Asset groups are theme-based creative containers in Performance Max: text, images, video, and audience signals bundled per landing page, not per audience.
What asset groups are
An asset group is the Performance Max equivalent of an ad group, but theme-driven. Each group holds up to 20 images, 5 logos, 5 videos, 15 headlines, 5 long headlines, 5 descriptions, and one final URL (Google Ads Help, 2025-09), plus audience signals, listing groups (for retail), and ad strength feedback.
The unit of meaning is theme, not audience. A theme is usually a landing page, an offer, or a coherent product cluster. If you cannot write 5 honest descriptions that fit every URL in the group, the theme is wrong.
How asset groups work
Performance Max generates ad combinations from your assets and serves them across Search, YouTube, Display, Discover, Gmail, and Maps. Each group is evaluated against the campaign-level bidding strategy and budget: the campaign is the spending entity, the group is the creative entity.
Audience signals are the most misunderstood part. Google's own help is explicit:
"Audience signals are suggestions to help Google's AI find new and relevant audiences faster. Performance Max may serve ads outside the audience signals you provide if it predicts a higher-value conversion." (Google Ads Help, 2025-09)
A Customer Match list, in-market audience, or custom segment is a learning hint, not a fence. The bidding strategy (tCPA or tROAS) plus the conversion signal drives spend. Asset-group-level reporting rolled out in 2024-Q4, finally exposing impressions, clicks, and conversion value per group instead of only per campaign (Search Engine Land, 2024-11-12).
Asset group structure best practices
The default is one asset group per landing page or per tightly themed product cluster. Split when assets, headlines, or descriptions diverge: a Black Friday LP deserves its own group because offer and creative differ from evergreen pages. Consolidate when you cannot write distinct headlines per group.
Conversion volume sets a hard floor. Smarter eCommerce's 2025 FAQ recommends at least 30 conversions in the last 30 days per asset group before splitting further; below that threshold each group sits in extended learning and CPA drifts (Smarter eCommerce: PMax 2025 FAQ, 2025-03). Accounts with limited conversion volume should consolidate aggressively.
Practical thresholds that hold across audits:
| Signal | Action |
|---|---|
| <30 conversions / 30 days per group | Consolidate into broader theme |
| 30-100 conversions / 30 days | Hold structure, monitor asset strength |
| >100 conversions / 30 days, distinct LP | Split candidate |
| Same LP, different angle | One group, vary assets |
Always supply audience signals: Customer Match, recent visitors, custom segments from competitor URLs, plus one in-market list. Empty signals slow learning because Google has nothing to seed exploration from.
Across Whitead audits of PMax accounts in 2024-2026, the dominant failure mode is over-splitting on persona: one group per audience segment instead of per landing page. Symptom repeats: 6-12 groups, each below 30 conversions per month, all stuck in learning, campaign CPA 30-50% above account average. The inverse failure is one giant group covering 40 product categories, surfacing as low ad strength and no way to attribute spend to a winning theme. The fix starts the same way both directions: list landing pages, count conversions per LP over 30 days, build groups bottom-up from URLs that clear the 30-conversion floor. Signals come last, as exploration hints.
Common misconceptions
"Audience signals restrict who sees the ad."
They do not. Signals bias exploration during learning but do not exclude users outside them. For real exclusions use content suitability, brand exclusions, and negative keywords (rolled out account-wide in 2024-03). The Gemini-powered Asset Studio added in 2025-Q2 generates more creative variants but does not change targeting either.
"More asset groups means better performance."
The opposite, below conversion volume thresholds. Each new group competes for the same campaign budget and restarts its own learning. Smarter eCommerce and Optmyzr both publish the 30-conversions-per-group floor; below it, splitting actively hurts. Audits consistently show 3-5 well-fed groups outperform 10+ starved ones at the same spend.
"Asset groups are like Search ad groups."
Structurally similar (creative under a campaign), but the analogy breaks at targeting. Search ad groups are defined by keywords (explicit intent). Asset groups are defined by themes and serve across all PMax inventory; targeting is delegated to bidding strategy plus signals. Treating an asset group like a keyword-tight ad group leads to over-splitting.
Related concepts
- Performance Max — campaign type that contains asset groups.
- Smart Bidding — the 30-conversion floor on asset groups exists because Smart Bidding cannot stabilise below it.
Sources
- Google Ads Help: About asset groups in Performance Max (2025-09) — asset spec limits, audience signal definition.
- Smarter eCommerce: PMax 2025 FAQ (2025-03) — conversion thresholds, splitting guidance.
- Smarter eCommerce: State of Performance Max 2025 (2025) — asset-group-level reporting + consolidation evidence.
- Optmyzr: Is PMax Cannibalizing Your Search Campaigns? (2025-07) — multi-account data study on PMax structure.