meta-ads-audience-architect

name

meta-ads-audience-architect

description

Build and optimize Meta Ads audiences for ecommerce targeting. Use when users ask about audience targeting, lookalike audiences, custom audiences, retargeting setup, interest/behavior targeting, audience sizing, or targeting strategy for Facebook/Instagram ads. Triggers include "build audience", "create lookalike", "retargeting strategy", "who should I target", or any audience-related Meta ads questions.

Audience Architect Skill

Build, research, and optimize Meta Ads audiences for ecommerce targeting.

Skill Overview

The Audience Architect skill enables strategic audience building for Meta Ads campaigns. It guides the selection, creation, sizing, and optimization of targeting approaches—from broad prospecting to precise retargeting.

Core Principle: The Modern Targeting Hierarchy

2026 TARGETING EFFECTIVENESS (Updated January 2026):

MOST EFFECTIVE → Broad + Andromeda-optimized diverse creative
                 Advantage+ Shopping Campaigns (full automation)
                 Advantage+ Audience (A+A) - NOW DEFAULT
                 Creative-as-targeting (Andromeda semantic matching)
                 Lookalikes as SIGNALS (not hard boundaries)
                 Custom audiences (retargeting/exclusions only)
DECLINING      → Manual interest/behavior stacking
DEPRECATED     → Detailed targeting EXCLUSIONS (removed March 2025)
                 ASC existing customer budget cap (removed 2025)

KEY INSIGHT: Meta's Andromeda algorithm processes 10,000x more data
per impression. Creative diversity > audience selection.
Let the algorithm work. Provide diverse creative, not restrictions.

Critical 2026 Platform Changes

⚠️  IMPORTANT UPDATES TO UNDERSTAND:

1. ANDROMEDA ALGORITHM (Full rollout October 2025)
   - Creative IS the targeting - Andromeda matches ad semantics to users
   - Similar creative variations clustered into single "Entity ID"
   - 50 similar ads may only get 1 auction opportunity
   - SOLUTION: Diverse creative (format, persona, benefit) > volume

2. ADVANTAGE+ AUDIENCE IS NOW DEFAULT
   - A+A automatically enabled on new ad sets
   - Expands beyond your targeting when better opportunities exist
   - Use "Audience Controls" for hard limits (age, geo, exclusions)
   - Turn OFF only for strict retargeting or special ad categories

3. ASC EXISTING CUSTOMER CAP REMOVED
   - No longer available in Advantage+ Shopping Campaigns
   - Workaround: Manual campaign with customer list exclusion
   - Or: Two ad sets (new vs existing) + Advantage Campaign Budget

4. DETAILED TARGETING EXCLUSIONS REMOVED (March 2025)
   - Cannot exclude by interests/behaviors anymore
   - Custom audience exclusions STILL WORK
   - Campaigns with old exclusions stop delivering Jan 15, 2026

5. INTEREST CATEGORIES CONSOLIDATED (June 2025)
   - Many specific interests merged into broader categories
   - Verify availability in Ads Manager before planning

Conversions API (CAPI) Requirement

🔴 CRITICAL FOR AUDIENCE QUALITY:

Without proper CAPI setup, you're losing 20-40% of conversion data.
This directly impacts:
- Custom audience accuracy
- Lookalike audience quality
- Algorithm optimization

CAPI CHECKLIST:
□ Server-side events sending alongside pixel
□ Event Match Quality >90% in Events Manager
□ Deduplication properly configured
□ All key events tracked (Purchase, ATC, VC, IC)

Monitor via: Events Manager → Data Sources → [Your Pixel] → Overview

When to Use This Skill

Use audience-architect when:

• Planning targeting strategy for new campaigns
• Building custom audiences for retargeting
• Creating lookalike audiences
• Researching interest/behavior options
• Troubleshooting audience-related performance issues
• Expanding to new markets or segments
• Optimizing audience overlap and exclusions

Audience Strategy Decision Tree

START: What is the campaign goal?
│
├─► PROSPECTING (New Customers)
│   │
│   ├─► Budget $100+/day AND 50+ purchases/week possible?
│   │   ├─► YES → Use ASC or Broad Targeting
│   │   │         Load: audience-sizing.md
│   │   │
│   │   └─► NO → Need more targeted approach
│   │             │
│   │             ├─► Have 1000+ customer data?
│   │             │   ├─► YES → Build Lookalikes
│   │             │   │         Load: lookalike-strategies.md
│   │             │   │
│   │             │   └─► NO → Use Interest/Behavior targeting
│   │             │             Load: interest-categories.md
│   │             │             Load: behavior-taxonomy.md
│   │             │
│   │             └─► Very niche product?
│   │                 └─► YES → Layer interests + behaviors
│   │                           Load: audience-layering-exclusions.md
│   │
│   └─► EXCLUSIONS NEEDED
│       └─► Exclude website visitors + purchasers from prospecting
│           Load: audience-layering-exclusions.md
│
├─► RETARGETING (Warm Audiences)
│   │
│   ├─► Website-based retargeting
│   │   └─► Load: custom-audience-types.md (Website Visitors section)
│   │
│   ├─► Engagement-based retargeting
│   │   └─► Load: custom-audience-types.md (Engagement section)
│   │
│   └─► Customer list retargeting
│       └─► Load: custom-audience-types.md (Customer Lists section)
│
└─► EXPANSION (Scale existing success)
    │
    ├─► Geographic expansion
    │   └─► New country lookalikes from existing customers
    │       Load: lookalike-strategies.md
    │
    └─► Audience expansion
        └─► Broader LAL percentages or new LAL sources
            Load: lookalike-strategies.md
            Load: audience-sizing.md

Reference Files

When to Load Each Reference

File	Load When
interest-categories.md	Researching interest targeting options, building interest audiences
behavior-taxonomy.md	Researching behavior targeting, purchase behaviors, device usage
custom-audience-types.md	Building any custom audience (website, engagement, customer list)
lookalike-strategies.md	Creating lookalikes, optimizing LAL performance, choosing sources
audience-sizing.md	Validating audience size, troubleshooting delivery, sizing decisions
audience-layering-exclusions.md	Combining audiences, building exclusion ladders, preventing overlap

Audience Building Workflows

Workflow 1: New Campaign Audience Setup

STEP 1: Define campaign type
├─► ASC Campaign → Minimal audience work (algorithm handles it)
│   - ⚠️ Existing customer cap REMOVED (2025)
│   - No detailed targeting needed
│   - For new customer focus: Use manual campaign instead (see below)
│
├─► Manual Campaign (New Customer Prospecting)
│   - Exclude customer list + website purchasers
│   - Use Advantage+ Audience with exclusions
│   - This replaces old ASC customer cap functionality
│
└─► Manual Campaign (Full Control) → Full audience setup needed
    Continue to Step 2...

STEP 2: Build prospecting audiences
├─► Primary: Broad (age/gender/geo only)
├─► Alternative: 1-3% Lookalike of purchasers
└─► Test: Interest stack or Advantage+ Audience

STEP 3: Build retargeting audiences
├─► Hot: Cart abandoners (7 days)
├─► Warm: Product viewers (14 days)
├─► Cool: All visitors (30 days)
└─► Past purchasers (for exclusion + cross-sell)

STEP 4: Set up exclusions
├─► Prospecting excludes: All website visitors, purchasers
├─► Each retargeting tier excludes: Higher-intent tiers
└─► Verify no overlap between campaigns

Workflow 2: Lookalike Audience Creation

STEP 1: Identify best source audience
├─► Preferred: Purchasers (all or high-value)
├─► Alternative: Add to cart, email subscribers
└─► Minimum: 1,000 people in source

STEP 2: Select percentage
├─► Testing: Start with 1%
├─► Scaling: Expand to 1-3%, then 3-5%
└─► Reach: 5-10% for maximum scale

STEP 3: Choose location
└─► Same country as source, or new expansion market

STEP 4: Validate size
├─► Too small (<100K): Consider broader %
├─► Optimal: 500K-5M for most campaigns
└─► Too large (>10M): May need narrowing

STEP 5: Create and apply
└─► Use in ad set targeting
└─► Exclude existing customers if prospecting

Workflow 3: Custom Audience Setup

WEBSITE VISITORS:
├─► All visitors (retention: 30-180 days)
├─► Product/category viewers (retention: 7-30 days)
├─► Cart abandoners (retention: 7-14 days)
├─► Purchasers (retention: 30-180 days)
└─► High-value page visitors (pricing, checkout)

ENGAGEMENT AUDIENCES:
├─► Video viewers (25%, 50%, 75%, 95%)
├─► Page/post engagers (liked, commented, shared)
├─► Instagram profile visitors
├─► Lead form openers/submitters
└─► Shopping/collection engagers

CUSTOMER LISTS:
├─► All customers (email + phone)
├─► High-value customers (top 25%)
├─► Recent purchasers (90 days)
├─► Lapsed customers (no purchase 180+ days)
└─► Email subscribers (non-purchasers)

MCP Tool Integration

Audience-Related Tools

When building audiences programmatically via MCP:

# Get available targeting options
targeting_options = meta_ads_mcp.get_targeting_options(
    account_id=account_id,
    targeting_type="interests"  # or "behaviors", "demographics"
)

# Search for specific interests
interests = meta_ads_mcp.search_interests(
    query="fitness",
    limit=25
)

# Create custom audience (website visitors)
custom_audience = meta_ads_mcp.create_custom_audience(
    account_id=account_id,
    name="Website Visitors - 30 Days",
    subtype="WEBSITE",
    rule={
        "inclusions": {
            "operator": "or",
            "rules": [
                {
                    "event_sources": [{"id": pixel_id}],
                    "retention_seconds": 2592000,  # 30 days
                    "filter": {"operator": "and", "filters": []}
                }
            ]
        }
    }
)

# Create lookalike audience
lookalike = meta_ads_mcp.create_lookalike_audience(
    account_id=account_id,
    name="LAL 1% - Purchasers - US",
    origin_audience_id=source_audience_id,
    targeting_spec={
        "geo_locations": {"countries": ["US"]}
    },
    ratio=0.01  # 1%
)

# Get audience size estimate
size_estimate = meta_ads_mcp.get_audience_size(
    account_id=account_id,
    targeting_spec={
        "geo_locations": {"countries": ["US"]},
        "age_min": 25,
        "age_max": 54,
        "genders": [2],  # Female
        "interests": [{"id": "6003139266461", "name": "Fitness"}]
    }
)

Quick Reference: Audience Type Selection

Scenario	Recommended Audience	Reference
ASC campaign	No detailed targeting (built-in)	—
Broad prospecting	Age/gender/geo only	audience-sizing.md
Data-driven prospecting	1% LAL of purchasers	lookalike-strategies.md
Niche product prospecting	Layered interests + behaviors	interest-categories.md, behavior-taxonomy.md
Cart abandonment	Website custom audience	custom-audience-types.md
Past purchaser cross-sell	Customer list custom audience	custom-audience-types.md
Video viewer retargeting	Engagement custom audience	custom-audience-types.md
Scaling existing campaign	Broader LAL or geo expansion	lookalike-strategies.md

Common Patterns

Pattern 1: Standard Ecommerce Audience Stack

PROSPECTING:
- Primary: Broad or ASC
- Secondary: 1% LAL Purchasers
- Exclusion: All site visitors (30 days), purchasers (180 days)

RETARGETING:
- Tier 1: ATC 7 days (exclude purchasers)
- Tier 2: VC 14 days (exclude ATC, purchasers)
- Tier 3: Visitors 30 days (exclude VC, ATC, purchasers)

RETENTION:
- Past purchasers 30-180 days
- Exclude recent purchasers (14 days)

Pattern 2: High-AOV / Long Consideration

PROSPECTING:
- Awareness: Broad, optimize for reach/video views
- Consideration: 1-3% LAL, optimize for content views
- Conversion: Warm audiences only

RETARGETING:
- Extended windows (30-60 days for each tier)
- More touchpoints allowed
- Lead capture integration

Pattern 3: New Account / Low Data

PROSPECTING:
- Week 1-2: Broad targeting, optimize for traffic/VC
- Week 3-4: Transition to purchase optimization
- Month 2+: Build first LALs from initial purchasers

RETARGETING:
- Start with any available data
- Expand as audiences build

Audience Performance Diagnosis

SYMPTOM: High CPM, low reach
├─► Audience too narrow
│   Action: Broaden targeting, remove restrictions
│   Load: audience-sizing.md
│
├─► Audience overlap/competition
│   Action: Check overlap tool, adjust exclusions
│   Load: audience-layering-exclusions.md
│
└─► High-value audience (competitive)
    Action: May be normal, test broader audiences

SYMPTOM: Low conversion rate from audience
├─► Wrong audience for product
│   Action: Review interest relevance
│   Load: interest-categories.md
│
├─► Audience too broad (low intent)
│   Action: Layer additional signals
│   Load: audience-layering-exclusions.md
│
└─► Creative mismatch
    Action: Align creative to audience (creative-studio skill)

SYMPTOM: Audience saturation (high frequency)
├─► Audience too small
│   Action: Expand LAL %, broaden targeting
│   Load: audience-sizing.md
│
├─► Budget too high for audience
│   Action: Reduce budget or expand audience
│   Load: audience-sizing.md
│
└─► Natural saturation
    Action: Refresh creative, expand audiences

Integration with Other Skills

Skill	Integration Point
campaign-launcher	Audience setup during campaign creation
creative-studio	Match creative to audience characteristics
performance-optimizer	Audience-level performance analysis
ecommerce-catalog	Audience signals from catalog engagement
duplicator	Audience variations for geo/demographic expansion

Key Principles

1. Broad > Narrow in most cases with sufficient budget
2. Creative diversity > audience selection - Andromeda matches creative to users
3. Lookalikes are signals - provide direction, not hard boundaries in A+A
4. Custom audience exclusions - the only exclusion type still available
5. CAPI is mandatory - without it, audiences degrade significantly
6. Size matters - too small = learning limited, too large = diluted
7. Diverse creative formats - avoid 50 similar variations (clustered by Andromeda)
8. Test before scaling - validate audience before major budget commitment

Andromeda-Era Creative Strategy

FOR BEST AUDIENCE MATCHING:

DO:
✓ Use diverse creative formats (static, video, carousel, UGC)
✓ Show different personas, environments, and benefits
✓ Let Andromeda's semantic understanding find your audience
✓ Provide 10+ meaningfully different creative concepts

DON'T:
✗ Create 50 variations of same concept (becomes 1 Entity ID)
✗ Over-rely on manual targeting to find customers
✗ Ignore creative diversity in favor of audience complexity
✗ Fight the algorithm with narrow restrictions