Performance Max Campaigns — When They Win, When They Burn

13 min · video · Intermediate

Performance Max sounds like magic: feed your products, set a budget, let the algorithm go. In reality it is a black box with expensive learning phases and a kill threshold where you lose money fast. Today we decode it — and show you the search-term insights Google hides.

Performance Max: the black box unpacked

Performance Max (PMax) is Google's algorithmic campaign type. You give it:

Product feed (from Merchant Center)
Budget (A$50–500/day typical)
Target ROAS (optional; "I want 1.5 ROAS" or leave empty for "maximize conversions")

Google then:

Selects products from your feed
Writes ad copy (if you do not provide it)
Selects audience
Chooses placements (Search, Shopping, YouTube, Gmail, Maps, Discover)
Sets bid automatically
Rotates creative daily

The trade-off: maximum automation, medium control. You cannot see which keywords triggered your ad. You cannot bid on specific products. You see only aggregated metrics (CPA, ROAS).

The learning phase: why day 1–7 ROAS is noise

PMax has a 7–14 day learning curve. In days 1–3, the algorithm is random (like new Advantage+ on Meta). ROAS will look terrible.

Real AU data (kitchen gadget, A$50/day PMax budget):

Day	Spend	Conversions	ROAS	Phase
1–3	A 50	2	0.4	Learning (ignore)
4–7	A