max consolidates multiple ad formats, Search, Display, YouTube, Discovery, Gmail, Maps, Shopping, into one campaign that Google's auction system distributes across all of them simultaneously. The aggregated reporting hides what is running where. You see total conversions. You do not see how many came from YouTube versus Search versus Display, which asset combinations drove results, or which placements ate your budget.
Google's argument is that fragmented channel reporting encourages bad decisions, that optimizing each channel independently misses cross-channel synergies. There is some truth to that. The practical problem is that opacity cuts both ways: it also prevents you from catching waste, flagging irrelevant placements, or understanding which creative direction is actually working.
The good news is that there are more controls than the interface suggests. They are just not surfaced prominently.
Asset Groups Are Not Ad Groups, Stop Treating Them the Same Way
The most common structural mistake I see: Practitioners create one asset group, load all their assets into it, and let PMax run. The result is a campaign with no internal differentiation, Google serves whichever combination it predicts will convert, across all inventory, with no signal about what you actually want to prioritize.
Asset groups are the primary way you communicate intent to PMax. Each asset group should represent a coherent combination of: a specific product or service, a defined audience, and a dedicated landing page. When these three things are aligned within a single asset group, you give the AI a clear signal about what a conversion looks like for that segment.
In practice, this means creating separate asset groups for different service lines or product categories, not lumping everything together because it seems simpler. A legal services firm should have separate asset groups for personal injury, corporate law, and family law, not one asset group for "legal services." An e-commerce brand should separate by product category, with each group pointing to the relevant category page rather than the homepage.
One specific tactic worth testing: the "pure Shopping" approach. If you create a PMax campaign with Shopping feed assets only, no image, video, or text assets, the campaign behaves significantly more like Smart Shopping. It focuses on Search and Shopping inventory, produces better reporting, and is considerably more controllable. For e-commerce accounts that primarily need Shopping performance without the Display and YouTube noise, this structure often outperforms a fully-loaded PMax campaign.
Search Themes: Your Replacement for Keywords
Search themes were introduced in late 2024 and remain underused. They are not keywords. They are signals, up to 25 per asset group, that tell Google's AI what search territory you want to prioritize. The AI can still venture outside them, but search themes meaningfully steer traffic toward the query space you care about.
The most practical way to populate search themes is to pull your best-performing search terms from your existing Search campaigns. Sort by conversion volume over the last 90 days and take the top 20 to 25 that represent distinct intent categories. These terms have already proven they convert in your account, feeding them as search themes into PMax gives the AI a head start with validated signal rather than having it explore from scratch.
If you are launching PMax without existing Search campaign data, use the search themes to define the intent territory explicitly. "Performance marketing agency Athens," "Google Ads management," "paid media strategy" are all legitimate themes for a performance marketing consultancy. Be specific. Broad terms like "digital marketing" generate volume without direction.
Audit your search themes monthly. If the Insights tab shows PMax is spending heavily outside your intended territory, revise the themes and add negatives to redirect.
Audience Signals: The One Input That Changes Everything
PMax uses audience signals as a starting point for its customer model, not as a targeting constraint. It will expand beyond your signals if it identifies conversion opportunities elsewhere. But the quality of your input signals determines how quickly the AI finds a productive pattern and how closely that pattern resembles your actual customer.
The hierarchy matters. Customer Match lists from your CRM are the highest-quality signal by a significant margin. They tell PMax exactly what your converted customers look like. If you have 500 or more email addresses from past customers, upload them before the campaign launches. Wait 24 hours for Google to process and match them before activating PMax, running the campaign before Customer Match has processed means the AI starts with weaker signal.
Website remarketing audiences come second, ideally with a 60-day or 90-day window to provide sufficient volume. GA4 audiences tied to high-intent page visits, service pages, pricing pages, contact pages, are more valuable than general site visitors.
What not to use: generic Google interest categories as primary signals. "Business professionals" or "marketing enthusiasts" are broad enough to be nearly meaningless as training signals for a B2B account. They dilute the model rather than sharpening it. Use them at most as secondary observation audiences if you want data, not as primary signals you are actively pushing the AI toward.
The Three Exclusion Layers Nobody Sets
Most PMax campaigns launch without any exclusions. This is one of the most expensive configuration oversights in paid media.
Brand exclusions. Go to Campaign settings and find Brand exclusions. If you run separate brand Search campaigns, which you should, add your brand here. Without brand exclusions, PMax will target your brand terms, claim credit for those conversions, and report an inflated ROAS while your separate brand campaign simultaneously bids on the same traffic. The result is budget waste and attribution chaos. This setting is non-negotiable for any account with brand campaigns.
Placement exclusions. PMax serves Display and YouTube ads by default. The default placement pool includes mobile app categories that perform extremely poorly for almost every B2B and lead gen account: gaming apps, entertainment apps, children's content, and app categories with no contextual relationship to your offering. Set placement exclusions at the account level, they apply to all campaigns, including PMax, and specifically exclude "Games," "Entertainment" (mobile apps subcategory), and "Online Communities" unless you have a specific reason to be there. Account-level exclusions persist across new campaigns, so you only need to do this once.
URL expansion exclusions. PMax can route traffic to any page on your site when URL expansion is enabled. Thank-you pages, form confirmation pages, thin category pages, PDFs, and any page you would not deliberately send paid traffic to should be explicitly excluded. Go to Campaign settings → URL expansion and add these pages. Left unchecked, URL expansion will drive traffic to pages that look like destinations to Google but do nothing for your conversion goals.
Reading the Insights Tab Like a Practitioner
The Insights tab is the closest thing PMax has to transparent reporting, and most accounts barely look at it. Here is what to actually extract from it.
The Search categories report shows you the general query territory PMax is targeting. It is not keyword-level, you will see categories like "performance marketing services" or "Google Ads management" rather than individual search terms, but it tells you whether the campaign is operating in the right space or drifting into irrelevant territory. Check this weekly for the first four weeks of a new campaign.
The Asset performance ratings show individual assets rated Best, Good, Low, or Learning. "Low" assets are dragging down the combination pool. Replace them with fresh creative rather than waiting, an asset rated Low will continue running but will be served less frequently. The issue is that a Low asset in rotation still consumes impressions that could go to better-performing combinations.
The Audience insights report shows the demographic and interest profile of your converting users. This is genuinely useful for two things: verifying that PMax is converting the audience you intended, and identifying audience characteristics you can feed back into your Customer Match strategy and broader campaign targeting.
PMax Alongside Search: The Budget Trap
When PMax and Search campaigns target overlapping query territory, PMax wins the auction in most cases. Google's system prioritizes PMax over Search for the same query when both are active. Add PMax without structuring your budget carefully and it will gradually consume the query volume your Search campaigns were handling, with Search impression share declining and no clear explanation in the interface.
The correct approach is to treat PMax as an incremental budget layer, not a replacement for Search. Keep your Search campaigns funded at their current level. Allocate a separate, additional budget to PMax and measure its performance against a clear incremental revenue target. If PMax is generating genuine new conversions rather than cannibalizing Search, that will be visible in your CRM data even if the platform-level attribution does not show it cleanly.
Use brand exclusions and negative keyword lists at the campaign level to enforce the boundary between Search and PMax wherever possible. The boundary is not perfectly enforceable, PMax will still win some queries you might prefer to handle in Search, but exclusions reduce the overlap significantly.
The 6-Week Rule and Why You Cannot Shortcut It
PMax requires a learning period of roughly four to six weeks before performance stabilizes. During this period, the AI is building its conversion model, calibrating audience signals, and testing asset combinations. Making structural changes, pausing asset groups, significantly altering targets, changing bidding strategy, within this window resets the learning period. Most accounts that report poor PMax results paused or significantly modified the campaign in the first two weeks based on early data.
When you first launch PMax, loosen your Target CPA or Target ROAS by 10 to 15 percent relative to your historical benchmark. This gives the system room to find new conversion territory without the bidding algorithm restricting reach during the period when it most needs data. After the learning phase stabilizes, tighten your targets back gradually, increments of 10 percent with two-week intervals between changes.
If the campaign is genuinely underperforming after six weeks, diagnose systematically: Are audience signals in place? Are brand exclusions set? Are search themes directing traffic appropriately? Is the conversion tracking firing correctly? The answer to underperformance is almost never to reduce the budget or pause the campaign in the first month, it is to audit the configuration.
Your 30-Day PMax Action List
- Week 1, Day 1: Set brand exclusions. Set account-level placement exclusions for gaming and entertainment app categories. Add URL expansion exclusions for thank-you and confirmation pages.
- Week 1, Day 2: Upload Customer Match list from your CRM. Wait 24 hours before activating the campaign.
- Week 1, Day 3: Structure asset groups by product line or service category. Create a dedicated landing page for each asset group, do not point everything at the homepage.
- Week 1, Day 4: Pull your top 20 converting search terms from existing Search campaigns. Add them as search themes, distributed across the relevant asset groups.
- Week 1, Day 5: Set your bidding target 10 to 15 percent looser than your historical CPA or ROAS benchmark. Launch.
- Week 2: Check the Insights tab. Are search categories on-target? Replace any assets rated Low. Do not make structural changes yet.
- Week 3: Review placement report if accessible. Add any irrelevant placements to the exclusion list. Check Search campaign impression share for signs of cannibalization.
- Week 4: First full performance review. Compare CRM revenue data to platform-reported conversions. Begin tightening your bidding target by 10 percent if performance is on track.
PMax is not going away. The accounts that do well with it are not the ones that trust Google's defaults, they are the ones that understand where the real controls are and use them consistently. Once PMax is running well, the next challenge is measuring its contribution honestly, platform-reported ROAS will almost always overstate what PMax is actually delivering.