Google Performance Max in 2026: What We’ve Learned Running It Across Client Accounts
Summary
Performance Max is now the largest single line item in most e-commerce Google Ads accounts, and it remains the least understood. Google presents it as “give us assets and a goal, we’ll handle the rest” — which is broadly how it works, but that framing hides the handful of controls that separate a PMax campaign that scales profitably from one that quietly spends budget on traffic you were already getting for free.
This article covers what we’ve actually learned running PMax across client accounts in 2026: how to structure asset groups, what signals genuinely influence it, the brand-cannibalization problem nobody warns you about, and the controls worth using. It goes deeper on PMax specifically than our Google Power Pack budget guide, which covers how PMax fits alongside AI Max and Demand Gen.
Quick answer: Performance Max works well when you feed it clean conversion data, strong creative assets, and a well-structured product feed — and when you actively manage the two things that quietly wreck it: brand traffic cannibalization (PMax claiming conversions you’d have won anyway through branded search) and asset group structure (one undifferentiated group can’t serve distinct product categories well). Use brand exclusions, segment asset groups by product category or margin, and judge PMax on incremental new-customer acquisition rather than its self-reported ROAS.
What PMax actually is
Performance Max runs a single campaign across Google’s entire inventory — Search, Shopping, YouTube, Display, Discover, Gmail, and Maps — using automated bidding and placement decisions. You provide: - a goal (usually conversions or conversion value) - asset groups (headlines, descriptions, images, video) - audience signals (hints, not targets) - a product feed (for e-commerce)
Google decides where, when, and to whom the ads serve. You cannot see or control placement mix with the granularity of legacy campaign types.
That trade — control for reach and automation — is the whole conversation.
The brand cannibalization problem
This is the single biggest issue we find in PMax accounts, and it’s usually invisible until you look for it.
What happens: PMax will happily serve on branded searches — people searching your brand name, who were going to find you anyway. Those conversions get attributed to PMax, inflating its apparent ROAS and making it look like your best campaign. In reality, you’re paying for traffic you’d have captured through cheaper branded search or organic.
Why it matters: if PMax reports 8× ROAS but a large share of that is branded traffic, its incremental contribution is far lower. Scale budget based on the inflated number and you’re pouring spend into demand that already existed.
The fix:
- Use brand exclusions at the campaign level to keep PMax off your branded terms
- Run a dedicated branded search campaign so you capture that traffic deliberately and cheaply
- Compare PMax performance before and after applying brand exclusions — the drop in reported ROAS tells you how much was cannibalization
This connects to the wider measurement problem in our attribution guide: platform-reported numbers systematically over-credit.
Asset group structure
The most common structural mistake is one asset group covering everything.
Why it fails: asset groups are how you give PMax differentiated creative and signals per segment. A single group serving a whole catalog means generic creative and undifferentiated signals across products with completely different buyers, margins, and price points.
Better structures:
| Structure | When to use |
|---|---|
| By product category | Most e-commerce with distinct categories |
| By margin tier | When profitability differs sharply across products |
| By price band | When buyer intent differs by price point |
| By audience/use case | When the same products serve distinct segments |
Each asset group gets creative, headlines, and audience signals tailored to that segment. This is the main lever you actually control in PMax — use it.
Feed quality: the underrated lever
For e-commerce PMax, the product feed does much of the targeting work. Weak feeds produce weak performance regardless of creative.
Feed checklist:
- Complete, accurate titles that front-load what people search
- Correct and specific product types and categories
- High-quality images meeting Merchant Center standards
- Accurate availability and pricing (mismatches cause disapprovals and lost impressions)
- Custom labels for segmentation (margin tier, seasonality, bestseller status) — these let you split asset groups or apply different strategies
One practical warning we’ve hit: substantial catalog changes (like bulk SKU restructuring) can reset performance history in Merchant Center. Plan large feed changes deliberately rather than mid-peak-season.
Audience signals: hints, not targets
Audience signals tell PMax where to start looking; the algorithm then explores beyond them. They are not targeting.
What genuinely helps:
- Customer lists (your actual purchasers) — the strongest signal you can give
- Website visitor segments with meaningful intent (cart abandoners, product viewers)
- Well-defined custom segments based on real search behavior
What helps less: broad demographic or interest signals — PMax will explore past them anyway.
The pattern mirrors Meta’s Advantage+ (see our Advantage+ guide): feed the algorithm your best first-party data and let it work.
The controls worth using
PMax gives less control than legacy campaigns, but more than most people use:
- Brand exclusions — essential (see above)
- Negative keywords at account level (and campaign-level where available) — keep PMax off irrelevant queries
- Location and language settings — genuinely enforced
- Ad schedule — where demand is time-dependent
- Final URL expansion — turn OFF if you want traffic going only to your chosen landing pages
- Custom labels in the feed — for segmenting asset groups
- Campaign-level budget separation — separate campaigns for genuinely different objectives rather than one giant PMax
How to judge PMax honestly
Because PMax over-claims, judge it on:
- Incremental new-customer acquisition — Google reports new-customer metrics; use them
- Blended efficiency (MER) — does total revenue ÷ total spend improve as PMax scales?
- Performance with brand excluded — the honest read on its non-branded contribution
- Search term insights — review what queries it’s actually serving on; it’s more visible than it used to be
If PMax’s reported ROAS is stellar but blended efficiency isn’t improving as you scale it, cannibalization is the likely explanation.
What we’d recommend doing next
- Check whether brand exclusions are applied. If not, this is probably your biggest hidden issue — apply them and watch what happens to reported ROAS.
- Audit your asset group structure. One group for everything? Split by category or margin.
- Review feed quality — titles, product types, images, custom labels.
- Judge PMax on blended efficiency and new-customer metrics, not its self-reported ROAS.
If you want senior operators restructuring your PMax campaigns, book a $100 audit. We’ll review structure, feed, signals, and cannibalization, and deliver a 90-day plan.
For how PMax fits alongside AI Max and Demand Gen, see our Google Power Pack budget guide. Learn more about our Performance Marketing service.
Frequently asked questions
Does Performance Max cannibalize branded search?
It can, significantly. PMax will serve on branded searches — traffic you’d likely have captured anyway through cheaper branded search or organic — and claim those conversions, inflating its reported ROAS. Apply brand exclusions at the campaign level and run a dedicated branded search campaign, then compare PMax performance before and after to see how much was cannibalization.
How should I structure Performance Max asset groups?
Avoid one asset group covering everything. Segment by product category, margin tier, price band, or audience use case, giving each group tailored creative, headlines, and audience signals. Asset group structure is the main lever you actually control in PMax, and an undifferentiated single group is the most common structural mistake we find.
Do audience signals control who sees Performance Max ads?
No — they’re starting hints, not targets. PMax explores beyond them. The signals that genuinely help are your strongest first-party data: customer lists of actual purchasers and high-intent website visitor segments. Broad demographic or interest signals help much less because the algorithm explores past them regardless.
How do I know if Performance Max is actually working?
Judge it on incremental new-customer acquisition (Google reports new-customer metrics), blended efficiency across your whole account (total revenue ÷ total spend), and performance with brand traffic excluded. If PMax’s reported ROAS looks excellent but blended efficiency doesn’t improve as you scale it, brand cannibalization is the likely explanation.
Should I turn off final URL expansion in PMax?
Turn it off if you want traffic going only to landing pages you’ve chosen. Final URL expansion lets Google send traffic to other pages on your site it judges relevant, which can be useful for discovery but reduces control over the landing experience. For most performance-focused e-commerce accounts, disabling it keeps traffic on optimized pages.
About Pixel Movers: We run Google Ads for brands across UAE, KSA, Pakistan, US, UK, and Canada, managing $2M+ in annual ad spend across Performance Max, AI Max, and Demand Gen. Recent work includes SerMobile (UAE e-commerce, 11× ROAS) and Grace Hospitality (+127% direct bookings). Learn more about us →
