What “custom audiences” means in Google Ads today (and what to build first)
In Google Ads, “custom audiences” usually refers to a few different (but complementary) audience tools. If you want to set them up effectively, start by deciding which of these you actually need, because each one solves a different problem.
Custom segments are your main “build a new audience from scratch” tool for prospecting. You describe the kind of person you want (using keywords, search terms, website URLs, or app names), and Google groups people likely to match that intent/behavior.
Your data segments (the newer name for what many advertisers still call remarketing lists) are for reaching people who already interacted with you—site visitors, app users, YouTube viewers, customer lists, and more.
Customer Match is the most important first-party audience for most advertisers right now. You upload first-party customer information (like email/phone/address in approved formats) and use it across Google surfaces, with the platform hashing/matching the data to signed-in users.
Combined segments let you create “AND/OR/NOT” logic between existing audiences, so you can get precision without making everything too narrow.
Finally, understand that modern Google Ads also has audience expansion mechanics (for example, optimized targeting in Display/Video and audience signals in Performance Max). These can use your audiences as “signals” and still go beyond them when it predicts better performance—so your setup must include smart exclusions and clean seed signals. Also note that “similar segments” are no longer generated in Google Ads (this change started May 1, 2023), so your first-party data and well-built custom segments matter more than ever.
Step-by-step: how to set up effective custom segments (prospecting)
1) Start with one clear audience job
Before you touch Audience Manager, decide what this audience is supposed to do. In practice, almost every effective audience falls into one of these jobs: prospecting (new users), warm prospecting (people actively researching), competitor conquesting (carefully), or remarketing (already engaged). When you mix jobs inside one audience, you blur intent and make optimization messy.
A reliable structure is to build separate custom segments for: “high intent” (closest to purchase), “mid intent” (problem-aware), and “category/affinity” (broader). This gives you levers to control reach, efficiency, and learning speed.
2) Create a custom segment in Audience Manager the right way
In your account, go to Audience Manager and create a new custom segment. Name it like you’ll name campaigns: include funnel stage, theme, and geo if relevant (for example, “HI | HVAC Repair | Search Terms | US”). Clear naming prevents wasted time six months later.
When you build the segment, you’ll typically choose between two keyword-driven interpretations:
Interests or purchase intentions (default) is best when you want scalable reach and you’re okay with Google interpreting your inputs as intent/interest signals across inventory.
People who searched for these terms on Google is best when you want the segment anchored to actual search behavior on Google properties. Just remember: that stricter “searched for” interpretation only applies on campaigns running on Google properties; elsewhere those terms are treated more like intent/interest inputs.
You can also feed the segment with:
URLs of websites your ideal customer visits (this helps model the audience; it does not mean your ads will appear on those specific sites).
App names your ideal customer uses (again, this influences who you reach; it does not guarantee placement inside those apps).
3) Build better seed inputs (the difference between “okay” and “excellent”)
The biggest mistake I see is using either (a) overly generic keywords that describe your product, or (b) a huge messy list of terms. A custom segment performs best when your inputs describe the user’s situation and intent, not your internal product taxonomy.
For a “high intent” custom segment, focus on action-oriented searches and solution-comparison behaviors (for example, “best”, “pricing”, “near me”, “provider”, “software”, “service”, “quote”, “estimate”, “implementation”). For “mid intent,” focus on the problem statement and alternatives (symptoms, use cases, “how to”, “vs”, “top tools”). For “category,” focus on adjacent behaviors and tools your buyer uses.
If you’re adding URLs, include a mix of: (1) direct competitors, (2) credible review/comparison sites in your niche, and (3) adjacent tools/platforms your buyer uses. If you only add competitors, you often bias the audience too narrowly and performance becomes volatile.
4) Use demographics as a refinement, not a crutch
You can narrow with demographics, but do it carefully. When advertisers over-filter (age, parental status, household income, etc.) they often choke delivery and force the system into higher-cost pockets of inventory. My rule: if you don’t have a proven conversion rate difference by demographic, keep it broad and let bidding/creative do the work—then refine once you have evidence.
Step-by-step: how to set up first-party audiences (remarketing + Customer Match)
1) Create “your data” segments (site/app/YouTube) with the right durations
Your data segments collect users via tagging (site/app) or via platform engagement (like YouTube viewers). Two settings matter most: membership duration and segment rules.
Set membership duration to match your real buying cycle. If you sell something quick-turn (like a local service), short durations can work well. If you sell high-consideration items (B2B, home projects, higher-ticket ecommerce), longer durations usually outperform because people research and return multiple times.
Also be aware that segments can be “Open” or “Closed,” and that audiences not used for a long period can be automatically closed (you can reopen them). This matters when you’re rebuilding campaigns and wonder why lists stopped growing.
Finally, don’t ignore eligibility thresholds. As a practical baseline, you need enough active users for the network you plan to use. If you’re trying to run RLSA-style tactics on Search, you’ll need substantially more volume than Display remarketing typically requires.
2) Build Customer Match like you’re building an asset, not a one-time upload
Customer Match is where I’d put most advertisers’ “custom audience” effort first—because it’s durable in a privacy-forward world and it’s the cleanest way to tell the system who your best customers are.
To set it up effectively, upload only first-party customer information (data customers shared directly with you). Include as many identifiers as you legitimately have per customer (email, phone, address, etc.) and keep them on the same row for that customer so matching works as well as possible. Don’t obsess over perfect match rates; use match rate mainly as a diagnostic signal for formatting/completeness issues, then improve it over time.
Make Customer Match an ongoing process. Lists decay fast if you don’t refresh them, especially for lead gen and subscription businesses where contact info and intent change. Frequent updates keep your segments usable and your exclusions accurate.
One more operational note: privacy and platform constraints matter. Certain traffic environments (notably iOS scenarios) can reduce how consistently customer-based audiences apply, and signed-in behavior is a real limiter. That’s normal—plan for it by pairing Customer Match with strong site-based audiences and solid conversion tracking.
3) Stay on the right side of personalized advertising rules
Effective custom audiences never require “creepy” targeting. Avoid building or using audiences based on sensitive information or on restricted site/app areas (for example, content implying medical conditions, negative financial status, or other sensitive categories). If you operate in a sensitive vertical, your safest path is often to lean more on policy-compliant, pre-defined audiences and contextual strategies, while keeping first-party data collection and usage strictly compliant.
Activating your audiences: campaign-by-campaign setup that actually performs
Search: use audiences for insights first, then for control
For Search campaigns, the most common “effective” setup is to add your first-party segments in Observation mode first. That lets you measure how different audience groups perform without restricting reach. Once you see consistent lift, you can apply bid adjustments (where available) or build dedicated audience-only Search campaigns when it makes sense.
When you do want strict control (for example, a “past converters” upsell campaign), that’s when you move from observation-style usage to audience-restricted targeting—while keeping keywords, ads, and landing pages aligned to the user’s stage.
Display & Video: expect expansion—so design the audience as a signal + protect with exclusions
On Display and Video, audience is powerful but it’s not always a hard boundary. With features like optimized targeting, the system can use your selected audiences (including custom segments and Customer Match) as signals and still serve beyond them if it predicts better results. This is exactly why your exclusions matter. If you don’t want existing customers, recent converters, employees, or low-quality leads included, exclude them explicitly—don’t assume targeting alone will prevent overlap.
Also separate prospecting and remarketing into different ad groups or campaigns whenever budgets are meaningful. It’s one of the fastest ways to prevent remarketing from “stealing” spend and making prospecting look weaker than it really is.
Performance Max: use audience signals to speed up learning, not to lock targeting
In Performance Max, you don’t “target” audiences in the traditional sense. You provide audience signals at the asset group level to guide the system. This is optional, but in real accounts it often reduces wasted early spend and helps the campaign learn what a good user looks like faster.
Your best starting audience signal stack is usually: (1) Customer Match of your best customers, (2) your highest-intent website visitor segments, and (3) a tightly themed custom segment built from search-term intent. From there, you iterate based on what the campaign actually finds—using creative, feed quality, and conversion quality as your primary levers.
Quick diagnostic checklist (when custom audiences don’t work)
- Audience too small to serve: confirm the list has enough recent active users for the network you’re running, and extend membership duration if it matches your buying cycle.
- Wrong audience mode: on Search, make sure you didn’t accidentally set audiences to restrictive targeting when you intended observation-based measurement (or vice versa).
- Overly narrow custom segment inputs: reduce hyper-specific terms, add a second “mid intent” segment, and avoid stacking too many filters at once.
- Policy/privacy limitations: remove sensitive-interest signals and avoid audience strategies tied to restricted content areas; rebuild using compliant themes and first-party segments.
- Performance Max expectation mismatch: remember signals guide learning but don’t hard-restrict delivery—use exclusions and conversion quality controls to shape outcomes.
How to iterate without wasting budget
The best custom audience strategies are built in cycles. Launch with one high-intent custom segment, one mid-intent custom segment, and one first-party segment. Let each collect statistically meaningful data, then refine by splitting winners (more granularity) and merging losers (less complexity). If you keep audiences clean, aligned to funnel stage, and backed by first-party data, you’ll usually see both better efficiency and more stable scaling.
Let AI handle
the Google Ads grunt work
| Section | Key concept | Practical how-to (from blog) | Common pitfalls | Related Google Ads documentation |
|---|---|---|---|---|
| Audience types & strategy | What “custom audiences” means today |
Break “custom audiences” into four main tools: • Custom segments for prospecting from scratch (keywords, URLs, apps). • Your data segments (remarketing) for past visitors/users/viewers and app/YouTube engagement. • Customer Match as your primary first‑party list across Google surfaces. • Combined segments to apply AND/OR/NOT logic across existing audiences. Plan what job each audience should do before building: net‑new prospecting, warm prospecting, competitor, or remarketing. |
• Treating all audience types as one catch‑all “custom audience.” • Ignoring the deprecation of similar segments and under‑investing in first‑party data. • Not planning exclusions for existing customers, employees, or low‑quality leads. |
• Audience Manager overview • About custom segments • How your data segments work • About Customer Match |
| Custom segments – planning | Start with one clear “audience job” |
Define a single, clear purpose for each segment before opening Audience Manager: • Build separate custom segments for high intent, mid intent, and category/affinity. • Align each segment to a funnel stage so you can control reach, bids, and budgets more precisely. |
• Mixing multiple intent levels (cold + warm + remarketing) into one segment, which blurs performance signals. • Using the same segment everywhere instead of tailoring to funnel stage and campaign type. |
• About custom segments • Add audience targeting to a campaign or ad group |
| Custom segments – build | Creating segments correctly in Audience Manager |
• In Audience Manager, create a new custom segment and name it with funnel stage, theme, and geo (for example, “HI | HVAC Repair | Search Terms | US”). • Choose the right keyword interpretation: – Interests or purchase intentions for scalable reach across inventory. – People who searched for these terms on Google when you want search‑anchored behavior on Google properties. • Add URLs of sites your ideal customer visits and apps they likely use to improve modeling (these are signals, not placement guarantees). |
• Assuming “people who searched” applies everywhere (it’s search‑anchored only on Google properties). • Expecting ads to show on the exact URLs/apps you enter rather than understanding them as modeling signals. • Vague or inconsistent naming, making later optimization and analysis difficult. |
• About custom segments • Audience Manager overview |
| Custom segments – inputs | Designing high‑quality seed signals |
• For high intent, prioritize action‑oriented and comparison terms (for example, “pricing”, “near me”, “quote”, “software”, “service”, “best”). • For mid intent, use problem/solution language, “how to”, “vs”, “top tools”, and alternatives your buyer is researching. • For broader category/affinity, focus on adjacent tools and behaviors buyers typically exhibit. • When adding URLs, mix: 1) Direct competitors 2) Credible review/comparison sites 3) Adjacent tools/platforms your buyer uses. |
• Using only generic product terms or internal taxonomy labels like feature names. • Dumping huge, uncurated keyword lists that dilute intent signals. • Feeding only competitor URLs, which can make audiences too narrow and volatile. |
• About custom segments • Audience insights |
| Custom segments – refinement | Using demographics as a secondary filter |
• Start broad on demographics unless you have proven conversion differences by age, gender, parental status, or income. • Let bidding strategies and creative do most of the selection work, then tighten demographics once you have clear data. |
• Over‑filtering demographics and choking delivery, which can push you into expensive, low‑volume pockets. • Assuming demographic differences without data support. |
• Add audience targeting to a campaign or ad group • Audience insights |
| Your data segments | Building site/app/YouTube remarketing correctly |
• Use your data segments to capture site visitors, app users, and YouTube engagement. • Choose membership duration to match your buying cycle: shorter for quick‑turn local services, longer for B2B/high‑consideration purchases. • Configure segment rules (for example, by URL or event) and understand “Open” vs “Closed” segment status; reopen closed lists when needed. • Ensure you meet minimum size thresholds for the network (Search requires more volume than Display). |
• Using default or arbitrary durations that don’t match how long users actually consider your product. • Forgetting that unused lists can auto‑close, then wondering why they stopped growing. • Trying RLSA‑style tactics on tiny lists that can’t serve. |
• How your data segments work • Set up your data segments for Search ads • Audience Manager overview |
| Customer Match – setup | Treat Customer Match as a core asset |
• Upload only first‑party customer data (information people shared directly with you). • Include as many identifiers per row as you legitimately have (email, phone, address, etc.) to improve matching. • Use match rate primarily as a diagnostic (formatting/completeness), then improve it over time rather than chasing a “perfect” number. • Refresh lists frequently so recency and exclusions stay accurate. |
• One‑time uploads that quickly decay and no longer represent active customers or leads. • Mixing in third‑party or non‑consented data, creating compliance and performance risk. • Under‑using Customer Match as a seed for prospecting and as a key exclusion list. |
• About Customer Match • Format your customer data file • How your data segments work |
| Customer Match – privacy | Operate within personalized advertising rules |
• Avoid audiences that rely on sensitive categories (health conditions, negative financial status, etc.) or restricted content areas. • For sensitive verticals, lean more on policy‑compliant predefined audiences, contextual targeting, and carefully handled first‑party data. • Pair Customer Match with strong site tagging and conversion tracking so performance remains resilient when match coverage is limited (for example, certain iOS scenarios). |
• Building “creepy” or sensitive‑interest audiences that violate personalized advertising policies. • Assuming customer‑based audiences will apply uniformly across all devices and environments. |
• Personalized advertising policy • How personalized ads work |
| Search activation | Use Observation first, then Targeting |
• Add your key first‑party segments to Search campaigns in Observation mode first to see how they perform without restricting reach. • Once you see consistent lift, apply bid adjustments or spin out dedicated audience‑only Search campaigns where appropriate. • Use strict targeting only when you intentionally want audience‑restricted campaigns (for example, upselling past converters). |
• Accidentally setting lists to Targeting when you meant Observation, which silently constrains reach. • Building audience‑only Search without ensuring keywords, ads, and landing pages align to that audience’s stage. |
• About “Targeting” and “Observation” settings • Set up your data segments for Search ads |
| Display & Video activation | Audiences as signals + exclusions |
• Expect optimized targeting and audience expansion to go beyond your selected segments when it predicts better performance. • Treat custom segments and Customer Match as strong signals, not hard boundaries. • Protect performance with explicit exclusions for converters, existing customers, employees, and low‑quality leads. • Separate prospecting vs remarketing into different ad groups or campaigns to keep budgets and optimization clean. |
• Assuming that adding an audience fully restricts where ads serve, then being surprised by broader reach. • Letting remarketing soak up most spend in mixed ad groups, making prospecting look weak. • Forgetting to exclude converters or Customer Match lists when running pure acquisition campaigns. |
• About optimized targeting • About audience expansion • Add audience targeting to a campaign or ad group |
| Performance Max activation | Audience signals, not fixed targeting |
• In Performance Max, add audience signals at the asset group level instead of traditional targeting. • Start with: 1) Customer Match of your best customers. 2) Highest‑intent site visitor segments. 3) A tightly themed custom segment based on top search intent. • Use signals to speed up learning and reduce early waste, then iterate based on actual results. |
• Expecting Performance Max to stay within your audience signals like a normal targeting setting. • Over‑focusing on audiences instead of improving creative, feed quality, and conversion tracking. |
• Add audience signals • Audience insights • About optimized targeting |
| Diagnostics | Quick checks when custom audiences underperform |
Use the blog’s checklist: • Confirm list size and recency; extend membership duration if it matches your buying cycle. • Verify audience mode on Search (Observation vs Targeting). • Loosen overly narrow custom segment inputs and avoid stacking too many filters. • Remove sensitive‑interest themes if you’re hitting policy limitations. • For Performance Max, remember that signals don’t hard‑restrict delivery; focus on exclusions and conversion quality. |
• Over‑reacting to short‑term performance without checking basic eligibility and settings first. • Iterating by adding more constraints instead of simplifying and re‑aligning to funnel stages. |
• About “Targeting” and “Observation” settings • About optimized targeting • Audience Manager overview |
| Iteration strategy | Build in cycles, not one‑offs |
• Launch with a simple structure: one high‑intent custom segment, one mid‑intent custom segment, and one strong first‑party segment. • Let them gather statistically meaningful data. • Split winners for more granularity and control; merge or drop losers to reduce complexity. • Keep audience definitions clean, mapped to funnel stages, and supported by fresh first‑party data. |
• Constantly rebuilding audiences before they collect enough data to learn. • Accumulating many overlapping segments that add operational overhead without clear strategic roles. |
• Audience Manager overview • Audience insights |
If you’re building custom audiences in Google Ads—splitting intent levels into clean custom segments, keeping “your data” and Customer Match lists fresh, and being careful with exclusions and settings like Observation vs Targeting—Blobr can help you stay consistent as things scale. It connects to your Google Ads account, continuously reviews audiences and performance signals alongside account structure, and uses specialized AI agents to surface practical recommendations (like where segments are too broad or too restrictive, when lists may be stale or under-sized, or when prospecting and remarketing are getting mixed), so you can iterate methodically without having to audit everything by hand every week.
What “custom audiences” means in Google Ads today (and what to build first)
In Google Ads, “custom audiences” usually refers to a few different (but complementary) audience tools. If you want to set them up effectively, start by deciding which of these you actually need, because each one solves a different problem.
Custom segments are your main “build a new audience from scratch” tool for prospecting. You describe the kind of person you want (using keywords, search terms, website URLs, or app names), and Google groups people likely to match that intent/behavior.
Your data segments (the newer name for what many advertisers still call remarketing lists) are for reaching people who already interacted with you—site visitors, app users, YouTube viewers, customer lists, and more.
Customer Match is the most important first-party audience for most advertisers right now. You upload first-party customer information (like email/phone/address in approved formats) and use it across Google surfaces, with the platform hashing/matching the data to signed-in users.
Combined segments let you create “AND/OR/NOT” logic between existing audiences, so you can get precision without making everything too narrow.
Finally, understand that modern Google Ads also has audience expansion mechanics (for example, optimized targeting in Display/Video and audience signals in Performance Max). These can use your audiences as “signals” and still go beyond them when it predicts better performance—so your setup must include smart exclusions and clean seed signals. Also note that “similar segments” are no longer generated in Google Ads (this change started May 1, 2023), so your first-party data and well-built custom segments matter more than ever.
Step-by-step: how to set up effective custom segments (prospecting)
1) Start with one clear audience job
Before you touch Audience Manager, decide what this audience is supposed to do. In practice, almost every effective audience falls into one of these jobs: prospecting (new users), warm prospecting (people actively researching), competitor conquesting (carefully), or remarketing (already engaged). When you mix jobs inside one audience, you blur intent and make optimization messy.
A reliable structure is to build separate custom segments for: “high intent” (closest to purchase), “mid intent” (problem-aware), and “category/affinity” (broader). This gives you levers to control reach, efficiency, and learning speed.
2) Create a custom segment in Audience Manager the right way
In your account, go to Audience Manager and create a new custom segment. Name it like you’ll name campaigns: include funnel stage, theme, and geo if relevant (for example, “HI | HVAC Repair | Search Terms | US”). Clear naming prevents wasted time six months later.
When you build the segment, you’ll typically choose between two keyword-driven interpretations:
Interests or purchase intentions (default) is best when you want scalable reach and you’re okay with Google interpreting your inputs as intent/interest signals across inventory.
People who searched for these terms on Google is best when you want the segment anchored to actual search behavior on Google properties. Just remember: that stricter “searched for” interpretation only applies on campaigns running on Google properties; elsewhere those terms are treated more like intent/interest inputs.
You can also feed the segment with:
URLs of websites your ideal customer visits (this helps model the audience; it does not mean your ads will appear on those specific sites).
App names your ideal customer uses (again, this influences who you reach; it does not guarantee placement inside those apps).
3) Build better seed inputs (the difference between “okay” and “excellent”)
The biggest mistake I see is using either (a) overly generic keywords that describe your product, or (b) a huge messy list of terms. A custom segment performs best when your inputs describe the user’s situation and intent, not your internal product taxonomy.
For a “high intent” custom segment, focus on action-oriented searches and solution-comparison behaviors (for example, “best”, “pricing”, “near me”, “provider”, “software”, “service”, “quote”, “estimate”, “implementation”). For “mid intent,” focus on the problem statement and alternatives (symptoms, use cases, “how to”, “vs”, “top tools”). For “category,” focus on adjacent behaviors and tools your buyer uses.
If you’re adding URLs, include a mix of: (1) direct competitors, (2) credible review/comparison sites in your niche, and (3) adjacent tools/platforms your buyer uses. If you only add competitors, you often bias the audience too narrowly and performance becomes volatile.
4) Use demographics as a refinement, not a crutch
You can narrow with demographics, but do it carefully. When advertisers over-filter (age, parental status, household income, etc.) they often choke delivery and force the system into higher-cost pockets of inventory. My rule: if you don’t have a proven conversion rate difference by demographic, keep it broad and let bidding/creative do the work—then refine once you have evidence.
Step-by-step: how to set up first-party audiences (remarketing + Customer Match)
1) Create “your data” segments (site/app/YouTube) with the right durations
Your data segments collect users via tagging (site/app) or via platform engagement (like YouTube viewers). Two settings matter most: membership duration and segment rules.
Set membership duration to match your real buying cycle. If you sell something quick-turn (like a local service), short durations can work well. If you sell high-consideration items (B2B, home projects, higher-ticket ecommerce), longer durations usually outperform because people research and return multiple times.
Also be aware that segments can be “Open” or “Closed,” and that audiences not used for a long period can be automatically closed (you can reopen them). This matters when you’re rebuilding campaigns and wonder why lists stopped growing.
Finally, don’t ignore eligibility thresholds. As a practical baseline, you need enough active users for the network you plan to use. If you’re trying to run RLSA-style tactics on Search, you’ll need substantially more volume than Display remarketing typically requires.
2) Build Customer Match like you’re building an asset, not a one-time upload
Customer Match is where I’d put most advertisers’ “custom audience” effort first—because it’s durable in a privacy-forward world and it’s the cleanest way to tell the system who your best customers are.
To set it up effectively, upload only first-party customer information (data customers shared directly with you). Include as many identifiers as you legitimately have per customer (email, phone, address, etc.) and keep them on the same row for that customer so matching works as well as possible. Don’t obsess over perfect match rates; use match rate mainly as a diagnostic signal for formatting/completeness issues, then improve it over time.
Make Customer Match an ongoing process. Lists decay fast if you don’t refresh them, especially for lead gen and subscription businesses where contact info and intent change. Frequent updates keep your segments usable and your exclusions accurate.
One more operational note: privacy and platform constraints matter. Certain traffic environments (notably iOS scenarios) can reduce how consistently customer-based audiences apply, and signed-in behavior is a real limiter. That’s normal—plan for it by pairing Customer Match with strong site-based audiences and solid conversion tracking.
3) Stay on the right side of personalized advertising rules
Effective custom audiences never require “creepy” targeting. Avoid building or using audiences based on sensitive information or on restricted site/app areas (for example, content implying medical conditions, negative financial status, or other sensitive categories). If you operate in a sensitive vertical, your safest path is often to lean more on policy-compliant, pre-defined audiences and contextual strategies, while keeping first-party data collection and usage strictly compliant.
Activating your audiences: campaign-by-campaign setup that actually performs
Search: use audiences for insights first, then for control
For Search campaigns, the most common “effective” setup is to add your first-party segments in Observation mode first. That lets you measure how different audience groups perform without restricting reach. Once you see consistent lift, you can apply bid adjustments (where available) or build dedicated audience-only Search campaigns when it makes sense.
When you do want strict control (for example, a “past converters” upsell campaign), that’s when you move from observation-style usage to audience-restricted targeting—while keeping keywords, ads, and landing pages aligned to the user’s stage.
Display & Video: expect expansion—so design the audience as a signal + protect with exclusions
On Display and Video, audience is powerful but it’s not always a hard boundary. With features like optimized targeting, the system can use your selected audiences (including custom segments and Customer Match) as signals and still serve beyond them if it predicts better results. This is exactly why your exclusions matter. If you don’t want existing customers, recent converters, employees, or low-quality leads included, exclude them explicitly—don’t assume targeting alone will prevent overlap.
Also separate prospecting and remarketing into different ad groups or campaigns whenever budgets are meaningful. It’s one of the fastest ways to prevent remarketing from “stealing” spend and making prospecting look weaker than it really is.
Performance Max: use audience signals to speed up learning, not to lock targeting
In Performance Max, you don’t “target” audiences in the traditional sense. You provide audience signals at the asset group level to guide the system. This is optional, but in real accounts it often reduces wasted early spend and helps the campaign learn what a good user looks like faster.
Your best starting audience signal stack is usually: (1) Customer Match of your best customers, (2) your highest-intent website visitor segments, and (3) a tightly themed custom segment built from search-term intent. From there, you iterate based on what the campaign actually finds—using creative, feed quality, and conversion quality as your primary levers.
Quick diagnostic checklist (when custom audiences don’t work)
- Audience too small to serve: confirm the list has enough recent active users for the network you’re running, and extend membership duration if it matches your buying cycle.
- Wrong audience mode: on Search, make sure you didn’t accidentally set audiences to restrictive targeting when you intended observation-based measurement (or vice versa).
- Overly narrow custom segment inputs: reduce hyper-specific terms, add a second “mid intent” segment, and avoid stacking too many filters at once.
- Policy/privacy limitations: remove sensitive-interest signals and avoid audience strategies tied to restricted content areas; rebuild using compliant themes and first-party segments.
- Performance Max expectation mismatch: remember signals guide learning but don’t hard-restrict delivery—use exclusions and conversion quality controls to shape outcomes.
How to iterate without wasting budget
The best custom audience strategies are built in cycles. Launch with one high-intent custom segment, one mid-intent custom segment, and one first-party segment. Let each collect statistically meaningful data, then refine by splitting winners (more granularity) and merging losers (less complexity). If you keep audiences clean, aligned to funnel stage, and backed by first-party data, you’ll usually see both better efficiency and more stable scaling.
