Google Ads PMax Creative Refresh: Asset Group Rotation Cadence That Works

The PMax mistake that cost a client 40 percent of their revenue

A home goods brand ran a single PMax campaign with one asset group for eighteen months. Assets hadn't been touched in fourteen months. Performance degraded by about 3 percent per month on a compounding basis, which nobody flagged because no single week looked catastrophic. By the time we audited, blended ROAS had dropped from 3.8 to 2.1. Thirty days of systematic creative refresh across four new asset groups brought ROAS to 3.4. The degradation was entirely creative fatigue.

TL;DR ▸ PMax asset groups decay slowly but continuously; silent decay is the dominant failure mode ▸ Partial refresh every 30 days, major refresh every 90 days, built into your calendar as a recurring task ▸ Maintain 3 to 6 asset groups per campaign, each with all asset slots filled ▸ Use the Asset Report performance labels to guide replacements, not feelings or recency

The PMax creative architecture that scales

One PMax campaign per major business segment. Inside each campaign, 3 to 6 asset groups, each built around a distinct theme, audience signal, or product line. Every asset group has every slot filled: 15 headlines, 5 descriptions, 20 images, 5 logos, 5 videos (or at least one video; Google will auto-generate the rest if needed, but your own video outperforms auto-generated video by 20 to 40 percent).

The theme of each asset group is what separates them. Not slight variations on the same creative. Genuinely different angles. For an apparel brand, asset groups might split as: everyday versatility, performance and technical, seasonal occasion, gifting, category specialist. Each group gets its own creative production and its own signal.

Audience signals attached to each asset group matter less than brands think. PMax will largely ignore them in favor of its own modeling, but the signal still helps the initial learning phase. Use your highest-value customer segment as the audience signal baseline.

The 30-day partial refresh protocol

Every 30 days, rotate 20 to 30 percent of the assets in each asset group. Replace the Low-rated ones first, then one or two Good-rated ones to introduce fresh signal. Leave Best-rated ones alone unless they are more than 90 days old.

The protocol in detail:

The 30-day cadence is the one that keeps ROAS from drifting downward. Skipping it does not break performance immediately; it degrades it slowly over months. By the time the degradation is visible in weekly reporting, you have lost a quarter of performance.

What a major refresh actually means

Partial refresh swaps assets. Major refresh creates a new asset group. The distinction matters because a new asset group restarts learning for that group (not for the campaign), which means a two-week underperformance window you have to plan for.

The trigger for a major refresh: when a campaign's performance has been flat or declining for 4 to 6 weeks despite partial refreshes. Or when you are expanding into a new product line or seasonal push that deserves its own theme. Or when you have hit a spend ceiling and adding a new asset group is the path to scale.

Execution: build the new asset group completely before launching. Fill every slot. Attach a specific audience signal that matches the theme. Launch at a modest budget allocation (15 to 20 percent of campaign spend) and let it learn for two weeks before judging. Do not mix new-group and old-group assets; that defeats the structural purpose.

Creative requirements by surface

PMax serves across YouTube, Gmail, Display, Discovery, Shopping, and Search. Each surface has different creative demands. Undersupplied creative across surfaces means PMax is forced to auto-generate, which reliably underperforms custom creative.

Video is the surface most brands under-invest in. A single 15-second vertical video outperforms zero videos by a large margin; five different videos outperform one significantly. Brands that treat PMax as a Shopping-only campaign are leaving the YouTube surface unserved, which caps their addressable inventory.

Shopping feed quality is often the biggest lever but lives outside the asset group. Product titles, descriptions, GTINs, high-quality images in the feed, and the Google Shopping optimization guide covers the feed side. PMax results without a clean feed are ceiling-capped no matter how good the asset groups are.

The Asset Report and how to read it

Each asset in each group gets a performance label: Best, Good, Low, or Pending. Pending means not enough data. Best and Good stay. Low gets swapped.

Three caveats. First, labels are relative within a group, not absolute. A group's Low can be another group's Good. Second, labels are slow to settle; a new asset may show Pending for 2 to 3 weeks. Third, labels do not tell you what to replace with. They tell you what to remove. You still need a creative brief for the replacement.

The replacement brief we use: "Low-rated headline being swapped for one that tests a different hook." Document the swap so you can learn across quarters which hooks work and which do not. Most teams swap without documentation and lose the institutional learning.

PMax versus Standard Shopping

Standard Shopping still has a role for certain catalogs, though narrower than it was in 2024. PMax wins for most catalog-wide D2C businesses. Standard Shopping wins for brands that want granular control over bid strategy per product and have the operational capacity to manage it.

The decision framework in the Google Shopping vs Performance Max comparison goes deep. The short version: if you are spending under $50K monthly on Google, stay in PMax unless you have a specific reason. If you are spending over $100K, consider splitting the catalog between Standard Shopping for the long tail and PMax for the head.

Our Performance Max for D2C essay covers the full campaign architecture for D2C brands running PMax as their primary Google channel.

Data-driven attribution and PMax

PMax performance signals come from Google's conversion tracking. If your conversions are misattributed or delayed, PMax will optimize on the wrong signal and spend inefficiently.

Three configuration items. First, enhanced conversions via the gtag or GTM server-side. This improves match rate by 15 to 25 percent, which matters for optimization quality. Second, data-driven attribution as the default model, not last-click. Third, accurate conversion values, including for the initiate-checkout and add-to-cart events if you want PMax to learn funnel proxies.

The ecommerce analytics GA4 server-side post covers the signal infrastructure. The attribution for D2C MER essay handles the blended measurement question that sits on top.

The scaling math for PMax

PMax scales more smoothly than ASC+ because it has more inventory and more optimization surfaces. The 20 to 30 percent budget increase rule still applies but the stability after each step is usually better than on Meta.

The ceiling on PMax scaling is usually creative volume plus feed quality, not budget tolerance. Brands that hit a ceiling almost always find that the diagnosis is "we have three asset groups, all with the same theme, and our product feed has 60 percent of products missing GTINs." Fix the input quality; scaling follows.

For brands with strong first-party data, customer lists uploaded as audience signals lift PMax performance measurably. Not as targeting (PMax ignores them as targets), but as learning seeds that help the algorithm find lookalike behavior faster. The paid ads services scope includes first-party data integration as a default for PMax accounts.

Brand versus non-brand separation

PMax's tendency to cannibalize brand search is well-documented. Two approaches keep this in check.

First, brand exclusion lists at the account level. Add your brand terms to the negative keyword list for PMax campaigns. This is not a perfect solution because PMax can still serve on Display and YouTube for brand-aware users, but it prevents the most egregious Search overlap.

Second, separate branded search campaigns running manual CPC with high bids and exact match. Starve PMax of the easy brand conversions and let it spend on incremental demand. The measurement work in the break-even ROAS guide helps you see whether PMax is earning incremental revenue or harvesting existing demand.

What to do this week

▸ Audit every active PMax campaign for asset group count, per-group slot fill rate, and asset age ▸ Pull the Asset Report for every group and list all Low-rated assets due for replacement ▸ Schedule a recurring 30-day calendar task named "PMax partial refresh" with named owner ▸ Plan one major refresh (new asset group with new theme) for the next quarter ▸ Confirm enhanced conversions and data-driven attribution are live on the account ▸ Add brand terms to the PMax negative keyword list if not already in place