Marketer pack
Claude Skill
Paid Ads Campaign Architect
Designs a paid campaign — channel, audience, creative direction, budget split, kill criteria — without burning $20K to learn.
What it does
For a defined offer and budget, picks the right channel(s), audience targeting, creative direction, and measurement setup. Builds an explicit kill-criteria so the campaign doesn't run on autopilot for 6 weeks past the point it should have been stopped.
When to use
- ✓New paid program — first $5-50K test
- ✓Existing campaigns plateauing and you want to redesign
- ✓Launching a new product / segment that needs targeted paid push
When not to use
- ✗You don't have proof people pay for this organically yet — paid will accelerate a non-product-market-fit problem
- ✗Budget < $1K — that's too small to learn anything; use organic
Install
Download the .zip, then unzip into your Claude skills folder.
mkdir -p ~/.claude/skills
unzip ~/Downloads/paid-ads-campaign-architect.zip -d ~/.claude/skills/
# Restart Claude Code session.
# Skill is now available — Claude will use it when relevant.SKILL.md
SKILL.md
---
name: paid-ads-campaign-architect
description: Use when designing a paid advertising campaign. Triggers on "paid ads", "Google Ads", "LinkedIn campaign", "Meta ads", "paid acquisition".
---
# Paid Ads Campaign Architect
Paid advertising is a learning machine OR a money pit. The difference is the design — specifically, the discipline around hypothesis, budget pacing, and kill criteria.
## Required inputs
1. **Offer + landing page** — what you're sending traffic to (page must already convert organic traffic at acceptable rate, or paid won't fix it)
2. **Total budget + timeline** — and what success looks like at end
3. **Target audience** — specific enough to find (not "B2B SaaS")
4. **Existing benchmarks** — CAC, conversion rate, AOV from organic
5. **Channel hypothesis** — where you think they are; we'll pressure-test
## Output
### Channel pick
Match offer to channel:
- **LinkedIn** — high-ACV B2B, specific senior roles, $50-200 CPC
- **Google Search** — high-intent queries, demand already exists
- **Meta** — high-volume B2C, broad audiences
- **TikTok** — younger demographic, demand creation
- **Reddit** — community-fit niche, smaller scale
- **Industry-specific** (newsletters, podcast sponsorships) — small but quality
Skip channels that don't match offer ACV / audience fit. Don't run multi-channel on first $15K.
### Audience targeting
- **Layer 1** — broad enough to learn (top of funnel)
- **Layer 2** — narrow lookalike or intent signal
- **Layer 3** — retargeting of website visitors / video-watchers
Budget split: typically 50/30/20 across layers in week 1, shifts toward winners
### Creative direction
- **Hook variants** — 3-5 distinct angles to test
- **Format** — static / video / carousel — match to channel norms
- **CTA per creative** — specific action, low friction first
- **Compliance check** — claims, before/after, social proof permissions
### Budget pacing
- Week 1: even spend across hypotheses (learn)
- Week 2-3: shift to working creative/audience
- Week 4+: scale or kill
### Kill criteria
The conditions under which we pause:
- CAC > X for 14 consecutive days
- Reach saturation in audience (frequency > 4x)
- Cost-per-conversion 2x our break-even threshold
State these BEFORE the campaign starts. Otherwise it runs forever.
### Reporting
Weekly: spend, leads, CAC, top creative, lessons. Not a vanity dashboard.
## Anti-patterns
- Launching with one creative variant
- "Boost this post" without a strategy
- Optimizing for CTR instead of conversions
- Trusting the platform's "AI optimization" without checking what it actually optimizes for
- Running a campaign past kill criteria because "we're learning"
Example prompts
Once installed, try these prompts in Claude:
- Design first paid test for a B2B vertical SaaS, $15K total budget, 6 weeks. Goal: 50 qualified demos.
- Our LinkedIn campaign is spending $4K/week but CAC has crept to $1200. Help me diagnose + redesign. [paste campaign data]