Performance Max Product Group Strategy

What it costs

PMax works best above roughly 30-50 conversions per month per campaign. Splitting a small account into six campaigns starves each of data - consolidate until volume justifies splitting.

Treat every range above as a starting benchmark, not a quote. Competition in your specific market, the current state of your website, how much production capacity you already have in-house, and how aggressive your growth target is will all move the number. The useful exercise is to work backwards from revenue: what is one additional customer worth, how many do you need per month to justify the spend, and does the plan on the table plausibly produce them?

How long it takes

Each structural change restarts a two-week learning period. Plan structure once, hold it for six weeks, then evaluate - not weekly tinkering.

Two things reliably compress timelines. The first is production volume: more genuinely useful pages and more consistent execution per month means faster results, which is why budget and speed are linked. The second is starting condition - a site with clean technical foundations and existing authority moves much faster than a new domain. Be honest about which one you are, then set expectations accordingly with whoever signs off on the budget.

What actually works

Segment campaigns by margin or ROAS target first, then by category. Products with 60% margin and 20% margin should not share a ROAS target.

Use listing groups to isolate bestsellers, new arrivals and clearance so budget follows business priority rather than algorithmic preference.

None of these are clever. They are the boring fundamentals executed at unusual quality and consistency, which is what separates programmes that compound from programmes that plateau after one quarter.

Add audience signals as accelerants, not targets: customer match lists, cart abandoners, high-value purchaser lookalikes and your best-performing search audiences. Signals guide early exploration; they do not restrict delivery.

Apply brand exclusions and account-level negatives so PMax stops harvesting brand traffic you already own, then compare like-for-like on non-brand performance.

Mistakes that cost the most

Judging PMax on blended ROAS while it cannibalises brand search.

Creating one asset group per SKU, leaving every group without enough data to learn.

Uploading weak creative assets and blaming the campaign type; asset quality is a primary PMax lever.

Every one of these is common enough that we expect to find at least two of them in any audit. Fixing them costs almost nothing compared with the spend they waste, so they belong at the top of the work plan rather than in a backlog.

Your next 90 days

Weeks 1-2: instrument everything first. Tracking, call tracking where relevant, a conversion definition that finance would accept, and a baseline you can defend later. Skipping this step is the single most expensive shortcut available, because without a baseline you cannot prove anything worked.

Weeks 3-8: ship the highest-leverage items from the 'what actually works' list above, in priority order, one at a time so each change stays attributable. Resist the urge to rebuild everything at once - it feels productive and it destroys your ability to learn.

Weeks 9-12: build the compounding layer - content cadence, review velocity, link acquisition, experimentation - and set a monthly review with a fixed agenda. Applying brand exclusions before evaluating Performance Max is the difference between measuring incremental performance and measuring your own brand traffic twice.

If you would like a second opinion before you commit budget, we run free 30-minute strategy calls and will walk through your current setup using the same diagnostic we use with paying clients.