What to Control

Campaign-Level Negative Keywords

This was the single most requested feature for PMax and it's finally here. Use it aggressively.

Pull your search term reports weekly and add negatives for:

- Brand terms (if you're running a separate brand campaign)

- Competitor terms (unless you're intentionally bidding on them)

- Irrelevant informational queries

- Low-intent broad matches that waste budget

This alone can improve ROAS by 10-25% on accounts that haven't been managing it.

Asset Group Structure

Don't throw everything into one asset group and hope for the best. Structure your asset groups around:

- Product category or collection

- Margin tier (high margin products get their own group with higher ROAS targets)

- Seasonal vs evergreen products

- Best sellers vs long tail

Each asset group should have its own creative assets, audience signals and product feed subset. This gives Google clear signals about what to serve where.

Audience Signals

Audience signals don't restrict targeting but they give Google a starting point. Layer in:

- Your customer lists (purchasers, high-value segments)

- Website visitors by behaviour (product viewers, cart abandoners)

- Custom intent audiences built around your top-performing search terms

- In-market and affinity audiences relevant to your category

The better your signals, the faster PMax finds the right people.

Creative Assets

Google serves your assets across Search, Shopping, Display, YouTube and Discover. That means you need:

- Multiple headline variations (test benefit-led vs feature-led vs urgency)

- At least 5-10 image assets per group including lifestyle, product and graphic styles

- Video assets (even simple ones outperform accounts with no video)

- Descriptions that reinforce your USP and include relevant keywords

Rotate creative monthly. Stale assets = declining performance.

What to Automate

Bidding

Let Google handle the bidding. Target ROAS or Maximise Conversion Value are the right strategies for e-commerce PMax campaigns. Set realistic targets based on historical data and give the algorithm 2-3 weeks to stabilise before adjusting.

Resist the urge to change ROAS targets every few days. Small, infrequent adjustments (5-10% at a time) work better than dramatic swings.

Placement Distribution

You can now see where PMax is allocating spend across channels (Search, Shopping, Display, YouTube, Discover). Monitor this for insights but don't try to force specific channel splits.

If Shopping is dominating and performing well, that's fine. If Display is eating budget with poor returns, the fix is usually better creative or tighter audience signals rather than trying to manually suppress a channel.

Dynamic Product Feeds

Let Google pull from your product feed dynamically. Make sure your feed is clean (titles, descriptions, images, GTINs, accurate pricing) and let PMax optimise which products to show to which users.

Focus your energy on feed quality rather than manual product selection.

What to Ignore

URL Expansion

PMax's URL expansion feature sends traffic to pages Google thinks are relevant, which often means blog posts, about pages or collection pages that don't convert. Turn this off for most e-commerce accounts and specify your landing pages manually.

Automatically Created Assets

Google can now auto-generate headlines, descriptions and even image assets. In our experience, the quality is inconsistent and dilutes your brand voice. Keep this off and supply your own creative.

Final URL Expansion for Brand Campaigns

If you're running PMax alongside a dedicated brand Search campaign, make sure PMax isn't cannibalising your brand traffic. Use brand exclusions in PMax and monitor the overlap. Brand clicks through PMax count toward its reporting but they would have converted anyway.

The Structure We Use

For most e-commerce accounts we run:

1. PMax - Core (high margin products): Tight ROAS target, strong creative, custom audience signals, aggressive negative keyword management

2. PMax - Long tail: Lower ROAS target, broader product feed, let Google discover what works

3. Standard Shopping (benchmark): Runs alongside PMax with a small budget to provide comparison data and catch any products PMax under-serves

4. Branded Search: Separate campaign, exact match, keeps brand traffic out of PMax

This structure gives you the benefits of PMax automation while maintaining enough control to steer performance.

Common Mistakes We See

- Running one giant asset group with the entire catalogue. Google can't optimise effectively with no segmentation signals.

- Setting ROAS targets too high too early, which starves the campaign of data and prevents it from scaling.

- Ignoring search term reports now that they're available. Free money is sitting in those reports.

- No video assets. Accounts with even basic video consistently outperform those without.

- Changing targets or structure too frequently. PMax needs learning time. Give it at least 2-3 weeks between major changes.

The Bottom Line

PMax in 2026 is a genuinely strong campaign type when managed properly. The key is controlling the inputs (structure, creative, signals, negatives) and letting Google optimise the outputs (bidding, placement, delivery). Treat it as a tool that needs strategic direction, not a black box you set and forget.