Performance Max hides the levers, not the outcomes. Across the PMax accounts we manage, the difference between a 2x and a 6x return almost never comes from bid tweaks — it comes from how the campaign is segmented, how clean the product feed is, and what conversion value the algorithm is optimising towards.
01
Segment asset groups by margin, not by category
Most accounts mirror their website menu inside PMax. That forces Google to average performance across products with wildly different margins. We split asset groups by contribution margin and inventory depth instead, so high-margin lines get their own creative, their own audience signals and their own budget ceiling. Reporting becomes readable, and scaling decisions stop being guesswork.
02
Feed hygiene is 60% of the work
PMax reads your Merchant Center feed far more literally than it reads your copy. Titles front-loaded with the search phrase, complete GTIN and brand fields, custom labels for margin tiers and seasonality, and product-level exclusions for out-of-stock SKUs consistently lift ROAS before you touch a single setting. We audit the feed weekly and treat disapprovals as revenue incidents.
03
Send value, not just conversions
Optimising to conversions teaches the algorithm to buy cheap orders. Passing real revenue — better still, gross-profit values through offline conversion imports — teaches it to buy profitable ones. Enhanced conversions plus a server-side tag keeps that signal intact after browser tracking loss.
04
Give it room, then judge it on incrementality
PMax needs roughly 4-6 weeks and enough daily budget to clear the learning phase. Run brand exclusions so it cannot claim credit for demand you already own, and validate with a geo holdout or budget-split test before you declare victory.