Pmax is not "set and forget"
The single most common Google Ads mistake we see in audits: a single Performance Max campaign, one asset group, no audience signals, running for six months against search, display, YouTube, Gmail, and Discovery in a single black box.
The CPCs look fine. The ROAS looks passable. Everything is "trending up and to the right." And the CAC is 40% higher than it could be because Pmax is cannibalizing brand search and buying low-quality placements the advertiser has never inspected.
What good Pmax looks like
Multiple asset groups. One per product line or audience segment. Not one monster group.
Audience signals. First-party data (CRM uploads, past purchasers, high-LTV customers), in-market segments, and detailed demographics — all fed in as signals. Signals are not targeting. They are hints to the algorithm.
Brand exclusions. Pmax will eat your brand search if you let it. Exclude branded queries at the account level.
Asset variety. 5+ headlines, 5+ descriptions, 5+ images, 1+ video. Missing any asset type leaves Pmax blind to a placement.
Conversion hygiene. Enhanced conversions enabled. Server-side tagging live. The primary conversion action is revenue-weighted, not a soft signal.
When to run Pmax — and when not to
Run Pmax for:
- Ecommerce with a strong product feed.
- Lead-gen with a clear high-intent keyword universe (plus traditional search alongside).
Do not run Pmax-only for:
- Brand-driven queries (keep those in a standalone search campaign).
- Complex B2B with long sales cycles (Pmax's conversion goal is too blunt).
- Accounts with broken conversion tracking (Pmax requires clean data more than any other campaign type).
The TACoS question
We measure Google Ads on total cost of sale (TACoS), not ad-attributed return. Pmax that "generates $3 ROAS" while cannibalizing 40% of your organic brand traffic is losing money.
A healthy Pmax setup moves blended CAC down without shrinking organic brand search. That is the metric that matters.