Ecommerce Performance Max Fix: Asset Group Restructure and Auto-Generated Video Replacement

Question 1

Why is my Performance Max ROAS stable but revenue contribution declining?

Accepted Answer

According to groas, the brand held target returns but generated fewer total conversions at higher cost per acquisition, with CPA creeping up by roughly 15-20% over 12 months. The article states this pattern means the algorithm was optimizing for easy conversions rather than incremental ones, effectively paying more to reach people who would have converted anyway while losing reach among new, high-intent buyers. Adjusting ROAS targets, pausing underperforming products, and refreshing ad copy did not move performance in a lasting way.

Question 2

How does having one asset group per Performance Max campaign push budget to low-intent placements?

Accepted Answer

The article states the account ran three Performance Max campaigns, loosely organized by margin tier, each containing a single asset group with a broad mix of listings, headlines, descriptions, and images from across the catalog. With no framework for matching creative to audience stage, the algorithm defaulted to broad Display and YouTube placements where impressions are cheap, including a significant share of spend to mobile app placements, auto-play YouTube pre-rolls on unrelated content, and low-quality Display inventory. The article describes this as treating the entire catalog as one homogeneous offer, where signal quality matters more than the maximize conversion value with target ROAS bidding on top.

Question 3

Why does the article say never let Google auto-generate your Performance Max video?

Accepted Answer

According to groas, when no custom video is uploaded, Google auto-generates video as a slideshow of static product images with text overlays and stock transitions, with no product demonstration, narrative, or hook. The article states these videos consumed real budget on YouTube placements with negligible click-through rates, while view-through conversions still attributed value and masked the waste in default reporting. The audit found roughly a third of total PMax spend was going to placements that generated fewer than 2% of conversions, with auto-generated videos as the biggest offenders.

Question 4

How did this ecommerce brand restructure Performance Max asset groups by product category and intent?

Accepted Answer

The article states the single-asset-group-per-campaign structure was replaced with multiple asset groups organized by product category, with creative, headlines, descriptions, and audience signals matched to the products in each group. High-margin hero products received dedicated asset groups with the most specific audience signals, mid-tier products were grouped by category with shared creative themes, and lower-margin or long-tail products were grouped broadly with tighter spend controls. Each asset group received its own images, videos, headlines, and descriptions, with audience signals from Customer Match and in-market segments layered to give the algorithm a more productive starting point.

Question 5

What happened to revenue and CPA after the rebuild without increasing budget?

Accepted Answer

The article states that within the first four weeks of the restructured campaigns exiting learning, total revenue from Google Ads increased meaningfully on the same monthly spend. The improvement came entirely from reallocation, with budget that had leaked into low-intent Display and YouTube placements flowing toward Shopping and Search placements where purchase intent was higher. CPA dropped as the algorithm spent less time exploring low-quality placements, and the hero product asset groups consistently outperformed the broader groups.