Performance Max
The campaign type asset groups live inside. The six-surface, machine-learning campaign that hides where spend goes.
Read the entry →Stan Consulting · Marketing Atlas · Reference · Google Ads
The Performance Max equivalent of an ad group. The structural unit where assets, audience signals, and listing groups live. The grouping decision that determines what the algorithm optimizes against.
Section 02 · Quick definition
An asset group is the structural unit inside a Performance Max campaign that holds the creative inputs, the audience signal, and (for retail) the listing-group product selection. It plays the role an ad group plays in a Search campaign: a thematic container for assets that should be served against the same audience and goal. A campaign can hold multiple asset groups, and the operator's grouping decision determines what the algorithm has access to learn about and what gets reported as a unit.
Section 03 · Why it matters
Asset groups are where Performance Max becomes diagnosable or undiagnosable. A campaign with one asset group covering all products, all audiences, and all goals reports one ROAS number. The number is true. The number is also the only thing the operator can see. Reallocating budget, identifying weak segments, isolating creative tests — none of those are possible at the campaign level. They are possible at the asset-group level, but only if the asset groups are themed enough to mean something.
The reverse failure mode also exists: an account with twenty asset groups inside one campaign, each with a handful of assets, each below the data threshold the algorithm needs to learn. The ROAS swings wildly, the algorithm cycles spend through groups looking for signal, and the operator concludes Performance Max is volatile. The campaign is not volatile. It has been starved of structure.
The practical stake is that asset-group structure is a precondition for diagnosing Performance Max. A campaign with poor asset-group structure cannot be optimized; it can only be rebuilt.
Section 04 · How it works
An asset group bundles four input types: text assets (headlines, long headlines, descriptions), visual assets (images, logos, videos), an audience signal, and (for retail accounts) a listing group selecting which products from the Merchant Center feed are eligible. Performance Max assembles ads from those inputs and serves them across the six surfaces the campaign has access to.
Each asset group should map to a coherent product theme, customer theme, or seasonal theme. A retail account often runs one asset group per product category. A lead-gen account often runs one per service line or geographic market.
For retail Performance Max, the listing group inside the asset group selects which Shopping feed products are eligible. The same campaign can hold one asset group for high-margin SKUs and a separate one for clearance, each with its own creative voice.
The audience signal attached at asset-group level seeds the algorithm with a description of who is most likely to convert against this product or theme. The signal is informational, not restrictive; the algorithm is allowed to ignore it.
Performance reports surface conversion volume and conversion value at asset-group level. That is the layer at which the operator can compare segments, kill the weakest, scale the strongest, and reason about what is working.
The grouping decision is the single most important Performance Max input. Get it right, and the campaign becomes a structured set of segment tests. Get it wrong, and the campaign becomes a black box that returns one number per month.
Section 05 · Common misunderstandings
“One asset group per campaign is fine because Google figures it out.”
Google figures out delivery; it does not figure out grouping. With one asset group, you cannot isolate a category, run an audience test, or reallocate budget by segment. You have given up the diagnostic surface that asset groups provide.
“Asset groups are just like ad groups.”
Ad groups in Search bid on keyword sets. Asset groups in Performance Max contain creative and audience signals across six surfaces. The grouping logic is closer to a thematic product line than to a keyword cluster.
“More asset groups is always better.”
Asset groups need conversion volume to learn. Splitting a campaign into ten thinly funded groups starves each one. The right count is governed by conversion density, not by how many themes the operator can imagine.
“Asset groups inherit settings from the campaign so structure does not matter.”
Asset groups inherit bid strategy and conversion goals, but they hold their own audience signal, their own listing group, and their own creative pool. Those are precisely the inputs that determine what the algorithm sees as a unit.
“The asset-group ROAS shows incremental performance.”
Asset-group ROAS shows aggregate ROAS for the auctions that group won. It does not isolate incremental contribution, does not strip out branded conversions, and does not reflect what would have happened without the campaign.
Section 06 · Diagnostic questions
How many asset groups exist in each Performance Max campaign, and what coherent theme does each one represent?
What is the conversion volume per asset group over the last 30 days, and is each group above the algorithm's learning threshold?
For retail accounts, how is the Shopping feed segmented across listing groups, and does the segmentation match margin or merchandising priority?
What audience signal is attached at each asset-group level, and is it first-party, custom-segment, or interest-based?
Is creative asset coverage (headlines, descriptions, images, video) at the upper rating in Google's asset-group strength view?
When was the last time an under-performing asset group was paused, and what was learned from the pause?
If the campaign were rebuilt today with a clean grouping decision, what segments would it contain?
Section 07 · Related Atlas entries
Section 08 · Five Cents
The most common Performance Max structural failure I see is one asset group asked to wear four hats: clearance and full-price, prospecting and remarketing, evergreen and seasonal, all dropped into the same container with one audience signal and one creative pool. The campaign reports a 3x ROAS, the operator celebrates, the team moves on. Six months later the brand has stopped growing and nobody can name why. The reason is structural: the asset group hid every segment behind a single ROAS number. There was no way to see clearance was eating prospecting's budget, no way to see the audience signal was wrong for half the products, no way to test creative against the segment it was meant for. Asset-group structure is not Performance Max housekeeping. It is the only way the campaign type stays accountable.
Stan · Marketing AtlasSection 09 · Sources