PMax Hasn't Taken Control Away.
It Moved It Upstream.
How do you control outcomes in a system where you don't control inputs? That's the wrong question. The right question is: which inputs still matter?
The Shift Nobody Wants to Admit
Performance Max changed the rules. You can no longer choose keywords, set match types, pick placements, or control which search queries trigger your ads. The levers that defined Google Ads management for fifteen years have been removed.
Most agencies responded in one of two ways: complain about the loss of control, or pretend it doesn't matter and call PMax "automated." Both responses miss the point.
Control didn't disappear. It migrated.
The system still needs instructions. It just reads them from different places now - your feed, your margin data, your audience signals, your creative. If you're not engineering those inputs deliberately, you're not managing PMax. You're spectating.
The real leverage in 2026 isn't inside the Google Ads interface. It's upstream of it.
Feed Quality: Titles, Pricing, Imagery
Your product feed is the single most consequential input into PMax. It determines what Google shows, to whom, in which format, and at what cost. A weak feed doesn't just reduce performance - it corrupts every downstream signal the algorithm uses.
Three attributes matter disproportionately:
- 1Titles. Google uses titles for query matching in Shopping. A title that says "Blue Dress" competes differently from "Navy Midi Dress - Organic Cotton - Size 8-16." The algorithm can only match you to queries your title supports.
- 2Pricing. PMax knows your price relative to competitors. If you're 20% higher with no brand equity to justify it, the algorithm suppresses you before you even see the impression data.
- 3Imagery. In Shopping and Discovery placements, your image is the ad. Low-quality lifestyle shots or generic white-background product photos affect CTR, which feeds back into Quality Score proxies and auction eligibility.
Feed optimisation isn't a "nice to have." It's the primary control mechanism. Everything else is downstream.
SKU-Level Profitability Signals
PMax optimises toward conversion value. If you feed it revenue, it optimises for revenue. If you feed it contribution margin, it optimises for profit. The system doesn't know the difference - it just maximises whatever number you give it.
This is where most accounts leak money. A £200 sale with 15% margin and a £50 sale with 60% margin look very different to the algorithm when you report revenue. The algorithm chases the £200 sale. Your P&L prefers the £50 one.
The Upstream Fix
Pass margin-weighted conversion values into PMax. Segment asset groups by commercial role - Scale, Profit, Liquidate, Gateway. Each group gets a target that reflects its job, not a universal ROAS that treats every SKU the same.
You're not bidding manually anymore. But you're choosing what the algorithm optimises for. That's a more powerful lever than keyword bids ever were.
Audience Layering and Exclusions
PMax accepts audience signals - first-party lists, custom segments, demographic overlays. Most accounts add a token customer list and leave it. That's not audience strategy. That's ticking a box.
The upstream play is more deliberate:
- Exclude existing customers from acquisition campaigns so you stop paying to convert people who'd buy anyway.
- Layer high-value customer signals - repeat buyers, high-AOV segments, subscribers - so the algorithm finds more people like them instead of like your one-time discount buyers.
- Separate brand from non-brand audiences to prevent PMax from claiming brand traffic as new customer acquisition.
The Brand Cannibalisation Trap
Without exclusions, PMax will happily spend 40% of your budget on brand queries and report a 12x ROAS. That looks like performance. It's actually waste - those customers were already coming.
Creative Testing Across Asset Groups
Asset groups are PMax's closest equivalent to ad groups. They contain your creative, your audience signals, and your listing groups. Most accounts run one or two asset groups with every product and a handful of generic assets.
That's not a strategy. That's a default setting.
Deliberate creative testing in PMax means:
- Segmenting asset groups by product category, margin tier, or commercial role - not dumping everything into one group.
- Testing value propositions, not just visual variations. "Free returns" vs "Next-day delivery" vs "Made in the UK" tells you what message converts, not just which image gets clicks.
- Monitoring asset performance ratings as directional signals, not gospel. Google's "Best" and "Low" labels reflect CTR, not profit.
Test the Message, Not the Pixel
Swapping a lifestyle image for a product shot is a visual test. Testing whether "saves you £X per year" outperforms "trusted by 10,000 customers" is a strategic test. The second type moves the needle.
Stop Chasing Search Term Reports
If your agency's main PMax activity is pulling search term reports and adding negatives, they're optimising for a version of Google Ads that no longer exists.
Search terms in PMax are a partial view of one channel in a multi-channel campaign. You're seeing a fraction of the picture and making decisions as if it's the whole thing. That's not optimisation. That's pattern-matching on incomplete data.
Where to Focus Instead
Feed quality. Margin-weighted values. Audience architecture. Creative strategy. These are the inputs PMax actually reads. Search term reports are a rearview mirror on a car that's already driving itself.
PMax isn't taking control away. It's just moving it upstream. The agencies that understand this are the ones delivering commercial outcomes. Everyone else is rearranging deck chairs.
Next Steps
If your PMax management still revolves around search terms and bid adjustments, you're managing the wrong layer. The commercial levers have moved. Your management approach needs to move with them.