meta-ads-optimizer

Process

1. Diagnose — Identify the specific performance issue (rising CPA, declining ROAS, learning phase stuck, placement inefficiency, etc.)
2. Check tracking first — Most Meta performance issues trace back to signal quality. Verify CAPI, EMQ, event deduplication before optimizing bids/audiences
3. Analyze root cause — Use the Diagnostic Decision Tree
4. Select optimization lever — Match problem to framework
5. Implement changes — Respect learning phase (don't touch ad sets <7 days old or with recent edits)
6. Monitor for 5-7 days minimum — Meta needs time to stabilize after changes
7. Iterate — Small, incremental adjustments

Diagnostic Decision Tree

Learning Phase Management

Meta's most critical optimization principle: keep ad sets OUT of learning phase as much as possible. Meta research: advertisers with <20% of spend in learning see up to 68% lower CPA.

Learning Phase Rules

Fixing "Learning Limited" Status

Consolidation rule of thumb: If you have 5+ ad sets stuck in Learning Limited, consolidate to 2-3 larger ad sets with pooled budget. The algorithm prefers fewer, larger ad sets with sufficient event volume.

Audience Optimization

Audience Refresh Triggers

Expansion Tactics

2025 best practice: Accounts with strong Pixel + CAPI signals typically find that broader audiences outperform narrower ones. The algorithm does the targeting work — your job is providing creative that converts.

Exclusion Maintenance

Post-March 2025: Detailed targeting exclusions removed. Use customer list exclusions instead.

Budget Scaling Playbook

Vertical Scaling (increase budget on winners)

Why 20%? Larger jumps can reset the learning phase or destabilize delivery by shifting the algorithm's pacing too quickly.

Horizontal Scaling (duplicate or expand)

CPM Context (Critical 2025 Factor)

Meta's average ad price rose 9% in 2025 (Meta Q1 2025 earnings). When diagnosing rising CPA, always check if CPM inflation is a contributing factor — it may be partial cause of efficiency decline, not creative or audience failure.

Placement Optimization

Advantage+ Placements (Default)

Let Meta distribute across all eligible placements. Recommended default unless:

• You have creative only in certain specs (e.g., only 1:1 square — missing 4:5 Feed and 9:16 Reels)
• You have strict brand safety requirements
• You've verified via placement breakdown that specific placements consistently underperform

Manual Placement Review

Check placement-level performance monthly:

Decision rule: Exclude a placement if its CPA is >2x account average for 30+ days with sufficient volume (100+ clicks).

Automation Rules

Set up rules in Meta Ads Manager OR use third-party tools (Birch — formerly Revealbot, Madgicx).

Core Automation Rules

Best practices:

• Start with notification-only rules before auto-actions
• Set conservative thresholds first — tighten over time
• Always require minimum volume (100+ clicks, 10+ conversions) before rules trigger
• Exclude learning phase ad sets from optimization rules

ASC Optimization

Advantage+ Shopping Campaigns have limited manual levers. Optimize these:

When to ditch ASC: If ASC underperforms manual campaigns after 30 days with strong creative and tracking, the account may not have sufficient conversion volume or signal quality for ASC to work. Fall back to manual 3-campaign structure (Prospecting + Retargeting + Testing).

Output Format

## Meta Ads Optimization Plan

### Current Performance Summary
| Metric | Current | Target | Gap |
|--------|---------|--------|-----|
| CPA | $[X] | $[X] | [gap] |
| ROAS | [X]x | [X]x | [gap] |
| Frequency | [X] | <4 | [status] |
| Learning phase spend % | [X]% | <20% | [status] |

### Root Cause Diagnosis
[Specific reason for the issue, tied to the Diagnostic Decision Tree]

### Optimization Actions (ICE-prioritized)
| Priority | Action | Lever | ICE Score | Timeline |
|----------|--------|-------|-----------|----------|
| 1 | [action] | [audience/budget/placement/automation] | [score] | [when] |

### Automation Rules to Add
[Specific rules with thresholds]

### Monitoring Plan
- Check [metric] after [X] days (minimum 5-7 for learning phase)
- Success criteria: [specific]
- Rollback plan: [if it doesn't work]

Example

Input: "I run an ecommerce store on Meta. Spending $12K/month. ROAS was 4.5x last month, now 2.8x. Frequency on my top ad set is 5.2. Using ASC. What should I do?"

Output:

Meta Ads Optimization Plan

Current Performance Summary

Root Cause Diagnosis

Creative fatigue is the primary driver. Frequency at 5.2 is well past Meta's fatigue threshold (>4). Meta's own research: 45% conversion rate drop at 4+ exposures. Your ROAS decline (-38%) is consistent with severe fatigue.

Secondary factor: Meta's average CPM rose 9% in 2025, contributing some background cost pressure.

Optimization Actions

Creative Refresh Plan

Use the meta-ads-creative-engine skill for systematic creative testing. Immediate priorities:

1. Pause 3 highest-frequency creatives today
2. Add 6 new variants across formats: 2 static, 2 video (15-30 sec, 9:16 vertical), 2 UGC-style
3. Ensure new creative has strong 3-second hooks and captions (85% of Facebook videos watched without sound)
4. Maintain 15-30 active creatives in ASC (max ceiling is 150)

Automation Rules to Add

Monitoring Plan

• Day 3: Verify new creatives are spending and delivering
• Day 7: Check frequency — should be dropping toward 3-4 range as new creatives absorb impressions
• Day 14: ROAS should start recovering toward 3.5-4.0x range as fatigue clears
• Success criteria: ROAS above 3.8x within 21 days, frequency stable below 4.0
• Rollback: If ROAS continues declining after creative refresh, the problem is deeper (tracking issue, offer/product problem, market shift). Use meta-ads-audit skill for full diagnostic.

Guidelines

• Don't optimize during learning phase. 5-7 days minimum. Changes reset learning. The only exceptions are clear tracking failures (stop the bleeding) or disapprovals.
• Don't change budgets by more than 20% at once. Even on winners, larger jumps destabilize Smart Bidding pacing.
• Don't ignore creative frequency. It's the #1 leading indicator of Meta performance decline. Measure at the CREATIVE level, not ad or ad set level.
• Don't over-segment ad sets. 2025 algorithm prefers consolidated ad sets with 50+ conversions/week. Combine warm and cold audiences. Let the algorithm sort.
• Don't run rules without volume minimums. Rules triggering on 5 clicks will false-positive constantly. Require 100+ clicks and 10+ conversions minimum.
• Don't skip tracking checks when performance drops. A "performance issue" is often a tracking issue. Verify Pixel fires, CAPI fires, EMQ is stable before touching campaigns.
• Don't forget CPM context. Meta's avg CPM +9% in 2025. Some rising-CPA cases are partial CPM inflation, not creative/audience failure.
• Offline Conversions API discontinued May 2025. If troubleshooting offline tracking, migrate to standard CAPI with action_source: physical_store or system_generated.
• Attribution API change (October 2025): 7d_view and 28d_view windows removed from Insights API. If reports pull these, update reporting.
• Instant Experience change (June 2025): No longer counts as a landing page view. Don't build funnel metrics assuming it does.
• Cross-references: For tracking issues, use the meta-ads-tracking-setup skill. For creative testing and fatigue management, use the meta-ads-creative-engine skill. For full account diagnostic, use the meta-ads-audit skill. For Google Ads optimization, use the google-ads-optimizer skill. For cross-channel budget allocation, use the ad-spend-allocator skill.
• Confidence: Automation rule thresholds are starting points — calibrate to your account's volume and variance. ASC's ~12% CPA reduction vs manual is a Meta benchmark; results vary significantly by account.