Performance Max
The campaign type audience signals attach to. The six-surface, machine-learning campaign.
Read the entry →Stan Consulting · Marketing Atlas · Reference · Google Ads
Audience Signals.
The first-party customer lists, custom segments, and interest signals an operator gives Performance Max as a “starting point” for the algorithm. The signal Google asks for and routinely ignores.
Section 02 · Quick definition
Definition.
An audience signal is a description of who is most likely to convert against a given Performance Max asset group, supplied by the operator at asset-group level. The signal is composed of customer-match lists, website-visitor lists, custom segments built from search and URL data, and Google's interest and demographic taxonomies. It is informational input. The campaign is allowed to serve outside the signal whenever the algorithm predicts a conversion, and routinely does. The signal accelerates learning rather than restricting reach.
Section 03 · Why it matters
Why it matters.
Audience signals matter because operators routinely interpret them as targeting and run their accounts as if they were. The wording in the Google Ads interface contributes to the misread: “audience signal” sounds like a direction the algorithm follows. It is a hint the algorithm consults. The campaign continues to serve to whomever the conversion model predicts will convert, signal or no signal. Confusing signal for targeting causes operators to attribute Performance Max performance to audience inputs that did almost no work.
The signal does have real effects, but at a different layer than operators typically expect. A high-quality first-party customer list seeds the conversion model with examples of who has bought before. A custom segment built from competitor URLs and search terms describes a research-stage audience the algorithm will look for during exploration. Both shorten the campaign's learning period. Neither restrict who the campaign serves to.
The practical stake is that signal quality compounds over the campaign lifetime. An asset group seeded with stale lists and generic interests learns slowly and never recovers. An asset group seeded with current customer data and well-built custom segments learns fast and stays well-calibrated.
Section 04 · How it works
How audience signals shape delivery.
Audience signals attach at the asset-group level inside a Performance Max campaign. Each asset group can hold one combined signal, which Google calls an “audience.” That audience can pool four types of input: customer-match data lists uploaded by the operator, website-visitor segments collected via the Google tag, custom segments built from search terms and URLs, and Google's built-in interest and demographic categories.
-
Customer-match lists
Hashed first-party data uploaded from the operator's CRM or ESP. The most concrete signal type. Quality scales with list size, recency, and how cleanly customer events were captured upstream.
-
Website-visitor segments
Audiences built from on-site behaviour: cart abandoners, repeat visitors, page-segment audiences, purchase audiences. Requires the Google tag to be implemented and remarketing collection to be enabled.
-
Custom segments
Audience definitions built from keywords, URLs, app downloads, and YouTube engagement. The most underused category. A well-built custom segment of competitor URLs and category search terms describes a high-intent prospecting pool the algorithm can lean on during exploration.
-
Built-in audiences
Affinity, in-market, life-event, and demographic categories from Google's taxonomy. Lowest signal quality of the four types because the categories are coarse and operator-agnostic. Used carefully, they help cold-start new campaigns; used as the only input, they teach the algorithm almost nothing.
The campaign treats the combined signal as a description of the audience most likely to convert. It does not treat the signal as a fence. Performance Max is allowed to serve to anyone the conversion model predicts will convert, signal-matched or not.
Section 05 · Common misunderstandings
What people get wrong.
-
“Audience signals tell Performance Max who to target.”
Performance Max serves to whomever the conversion model predicts will convert. The signal accelerates learning by giving the model concrete starting examples. It does not restrict the auctions the campaign enters.
-
“Adding more interests to the signal makes it stronger.”
Adding more low-quality categories dilutes the signal. The model learns from the most concrete inputs first — customer lists and custom segments — and underweights generic interests. A focused signal trains faster than a broad one.
-
“The same signal works across asset groups.”
Signals are tied to asset groups for a reason. A clearance asset group should be seeded with discount-responsive customer segments. A premium-line asset group should be seeded with high-LTV first-party data. Reusing one signal across asset groups defeats the structure.
-
“Customer-match lists are too small to matter.”
Google's minimums for delivery are different from minimums for signal. A small first-party list still teaches the conversion model what the operator's actual buyers look like, which generic categories cannot do at any size.
-
“Once the campaign is learning well, the signal does not matter.”
The signal continues to shape who the algorithm explores against. Over time it interacts with the conversion data Google receives back, and the two together determine the steady-state audience the campaign delivers to.
Section 06 · Diagnostic questions
Questions a Stan Consulting diagnostic asks.
What audience signal is attached to each asset group, and what proportion of it is first-party data versus generic interest categories?
When were the customer-match lists last refreshed, and what conversion or LTV criteria define them?
Is the Google tag implemented site-wide, and are remarketing audiences capturing the segments the asset groups need?
Does each asset group have a distinct audience signal aligned with its theme, or is one signal reused across asset groups?
Are custom segments built from competitor URLs and category search terms, and are they refreshed quarterly?
What share of the asset group's actual conversions came from the audience signal versus from outside it?
If the operator could only give the campaign one input, would they trust their own audience signal enough to use it?
Section 07 · Related Atlas entries
Adjacent reading.
Section 08 · Five Cents
The most expensive misread I see in Performance Max accounts is the operator who points at the audience signal field and says “that is who we are targeting.” The operator is not targeting anyone. Performance Max is allowed to serve to whomever its conversion model decides will convert, and the signal is a hint the model takes into account during learning. When the operator believes they are targeting and the campaign is actually serving outside the signal, the result is a cluster of decisions made on the wrong premise: budget defended on a fake targeting basis, creative built for the audience signal rather than for the real serve mix, and reporting framed around an audience that is not the audience. Treat audience signals as what they are. A starting hint. Build them well, refresh them regularly, and let them speed up the learning. Do not pretend they are a fence.
Stan · Marketing AtlasSection 09 · Sources
Sources.
- Google Ads Help · About audience signals in Performance Max Official documentation describing the four input types and how Google treats the signal during learning and steady state.
- Google Ads Help · About Customer Match Reference for first-party customer list uploads, hashing requirements, and matching rates.
- Google Ads Help · About custom segments Documentation for custom segments built from search terms, URLs, app downloads, and YouTube engagement.
- Search Engine Land · Performance Max coverage Industry reporting on audience-signal effectiveness, custom-segment construction, and signal-versus-targeting confusion.
- Search Engine Journal · Google Ads coverage Practitioner reporting on audience-signal strategy and customer-match list management.