Ad Campaign Optimization

Ad Campaign Optimization

Overview

Optimize paid advertising across platforms — Google Ads, Meta (Facebook/Instagram), TikTok, LinkedIn, Twitter/X. Improve ROAS, reduce CAC, and scale winning campaigns.

Instructions

Campaign structure

Organize campaigns by objective, then ad sets by audience, then ads by creative variant:

Account
├── Campaign: Prospecting (Cold)
│   ├── Ad Set: Lookalike 1% (interest-based seed)
│   │   ├── Ad: Video A — problem/solution hook
│   │   ├── Ad: Video B — testimonial hook
│   │   └── Ad: Static C — benefit-focused
│   ├── Ad Set: Interest targeting (competitor audiences)
│   │   ├── Ad: Video A
│   │   └── Ad: Static D — data-driven hook
│   └── Ad Set: Broad targeting (algorithm-optimized)
│       ├── Ad: Video A
│       └── Ad: Video E — UGC style
│
├── Campaign: Retargeting (Warm)
│   ├── Ad Set: Website visitors 7-30 days
│   ├── Ad Set: Video viewers 50%+ (14 days)
│   └── Ad Set: Cart abandoners (7 days)
│
└── Campaign: Retention (Existing customers)
    ├── Ad Set: Upsell (purchased product A)
    └── Ad Set: Win-back (inactive 60+ days)

Key principles:

• Separate cold, warm, and hot audiences into different campaigns (different budgets, different optimization)
• Use Campaign Budget Optimization (CBO) within each campaign
• Exclude audiences across campaigns (retarget pool excluded from prospecting)
• Keep 3-5 ads per ad set minimum for creative rotation

Audience strategy

Prospecting (cold):

• Lookalike audiences: Seed from highest-value customers, start with 1% lookalike, expand to 2-5% as you scale
• Interest-based: Layer interests with demographics. Instead of "fitness" (too broad), use "fitness AND CrossFit AND 25-44"
• Broad targeting: On Meta, broad targeting often outperforms detailed targeting at scale

Retargeting (warm) — build exclusion-layered audiences:

Tier 1 (hottest): Cart/checkout abandoners, 0-7 days
Tier 2: Product page viewers, 7-14 days
Tier 3: Any website visitor, 14-30 days
Tier 4: Video viewers (50%+), 14-30 days
Tier 5: Social engagers, 30-60 days

Each tier excludes all tiers above it.
Tier 1 gets highest bid/budget (closest to conversion).

Lookalike seed quality (in order): Top 25% LTV customers > Repeat purchasers > All purchasers > Add-to-cart users > High-engagement visitors. Minimum seed: 1,000 users.

Creative strategy

Break winning ads into components:

HOOK (first 3 seconds)
├── Pattern interrupt: unexpected visual/sound
├── Curiosity gap: "I tried X for 30 days..."
├── Problem callout: "Tired of [specific pain]?"
└── Social proof: "500K people already switched"

BODY (next 10-20 seconds)
├── Problem amplification → Solution introduction
├── Proof elements: testimonials, data, demos
└── Differentiation: why this, not alternatives

CTA (final 3-5 seconds)
├── Direct: "Start your free trial"
├── Urgency or risk reversal
└── Social: "Join 50,000 happy customers"

Formats by platform:

• Meta: 15-30s vertical video, carousels (3-5 cards), static images, UGC-style
• TikTok: Native-feeling video, 1-2s hook, text overlays, Spark Ads
• Google: Search (headline = keyword match + benefit + CTA), Performance Max (diverse assets), YouTube bumpers
• LinkedIn: Document ads, thought leadership ads, lead gen forms

Creative testing:

• Phase 1: Test 3-5 hooks/angles, $20-50/day each, 3-5 days → winner by CTR and CPA
• Phase 2: Test 3-5 variations of winner, $30-75/day, 5-7 days → winner by CPA and ROAS
• Phase 3: Scale winners 20-30%/day, refresh at frequency >3.0

Bid strategy and budget

Awareness:    CPM bidding, optimize for reach
Consideration: CPC bidding or landing page view optimization
Conversion:   CPA/ROAS bidding (need 50+ conversions/week)
Retention:    Value-based bidding (optimize for LTV)

Start with 70/20/10 split: 70% prospecting, 20% retargeting, 10% testing. Scale winners by increasing budget 20-30% every 3 days.

Meta and Google need 50 conversion events per ad set per week to exit the learning phase. If not hitting this: consolidate ad sets, move optimization event up the funnel, or increase budget.

Attribution

Last-click:       Simple but undervalues awareness
First-click:      Values discovery but ignores nurturing
Time-decay:       More credit to recent touchpoints
Data-driven:      ML-based, available at scale (Google, Meta)

Cross-platform solutions: UTM parameters (tag every link), incrementality testing (10% holdout), Marketing Mix Modeling (statistical model), post-purchase surveys.

Performance metrics

EFFICIENCY: CPA (3:1), CTR (1-2% Meta, 3-5% Google Search), CPC
QUALITY: Conversion rate, bounce rate, frequency (3.0)
- Run incrementality tests quarterly to validate platform-reported attribution
- Start with proven formats (UGC video, testimonial) before testing experimental creative
- Keep at least 3 ads per ad set for rotation and learning

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [TerminalSkills](https://github.com/TerminalSkills)
- **Source:** [TerminalSkills/skills](https://github.com/TerminalSkills/skills)
- **License:** Apache-2.0
- **Homepage:** https://terminalskills.io/

Install and usage instructions live in the source repository linked above.