A stack of 4 combined audience families (Affinity + In-Market + Custom + Similar) performs +28 to +42% conversions vs 1 family alone. The key: knowing when to stack them, and especially which family to switch to Targeting vs keep in Observation — choice that determines 70% of the final result.
Google Ads audiences form the invisible skeleton that determines who your ads speak to. In 2026, five families coexist — Affinity, In-Market, Custom, Detailed Demographics, Similar Segments — alongside Remarketing and Customer Match on the first-party side. Combining them intelligently is the difference between a mediocre ROAS and a ROAS that scales. Across the 2,000-account audit sample, a stack of 4 combined audience families generates +28 to +42% conversions (per vertical) vs a single activated family, at iso-CPA budget.
This strategic guide reviews each family: internal mechanics, typical volumes, recommended bid modifiers, pitfalls to avoid. It picks by account type (e-commerce, B2B SaaS, local services, retail) with proven combinations. To go further on the algorithmic pilot that consumes these audiences, follow up with our Smart Bidding guide.
What are the 5 families of Google Ads audiences in 2026?
Before diving into each family's detail, let's map the full landscape. Google Ads distinguishes five families of "interest" audiences — plus two first-party families (Remarketing and Customer Match) that we cover in dedicated articles. Each family answers a precise funnel moment, and none is universal.
- Affinity Audiences — long-term signals tied to identity and lifestyle (e.g. "Food Enthusiasts", "Sport Lovers", "Luxury Shoppers"). 100+ categories. Dominant upper funnel, awareness, broad prospecting.
- In-Market Audiences — active purchase intent over 24-72h (e.g. "Auto Vehicles/SUV", "B2B Services/CRM"). 400+ precise categories. Dominant mid-to-lower funnel, controlled CPA.
- Custom Audiences — custom signals provided by the advertiser: keywords, competitor URLs, apps used. 2 subtypes (Custom Intent, Custom Affinity). Powerful for B2B and niches.
- Detailed Demographics — socio-demographic segmentation: household income, parental status, education, homeownership. Useful when the product has a clear demographic target.
- Similar Segments (Lookalike) — automatic extensions of first-party lists. Quality variable per source. Coverage across Search, Display, YouTube, Demand Gen.
The two first-party families not detailed here (Remarketing, Customer Match) are covered in specific articles — they remain the foundation of any performant stack. Complete official documentation on Google Ads support.
When should you use Affinity Audiences?
Affinity Audiences represent the oldest and broadest layer of the Google system. Their logic: categorize users based on durable interests tied to their identity or lifestyle. Typical examples: "Food Enthusiasts", "Sport Lovers", "Luxury Shoppers", "DIY Enthusiasts", "Business Professionals". Google today counts more than 100 Affinity categories and around a hundred customizable sub-categories.
Dominant use: upper funnel, awareness, broad prospecting. You target a lifestyle, not an immediate purchase intent. Direct consequence: Affinity conversion rates are mechanically lower than In-Market or Custom Intent, but cost per thousand impressions is 30 to 40% lower. Perfect for feeding the top of the funnel on a contained budget, especially on YouTube TrueView and Display.
Typical bid modifier when Affinity is activated in Observation: +10% if per-segment performance exceeds +30% vs global account, 0 otherwise. Never switch Affinity alone to Targeting on a direct response campaign: reach dilutes, Quality Score doesn't compensate for the lack of intent signals, and CPA climbs 40 to 60% in 30 days on poorly configured accounts observed in audit.
Use cases where Affinity excels: an organic cosmetics brand wanting to reach "Health & Wellness Enthusiasts" to prepare the ground before retargeting. A local gym targeting "Fitness Enthusiasts" within a 10 km radius. A car manufacturer activating "Car Enthusiasts" ahead of a product launch. Cases where Affinity fails: short-cycle B2B direct response, pure transactional e-commerce, any campaign piloted on strict Target CPA.
When should you use In-Market Audiences?
In-Market is the most exploited intent-driven audience family in 2026 — and the most profitable for the majority of advertisers. Its logic: identify users who currently show purchase intent on a category, via their recent searches, visited sites, videos watched, consulted apps. Google updates assignments every 24 to 72 hours — the rotation is constant, guaranteeing signal freshness.
More than 400 precise categories cover all verticals: "Auto Vehicles/SUV", "B2B Services/CRM Software", "Travel/Cruises", "Financial Services/Credit Cards", "Home Improvement/Kitchen & Bath", etc. Granularity far exceeds Affinity's — where Affinity gives you "Car Enthusiasts", In-Market refines to "Vehicles by Type/SUV" or even "Luxury Vehicles".
Dominant use: mid-to-lower funnel, controlled CPA. On our aggregated data over 90 days, In-Market signals generate +18 to +26% median CTR in Display vs equivalent Affinity (per seasonality), and -15% CPA on well-configured Demand Gen campaigns. Typical bid modifier when In-Market performs: +20 to +30% on segments validated in 14 days of Observation. On Search, activating In-Market in Observation is almost always a win — data surfaces faster.
Classic pitfall: stacking 3-4 In-Market with strict Targeting. Reach collapses (10-20% of potential), algo lacks signals, CPA explodes. Gold rule: 1 to 2 In-Market in Targeting maximum, the rest in Observation. To deepen In-Market integration in an e-commerce setup, see our 2026 Google Ads e-commerce playbook.
How do you build a performant Custom Audience?
Custom Audiences is the most flexible family — and the most powerful for advertisers who know how to leverage it. You provide the signals, Google builds the audience. Two subtypes have coexisted since the 2022 merger.
Custom Intent — the most used. You list a combination of keywords (15-30 transactional high-intent terms) + URLs (5-10 competitor or comparison pages). Google targets users who show behaviors associated with these signals: recent searches, visits to similar pages, engagement with similar content. B2B example: keywords ["B2B CRM", "Salesforce alternative", "HubSpot pricing"] + URLs [competitors, comparison pages] → ultra-intentional audience of decision-makers in evaluation phase.
Custom Affinity — rarer, for specific lifestyle audiences built on apps or sites not covered by standard Affinity categories. Example: an outdoor brand targeting hiking app users and topo-guide site visitors. Secondary use, CPA higher than Custom Intent.
In practice, Custom Intent with competitor keywords shows -18% CPA vs generic Affinity, and +12% CTR vs equivalent In-Market on B2B accounts. It's the #1 lever to scale intent-driven prospecting outside remarketing. Complete technical documentation available on the Google Ads API docs.
Build rule: always ≥ 15 keywords and ≥ 5 URLs so Google has enough signals. Fewer than 10 keywords = audience too narrow, algo refuses to serve. For the complete B2B methodology, follow with our B2B SaaS Google Ads strategy.
Is Detailed Demographics GDPR-compatible?
Detailed Demographics (DD) adds a socio-demographic layer to any campaign. Available in Search, Display, YouTube, Demand Gen. Five main categories exposed by Google:
- Household income — estimate by tiers (top 10%, top 11-20%, top 21-30%, top 31-40%, top 41-50%, lower 50%).
- Parental status — parents (children 0-18 by age tier) vs non-parents.
- Marital status — single, in a relationship, married.
- Education — educational level achieved (high school, bachelor's, graduate).
- Homeownership — owner vs renter.
Relevant use when the product has a clear socio-demographic target. Luxury: high income top 10-20% + 25-54 years. B2B ICP: high income tier (proxy for decision-maker). B2C family: parental status with children's ages. Real estate: homeownership + income. In practice, Detailed Demographics delivers +15% conversions when targeting is relevant, but -8% ROAS when miscalibrated (e.g. income top 10% for a mass-market product).
Technical precaution: if the targeted segment represents less than 1% of the population in the geographic area, Google may ignore the signal — too little data to bias the algorithm. Observe first, restrict only if performance exceeds +40% vs global.
certain Detailed Demographics categories aren't available on all European markets (notably Household Income in Germany). Google's Personalized Advertising Policies also prohibit targeting on health, political orientation, or personal financial situation. Verify local availability before basing a strategy solely on DD.
Which sources should you choose for Similar Segments?
Similar Segments (formerly "Similar Audiences") extend your first-party data by finding users with similar behaviors. Available on Search, Display, YouTube, Discovery and Demand Gen. Their quality depends entirely on the source — it's the most underestimated rule of the lever.
Source hierarchy (best to worst):
- Customer Match — paying customers LTV > 6 months. The reference source. Google finds users who resemble your best customers, not any visitor.
- Remarketing list converters 30 days. Second best. Less precise than Customer Match (some "converters" churned), but wider volume.
- GA4 converter audience imported into Google Ads. Equivalent to remarketing converter list, depends on GA4 tracking quality.
- High-intent Custom Audiences. Weaker quality — Custom signals don't replace pure first-party.
- All site visitors. To avoid. The lookalike resembles your visitors, not your converters — highly diluted quality.
Synthetic rule: the more qualified the source, the better the Similar Segment. Across most cases, Similar segments built from Customer Match customers LTV > 6 months show a CPA 30-40% lower than those built from "all visitors" (per vertical). Investing the time to structure a clean Customer Match list is the best strategic ROI of audience management in 2026.
Since 2023, Google has progressively automated Similar Segments in PMax and Demand Gen. In Search and classic Display, you retain manual control — use it. To dive deeper, see the article complete Performance Max guide.
Which audience combinations win per sector?
Let's get concrete. Here are 4 audience combinations proven on more than 500 accounts in 2025-2026, per vertical. Guiding principle: stack 3 to 4 complementary families, with 1 to 2 in Targeting and the rest in Observation for continuous data collection.
E-commerce — cart abandoner to recover
- Remarketing 1-7 days (Targeting) — warm recovery.
- In-Market "Online Shopping" (Observation, bid +20%) — active intent signal.
- Detailed Demographics income top 30% (Observation) — purchase capacity.
- Similar Segments Customer Match buyers > 2 orders (Observation, bid +15%) — qualified extension.
B2B SaaS — decision-maker intent prospecting
- Custom Intent competitor keywords + comparison URLs (Targeting) — acquisition core.
- Similar Segments Customer Match paying customers > 6 months (Targeting, bid +25%) — high-quality lookalike.
- In-Market "B2B Services" or vertical segment (Observation, bid +20%) — intent validation.
- Detailed Demographics education + income (Observation) — ICP decision-maker proxy.
Local services — plumber, lawyer, physical therapist, coach
- In-Market "Home Services" (Targeting, bid +25%) — immediate intent.
- Detailed Demographics homeowners (Observation, bid +15%) — owner relevance.
- Proximity audiences (geo 10-30 km) — catchment area.
- Remarketing 30 days (Targeting) — local site visitors.
Local retail — physical store with web presence
- Relevant Affinity lifestyle (Observation, bid +10%) — local awareness.
- Detailed Demographics income + education (Observation) — purchasing power.
- Proximity audiences geo ≤ 15 km — effective zone.
- Store visits remarketing if LSA active — recurrence.
CVR ranges are indicative, depend on vertical and average basket. They illustrate qualitative hierarchy: the further down the table, the wider the audience and the weaker the intent signal — so CVR drops and the bid modifier must stay measured. To finely calibrate your bids per audience, follow with our guide to reducing Google Ads CPA.
Observation vs Targeting: which mode when?
The difference between Observation and Targeting (formerly "Audience observation" vs "Audience targeting") conditions the entirety of your audience strategy. Misunderstood, it can sabotage a perfectly designed stack. Let's unpack.
Observation mode. You add an audience to the campaign without restricting reach. All users continue seeing ads; you simply observe performance by audience to adjust bids a posteriori. Perfect for testing an audience without risking volume loss.
Targeting mode. You restrict ads to audience members only (or exclude an undesirable segment). Reach mechanically drops but precision explodes. Use: ultra-qualified audience (Customer Match paying customers), or exclusion (existing customers on acquisition campaign).
Workflow rule:
- Activate all candidate audiences in Observation for 14 to 28 days.
- Measure performance per audience (CTR, CVR, CPA) vs global account.
- Identify audiences with a delta > +40% vs average.
- Switch only these to Targeting with adapted bid modifier.
- Keep the others in Observation for continuous data.
- Resume evaluation cycle every 6 weeks.
Classic pitfall: moving an audience to Targeting on day 1 "just in case." Result: reach divided by 3 to 10, Smart Bidding learning phase that blocks, erratic CPA over 30 days. On our 2025 sector benchmark, 47% of audited accounts use Targeting where Observation would suffice — an immediate volume gain possible on nearly one account in two.
our free Google Ads audit scans your audiences in 72h: detection of miscalibrated Targeting, missing audiences, sub-optimal Similar Segments sources, inconsistent bid modifiers. For continuous adjustment, our Auto-optimization module recalibrates bids per audience every 24h based on performance signals.
To deepen the continuous pilot mechanics consuming these audiences — and avoid the attribution traps that bias their evaluation —, follow with our Google Ads audit checklist. For a macro view on 2026 trends, Think with Google regularly publishes useful sector studies. For the latest product news, see Search Engine Land PPC.
Sources
Official sources consulted for this guide:
FAQ
How do you choose between Affinity and In-Market Audiences?
The rule holds in one sentence: Affinity for latent demand, In-Market for active demand. Affinity categorizes users on durable interests (e.g. 'Food Enthusiasts', 'Luxury Shoppers') — perfect for awareness and upper-funnel prospecting. In-Market detects recent purchase behavior (searches, sites visited over 24-72h) — perfect for mid-to-lower funnel, controlled CPA. On our internal SteerAds benchmark (2,000+ accounts 2025-2026), In-Market shows +22% median CTR in Display vs Affinity, but Affinity remains 30% cheaper in CPC on YouTube TrueView. Use both in parallel per funnel stage, never in competition on the same campaign.
Custom Intent vs remarketing: which to use?
Both, and especially not one in place of the other. Remarketing targets users who already came to your site — it's warm, low CPA but volume capped by your organic traffic. Custom Intent targets users who never came but show intent signals (keywords, competitor URLs) — it's intentional prospecting, wider volume but higher CPA. On our internal SteerAds benchmark, Custom Intent with competitor keywords shows -18% CPA vs generic Affinity, but stays 2 to 3× more expensive than pure remarketing. Allocation rule: 50% prospecting budget (Custom Intent + Similar Segments), 30% remarketing, 20% awareness (Affinity). Never a binary choice.
Is Detailed Demographics GDPR-compatible?
Yes, in its current configuration. Google aggregates data (estimated household income, parental status, education) from anonymized signals — no PII exposed on the advertiser side. You target segments, not named individuals. However, certain sensitive categories (health, personal financial situation, political orientation) are prohibited by Google's Personalized Advertising Policies since 2020. In Europe, you won't see 'Household Income' on all markets — availability depends on local regulators. On our internal SteerAds benchmark, 68% of accounts use Detailed Demographics in Observation mode only, to collect data without restricting reach — prudent and compliant approach.
Can you stack 5 audiences in the same campaign?
Technically yes, strategically rarely. Stacking 5 audiences in Targeting (AND logic) dramatically reduces reach — on Display, a campaign with 5 stacked audiences often reaches less than 5% of the initial target population, algorithm blocked by lack of signals. The 4 families combined remain our recommended maximum: +35% conversion vs 1 single family on the internal SteerAds benchmark. Optimal approach: 2-3 audiences in Observation (collect data, adjust bid modifiers) + 1 key audience in Targeting if ultra-qualified (Customer Match paying customers). Multiplying criteria doesn't make the audience 'better', it makes it invisible.