Performance Max in 2025: What Is Actually Working

Performance Max has been misunderstood since it launched and mis-managed since it became mandatory. Two years of running PMax across hospitality, e-commerce, and service businesses has given us a clearer picture of what the campaign type actually does well, where it falls down, and the specific configuration decisions that separate profitable campaigns from ones that quietly burn budget.

Here is what is working in 2025.

Asset Group Structure: How Many and How to Organise

The single biggest structural mistake we see is treating a Performance Max campaign like a standard ad group: one asset group, everything in it. PMax uses asset groups to understand what you are advertising and to whom. A single asset group forces the algorithm to serve every asset combination against every audience, which produces a lowest-common-denominator result.

The right structure depends on your business, but the principle is consistent: organise asset groups around products or services that have meaningfully different audiences or conversion intents.

For a restaurant running PMax, we use 2 asset groups: one for walk-in dining with creative oriented around the atmosphere and cuisine, and one for private event bookings with creative oriented around capacity, catering, and function room specifics. The audience signals, the assets, and the landing pages are different for each. The algorithm gets cleaner signals.

For an e-commerce store with multiple product categories, we separate asset groups by category rather than by product. Grouping skincare products separately from supplements, for example, gives the algorithm distinct creative clusters and distinct final URLs to work with. We typically run 3 to 6 asset groups per PMax campaign rather than 1 or 12.

The ceiling matters too. Too many asset groups fragment your budget below the threshold where the algorithm can generate useful learning data. A PMax campaign with $50 per day and 8 asset groups is starving each one. Consolidate.

Audience Signals: What Works and What Google Ignores

Audience signals in PMax are not targeting. This distinction is important and consistently misunderstood. When you add an audience signal, you are telling Google's algorithm where to start looking. The algorithm will expand beyond your signals if it finds conversions elsewhere. You cannot force PMax to show only to your signal audiences.

With that caveat: signals do matter. They accelerate the learning phase and push initial budget toward the most relevant audiences before the algorithm has enough conversion data to self-optimise.

The audience signals that consistently improve early campaign performance:

Your own customer lists. Upload your customer email list and create a similar audiences signal from it. This is the highest quality signal you can give the algorithm. It has seen real buyers. It knows what to look for.

Website visitor segments. Retargeting signals from Google Analytics audiences, specifically purchase-intent segments like cart abandoners or product page viewers, give PMax a high-intent starting point.

Custom intent audiences from search terms. Building a custom segment around competitor brand terms or category search terms (not your own brand) tells the algorithm what commercial intent looks like for your category.

What Google appears to de-weight or largely ignore: interest-based affinity audiences with no purchase history correlation to your account, demographic-only signals without behavioural data attached, and very small lists (under 100 users) that do not provide statistical signal.

Conversion Goal Configuration

PMax optimises toward whatever conversion events you tell it to. This is obvious in theory and frequently misconfigured in practice.

The issue is conversion goal stacking. Accounts that pass multiple conversion actions into a PMax campaign, phone calls, form fills, page views, and purchases all at once, are giving the algorithm a confused objective. If a 2-second page view counts as a conversion alongside a $300 purchase, the algorithm will optimise toward whichever event is most frequent and easiest to generate. Page views win. Purchases suffer.

Clean PMax conversion configuration means: one primary conversion goal that represents actual business value, configured as the target. Secondary conversion goals for reporting only, not for optimisation. For e-commerce, that primary goal is purchase, full stop. For a lead generation business, it is a qualified form submission, not every form submission combined with page views.

If you are running a lead generation campaign and your conversion quality is poor, the most likely explanation is a diluted conversion goal configuration. Fix the goal before adjusting bids.

Budget Allocation and When to Use PMax vs Search

PMax and Search campaigns compete for the same budget and sometimes for the same queries. Google prioritises Search campaigns over PMax for exact and phrase match queries when both campaigns are eligible. But PMax picks up queries that your Search campaigns miss.

The budget decision we recommend: if your Search campaigns are performing well and consuming budget efficiently, do not cannibalise them by running PMax with the same conversion goals on the same budget. Run PMax as an incremental layer with separate budget to capture demand your Search campaigns are not reaching.

If your Search campaigns are limited by budget, the right move is to increase Search budget before adding PMax. PMax's broad matching behaviour is harder to control and wastes budget on irrelevant queries during the learning phase.

PMax is clearly the right choice over Search in 4 situations: you are entering a new market and do not have strong keyword data, you have strong creative assets and want access to YouTube and Display inventory, your product is discovery-driven (people do not search for it specifically), or you are running a shopping campaign and want Merchant Center integration with full funnel coverage.

The Brand Safety Controls Most People Miss

Two controls that most PMax campaigns are not using.

Brand exclusions. PMax will serve on your own brand terms by default. This cannibalises your branded Search campaigns (which typically have better performance data and lower CPCs) and inflates PMax's reported conversion numbers by claiming credit for conversions driven by brand intent. Add your brand name and common variations as a brand exclusion in PMax.

URL exclusions. PMax will send traffic to any page on your website that it considers relevant. If you have pages you do not want driving ad traffic (legal pages, internal tools, thin category pages), exclude them explicitly. Use the URL expansion settings within the campaign to restrict or exclude specific URL patterns.

The placement exclusion list is also worth reviewing. PMax serves on Search, Shopping, YouTube, Display, Gmail, and Maps. If your vertical has particular Display or YouTube inventory that performs poorly (common for B2B and high-consideration purchases), you can submit placement exclusions to Google Ads support. This is not available through the standard interface but can be applied through a support request.

The Bottom Line

PMax is not a set-and-forget campaign type. The accounts that perform well run it with clean asset groups, high-quality audience signals, single focused conversion goals, proper brand exclusions, and regular creative updates. The accounts that perform poorly have one asset group, mixed conversion goals, and no exclusions.

The algorithm is capable. The configuration is the variable you control.

If you want a performance audit of your current Google Ads structure, including PMax configuration, see our Google Ads service or get in touch.