Why PMax dominates eCommerce
For online stores, Performance Max has become the primary way Google sells your products. It spans Shopping, Search, YouTube, Display, Gmail, and Discover from one campaign, and Google has steadily pushed Shopping budget into it. For a store with a healthy feed and real conversion data, it can produce excellent efficiency, because it matches products to intent across surfaces faster than a human managing separate campaigns could.
The catch is that the same automation that makes it powerful also makes it opaque, and opacity plus default settings is where eCommerce accounts quietly lose margin. The goal of this playbook is to keep the power while removing the ways it wastes money.
The feed is the campaign
Performance Max for eCommerce does not use keywords. Google decides which searches your products match by reading the product feed, which means your product titles, descriptions, attributes, and identifiers are the targeting. This is the single most important thing to understand: in PMax, feed work is campaign work.
A vague title like Blue Shirt matches poorly; a precise Mens Slim Fit Oxford Shirt Navy Cotton matches far more relevant, higher-intent queries. Complete attributes, correct GTINs, and clean product types let Google understand and place items accurately. Custom labels let you segment products by margin, bestseller status, or season for smarter structure. Before touching budgets or targets, iClick audits and optimises the feed in Merchant Center, because it is the highest-leverage work on a Shopping account by a wide margin.
In Performance Max, the product feed is the targeting. Optimise the feed before you optimise the campaign.
Brand exclusions are non-negotiable
The most expensive default in eCommerce PMax is letting it absorb your brand search. Left unfenced, PMax will happily serve on people searching your brand name, take credit for those cheap, high-converting clicks, and report a flattering blended ROAS built partly on demand you already owned.
iClick applies brand exclusions from day one, at the account and campaign level, so PMax competes for new demand rather than harvesting existing demand and calling it performance. Branded search is then handled deliberately, usually in a dedicated low-cost brand campaign you can see and control. Skipping this one step is the difference between a PMax ROAS that reflects real incremental sales and one that flatters itself with conversions that would have happened anyway.
Structuring by margin and priority
A single PMax campaign dumping every product together hides your economics. Margins in eCommerce swing from 15 to 60 percent, and a blended target overpays for the high-margin winners while starving the low-margin lines.
The fix is to segment. Use custom labels to group products into margin tiers or priority bands, then run separate asset groups or campaigns with their own target ROAS derived from each tier's margin. High-margin products can tolerate a lower ROAS and buy more growth; thin-margin products need an aggressive target just to stay profitable. This margin-tier structure, applied through the feed and campaign layout, is how you make Performance Max respect your unit economics instead of averaging them into a number that is wrong for almost every product.
Asset groups and audience signals
The two creative levers you control are asset groups and audience signals. An asset group is a themed set of headlines, descriptions, images, and videos tied to a set of products, roughly like an ad group. Several tightly themed asset groups, aligned to product categories or margin tiers, beat one catch-all group stuffed with everything, because they let Google assemble more relevant ads for each product set.
Audience signals are not targeting; they are hints. Feeding PMax your customer lists, high-value segments, and relevant in-market audiences tells the model who your good customers resemble so it finds more of them faster. Strong first-party audience signals, especially value-weighted ones, meaningfully speed up how quickly a new PMax campaign finds profitable buyers.
Prying open the black box
Performance Max reports less than any other campaign type, which is exactly why it needs the most deliberate monitoring. You have to go and get the visibility it does not volunteer.
iClick uses scripts and the available reports to see where spend and conversions actually land: the search terms and search categories insight to catch off-intent or branded queries, placement reporting to spot wasted Display and YouTube spend, asset group and listing group performance to find the products carrying or dragging the campaign, and channel-level checks to confirm it is not simply leaning on Shopping while starving everything else. Without this work, a PMax campaign can report a healthy number while quietly wasting a third of its budget where no one is looking.
How to scale without breaking it
Scaling PMax is where discipline pays off. The temptation is to raise budgets fast and tighten targets hard at the same time, which throws the campaign back into learning and stalls it.
The reliable approach is to move one lever at a time and let each change settle over a full conversion cycle. Raise budget in steps the campaign can actually spend at target rather than doubling it overnight. Tighten target ROAS toward the margin-derived goal in 10 to 15 percent increments, not in one jump. Add products and asset groups deliberately, watching that new additions do not cannibalise proven ones. Feed improvements, better titles, more complete attributes, keep compounding as you scale. Growth comes from steady, measured pressure, not sudden moves.
One lever, one conversion cycle, then read the result. Budget and target changes made together just reset learning.
The eCommerce PMax mistakes that cost the most
1. Running without brand exclusions, so PMax harvests branded search and inflates its reported ROAS.
2. Neglecting the feed, which is the actual targeting. A weak feed caps performance no bid change can lift.
3. One blended target across products with wildly different margins, wrong for almost every product.
4. A single catch-all asset group instead of themed groups aligned to categories or margin tiers.
5. Trusting default reporting. PMax hides where conversions come from, so without scripts and placement review you cannot tell incremental sales from cannibalised demand. Verify, do not trust.
