Performance Max in 2026: what's actually working for ecommerce
PMax has matured. The brands winning are the ones treating it like a portfolio, not a single campaign.
After three years of Performance Max in the wild, the playbook has firmed up. The winners aren't the ones using more automation — they're the ones giving the algorithm sharper inputs and harder constraints. Treat PMax like a black box that needs careful framing, not a magic button, and the ROAS curve flattens out at a much higher number.
Asset groups as a portfolio
Split by margin tier or brand line, not by season. Feed each group hero creative and a tight audience signal. Let the algorithm find the rest. Most accounts we audit have a single asset group covering 80% of SKUs — that's the easiest 20% ROAS gain on the table.
The exclusion list is the strategy
Brand terms, low-margin SKUs and out-of-stock items belong in account-level exclusions. This single change typically lifts blended ROAS by 12–18%. Update the negative list weekly; PMax will quietly re-spend into excluded territory if you let the list go stale.
Creative requirements that matter
- At least 5 portrait videos per asset group (Demand Gen surfaces).
- 15 square + 15 landscape images, refreshed every 30 days.
- 5 logo variants — square, wide, and inverted for dark surfaces.
- Headlines mapped to product benefits, not generic CTAs.
Reporting that holds up
Insights tab is useful for direction, useless for accountability. Pull asset group spend and conversions via the Google Ads API into Looker Studio and report against blended MER, not in-platform ROAS. Anything inside Google's UI is optimistic by design.
