Google Performance Max for E-commerce: What Actually Works

Performance Max is now the default Google Ads campaign type for most e-commerce brands. It's also the most opinionated, opaque, and easy-to-screw-up campaign type Google has ever launched. After running PMax across dozens of UK e-commerce accounts, this is the structure and approach we've settled on as our default starting point.

Why PMax now

Standard Shopping campaigns still exist, but Google has progressively starved them of inventory and creative options. Performance Max runs across all Google surfaces (Search, Shopping, YouTube, Discovery, Gmail, Maps) from a single campaign, using Google's machine learning to optimise across all of them.

This is a strength and a weakness. The strength: when it works, you reach customers Google has signal on that you'd never reach manually. The weakness: you have very limited visibility into where your spend goes and which products get shown.

How to structure asset groups

The biggest decision in PMax is how you split products into asset groups. Wrong structure = wasted spend. Right structure = clean reporting and machine learning that can actually optimise.

Default starting structure

For brands with 100+ SKUs, we start with this:

• Best Sellers asset group — top 10-20% of products by revenue from the last 90 days. These have the data Google needs to optimise quickly.
• Category asset groups — one per major product category (typically 3-6 groups).
• New Products asset group — products launched in the last 60 days, where you want Google to push for early data.
• Long Tail asset group — everything else, lower bid, "let it find its level."

For brands with under 100 SKUs, just split by category. Don't over-segment — PMax needs scale to optimise.

Audience signals (the under-used lever)

"Audience signals" tell PMax which customers you think will convert. Critically, they're hints, not constraints — Google can ignore them. But brands that ignore audience signals entirely leave significant performance on the table.

The four signal types, ranked by impact:

1. Customer match (uploaded customer lists). Your past purchasers. The highest-converting signal you have. Upload your full customer list and let Google find lookalikes.
2. Custom segments (search intent + URLs). Tell Google "customers searching for X" or "customers who've visited [competitor URL]."
3. In-market audiences. Google's pre-built segments of people actively shopping.
4. Affinity audiences. Broad interest segments. Use last — these are noisy.

Assets that actually convert

Each asset group needs a stack of creative. Minimum requirements (Google rejects asset groups with less):

• 5 headlines (30 characters each)
• 5 long headlines (90 characters each)
• 5 descriptions (90 characters each)
• 20 images (mix of product, lifestyle, and offer-focused)
• 5 logos (square + rectangular variants)
• 1+ video (Google generates one for you if you don't upload, but the auto-generated ones are weak)

What we've found works

From running PMax across categories from baby products to industrial packaging:

• Product images on white background consistently outperform lifestyle on PMax (different from social — don't transfer assumptions).
• Headlines with prices (when stable) lift CTR significantly. "£24.99 Premium Wipes" beats "Premium Wipes."
• Urgency language ("Limited Stock", "Free UK Delivery") underperforms on PMax compared to Meta. Don't overdo it.
• UGC video assets consistently outperform polished commercial videos. Authenticity beats production value.

Budget allocation

PMax requires more daily budget than you'd think to optimise effectively. Below these floors, the algorithm doesn't have enough conversion data to learn:

If you can't hit these floors, run a single asset group with all products and don't try to segment. Splitting a £50/day budget across 4 asset groups gives Google £12.50/day to optimise each — too little to find signal.

Reporting inside a black box

The frustration with PMax is reporting. You can't easily see what's working at the placement level (Search vs Shopping vs YouTube), and you can't see which specific keywords converted. Here's what to track instead:

Per asset group (weekly)

• Spend
• Revenue
• ROAS
• Conversion volume
• Top-performing assets (which images, headlines)

Account level (weekly)

• Blended new vs returning customer ratio
• Search Terms (PMax has limited but useful search term data — check weekly and add negative keywords)
• Asset group impressions distribution

Common PMax mistakes

1. Brand traffic eating budget

PMax aggressively bids on brand searches by default (your brand name as a keyword). It looks great in reports — high ROAS, low CPC — but you're paying for customers who would have come to you organically. Use brand exclusion lists (request from Google rep) or a separate brand-only Search campaign with negative keywords pointing to PMax.

2. Over-segmenting too early

Splitting into 8 asset groups before you have the conversion volume is the #1 reason for "PMax isn't working" complaints. Start consolidated, scale up.

3. Setting target ROAS too tight from launch

PMax needs a learning period (typically 2-3 weeks). If you launch with tROAS = 4.0× and never see traction, you're starving it. Launch with "Maximise Conversions" or a wider tROAS, let it gather data, then tighten.

4. Ignoring the negative keyword list

You can add negative keywords at the account level. Common ones for e-commerce: "free", "cheap", "review", "vs", competitor brand names. Without these, you waste spend on intent that won't convert.

"PMax rewards patience and data. Brands that change settings every 3 days get worse results than brands that let it run for 4 weeks before optimising."