Performance Max Is Not a Strategy, It's a Slot Machine (Unless You Feed It the Right Signals)
Google sells Performance Max as "AI that finds customers across all of Google's channels." The marketing is compelling. The reality is more nuanced: PMax is AI that finds the cheapest inventory to hit your ROAS target. Whether that inventory delivers real business value is not its primary concern.
Without constraints, Performance Max will dump your budget into Display networks and cheap placements that drive clicks but not value. It loves "garbage traffic" because it is cheap. A Display impression costs a fraction of a Search click. If both eventually lead to a conversion (even if Display just took the last-click credit), the algorithm sees efficiency.
The problem is that you cannot see where your money is going. PMax is deliberately opaque. Google provides aggregated reporting that obscures placement-level performance. You hit your ROAS target but cannot explain how.
How PMax Actually Works
Performance Max is not one campaign. It is access to all of Google's inventory under a single goal. That includes:
High Value Placements
- • Google Search
- • Shopping tab
- • YouTube (some)
Mixed Value
- • Gmail
- • Discover
- • Maps
Low Value Inventory
- • Display Network
- • Partner sites
- • YouTube shorts
The algorithm's job is to hit your ROAS target with the budget available. It does not care about your brand, your margins, or your customer quality. It cares about finding the cheapest path to a conversion event.
The Attribution Shell Game
PMax takes credit for conversions it touched, even if the customer was already on the path to purchase. If someone saw a Display ad, then searched your brand and bought, PMax claims the sale. This inflates attributed performance while cannibalising genuinely incremental traffic.
What "Feeding the Right Signals" Means
You cannot fully control PMax. That is by design. But you can influence its behaviour by giving it better information about what you actually want.
1. Feed Profit, Not Revenue
Set up conversion value rules or feed gross profit as the conversion value. When the algorithm optimises for profit, it stops celebrating high-volume, low-margin sales.
2. Use Audience Signals Aggressively
Add first-party data: customer lists, website visitors, purchasers. These signals tell the algorithm "people like this are valuable." Without signals, it guesses.
3. Structure Asset Groups by Margin Profile
Do not lump all products into one asset group. Segment by margin: high-margin products get aggressive targets; low-margin products get conservative targets or exclusions.
4. Exclude Low-Value URLs
Use URL exclusions to prevent PMax from sending traffic to blog posts, info pages, or other URLs that do not convert. Force it to focus on product and category pages.
The Asset Group Architecture
This is the difference between PMax as a slot machine and PMax as a structured investment vehicle. The architecture determines behaviour.
Naive Structure (Most Agencies)
Single target ROAS. Algorithm averages everything together. High-margin products subsidise low-margin products.
Commercial Structure
Each asset group has explicit objectives. Algorithm behaviour is constrained by structure.
The Transparency Problem
Even with good structure, PMax remains opaque. Google deliberately limits reporting to prevent you from understanding exactly where your money goes. This is not a bug; it is a feature that benefits Google.
- No placement-level data. You cannot see which websites, apps, or YouTube videos showed your ads.
- No search term reports. Unlike Search campaigns, you cannot see what queries triggered your ads.
- Aggregated asset reporting. You can see which assets perform best, but not in which contexts.
- Limited channel breakdown. High-level splits between Search, Display, YouTube, etc., but not actionable detail.
This opacity is why structure matters so much. You cannot optimise what you cannot see. But you can control the inputs and constraints that shape algorithm behaviour.
Performance Max is not good or bad. It is a tool. Like any tool, it does what you tell it to do. If you tell it to "maximise conversions at 4.0 ROAS" with no further constraints, it will find the cheapest path to that goal. If you tell it what you actually value, through profit signals, margin-based structure, and audience data, it becomes a more useful partner.
The Question to Ask
Ask your agency: "How are our PMax asset groups structured, and what conversion value are we feeding the algorithm?"
If the answer is "one asset group with all products, optimising for revenue", you are running a slot machine. If the answer is "margin-segmented asset groups feeding profit as conversion value", you have a strategy.
Turn PMax From a Slot Machine Into a Strategy
We will audit your PMax structure and show you where the algorithm is being lazy versus where it is being effective.
Get a PMax Architecture Review