Definition
Performance Max (PMax) is Google's automated campaign type that uses machine learning to run a single campaign across every Google ad surface — Search, Display, YouTube, Discover, Gmail, Maps — optimising toward a defined goal.
PMax launched in November 2021 and is now Google's strategic direction — Smart Shopping was deprecated and replaced with PMax in 2022. The pitch is simple: provide assets and a goal; Google's ML decides surfaces, audiences, and bids. Marketers trade control for reach and (in theory) better optimisation.
PMax is controversial. The reporting is opaque (asset-group-level data; limited search-term visibility), the algorithm cannibalises branded search if not carefully constrained, and audience signals get used in ways advertisers can't fully control. Despite the criticisms, when configured well it consistently outperforms manual campaigns at scale — particularly for ecommerce. The key is feed quality, audience signals, and conversion-value tracking.
Origin
Beta launched October 2020; general availability November 2021. Replaced Smart Shopping in 2022. Has expanded steadily since with new asset types, brand-exclusion controls (added 2023), and search-term visibility improvements.
How it works
- Set up conversion tracking with values (PMax is bid-based on value).
- Build asset groups around themes (often by product category or audience segment).
- Provide audience signals — customer lists, custom segments, similar audiences.
- Add brand exclusions to prevent cannibalising branded search.
- Add account-level negative keywords for irrelevant queries.
- Run for 4–6 weeks before drawing conclusions — PMax needs learning time.
- Track at asset-group level; compare to non-PMax baselines.
When to use it
Use when
- For ecommerce with strong product feeds and conversion values.
- When the account has scale and conversion volume to support ML learning.
- When manual campaigns have plateaued.
Skip when
- For tiny accounts. PMax needs 30+ conversions/month minimum to learn.
- For pure brand defense. PMax cannibalises branded queries — exclude them or run separate brand campaigns.
- When conversion tracking is unreliable. Bad signal in, bad performance out.
Key metrics
- Conversion value (PMax optimises on this).
- ROAS at the account level.
- Asset-group-level conversion volume.
- Branded vs non-branded query distribution (use brand exclusions if branded creeps in).
Examples
- PMax delivered 22% lower CPA than our prior Standard Shopping setup at 30% higher volume.
- We forgot brand exclusions and PMax ate our branded search. Spend doubled with no incremental conversion.
- PMax is opaque, but at scale the results often justify the loss of control.
In practice at Makreate
Makreate's paid acquisition work for ecommerce clients leans heavily on PMax — properly configured, it usually outperforms equivalent manual campaigns. The discipline is in the setup: clean conversion-value tracking, brand exclusions on day one, audience signals seeded from existing customers, account-level negatives. A recent DTC client moved from Standard Shopping to PMax and saw a 28% ROAS lift over 6 months — but the first month was -15% as the algorithm learned. Patience and the right setup were what made the difference; many advertisers kill PMax in the learning phase.
Advertising →Common mistakes
- Not adding brand exclusions. PMax will cannibalise your branded search.
- Killing the campaign in week 2. PMax needs 4–6 weeks of learning.
- Sending bad conversion data. The algorithm optimises on what you tell it.
- Treating PMax like Search. The reporting and controls are different and you'll miss what's working.
Frequently asked
Does PMax replace Search and Shopping?
Not entirely. PMax cannibalises Standard Shopping (Google deprecated Smart Shopping in PMax's favour) but coexists with Search. Most accounts run both.
How is PMax different from Smart Shopping?
Broader surface coverage (Search, YouTube, Discover, Gmail, Maps — not just Shopping), more asset types, more ML autonomy, less reporting transparency.
How do I see what PMax is doing?
Asset-group-level reports, performance insights tab, search-term insights (added in 2023). Still less transparent than manual campaigns — that's the trade-off.