Performance Max asset groups: stop loading every campaign with the same 10 assets

Most accounts we inherit have one Performance Max campaign, one asset group, ten headlines, four descriptions, three images, and a logo. Then the owner asks why the campaign isn't performing and the answer is: because there's nothing for it to test.

PMax is a machine that needs raw material. If you give it the same thin pile of assets every time, you get the same thin results.

How asset groups actually work

An asset group inside a Performance Max campaign is a creative + audience signal bundle. The campaign decides which channel to serve on (Search, YouTube, Display, Discover, Maps, Gmail), and then picks an asset group based on what the user looks like.

If you only have one asset group, every channel and audience gets the same creative. That's like running a single radio ad on a station that plays four formats. Some of the time it lands. Most of the time it doesn't.

The structure we run

For a local service business with multiple service lines and service areas, we build asset groups along two axes: what the service is and who the customer is.

Service-line groups

If you're an HVAC company, you have at least three asset groups:

• AC repair (summer-coded creative, urgency, before/after photos of clean coils)
• Heating repair (winter-coded creative, comfort framing, family in warm house)
• New install / replacement (longer consideration, warranty messaging, financing)

Each gets its own headlines, descriptions, images, and YouTube clip. Each points to a landing page that matches.

Geographic groups

If you cover six cities and one of them is your home base, that one gets its own asset group. Headlines mention the city. The landing page is city-specific. The image is recognizable to a local.

The other five cities can share an asset group if you're early. Once a city gets meaningful spend (>$1k/month), it earns its own.

What goes in each group

Google's interface lets you upload more assets than most operators realize. Use the room.

• 15 headlines (max). Mix promise, proof, geography, and CTA.
• 5 long headlines. These show on Discover and Gmail.
• 5 descriptions. Short and direct.
• 20 images at 1.91:1 and 20 at 1:1. Mix of staff, work, customer-facing photos.
• 5 logos.
• 3+ YouTube videos under 30 seconds. Even a phone-shot vertical works.
• A site links list with 4–6 specific links per service line.
• Audience signal: a custom segment built from your converters list, supplemented with a small set of in-market audiences relevant to the service.

You don't need every asset to be agency-polished. You need variety. Google rotates them. The boring ones lose, the good ones win, and the algorithm tells you which is which.

The two mistakes we see most

The first is reusing the same image across every asset group. PMax notices that the image isn't differentiating — it gets de-prioritized. Same with headlines.

The second is leaving the audience signal blank. Google says signals are optional, and they technically are, but a campaign without a signal takes 4–6 weeks longer to converge. Give it a starting point: your customer match list, a custom segment built from your competitor URLs, and one or two in-market audiences.

Reading the asset report

Inside each asset group, Google grades each individual asset Best, Good, Low, or Pending. The instinct is to delete every Low.

Don't. Delete only the Lows that have meaningful impressions (1,000+) and have been live more than 30 days. Anything newer hasn't had time. Anything with low impressions never got a fair test — Google might serve it more in a different context.

Replace, don't just delete. The campaign loses learnings if you cut assets without adding equivalents.

What you'll see

We rebuilt PMax for a regional landscaping company last quarter. They had one asset group covering all four service lines and twelve cities. We split it into seven asset groups (four service lines, plus three geo-specific groups for their top cities).

Three weeks in:

• CPA dropped 22%
• Lead volume up 38% on the same spend
• Mix of services in the lead form looked more like the actual business mix, instead of 80% one service

The work was a Saturday afternoon of writing copy and uploading images, followed by three weeks of watching it settle. That's the trade.

The point

Performance Max is going to keep eating budget. The accounts that win are the ones that feed it differently. Build asset groups for each service line. Build asset groups for your highest-value cities. Give every group enough creative variety that the algorithm has a real test to run. Then let it run.