Meta Ads Campaign Builder

8 steps · start to finish.

1. 1
   Step 1

   Environment Setup

   ▶
   1. Create the results directory.
   2. Validate required parameters are present: product name, product URL, objective, ICP, monthly budget, landing pages, and business type.
   3. Normalize the objective to one of awareness, traffic, lead_generation, conversions, or app_installs.
   4. Initialize validation_report.json with a setup stage.

   mkdir -p "{{ results_dir }}"
   python - <<'PY'
   import json, os, pathlib
   results = pathlib.Path(os.environ.get('RESULTS_DIR', '{{ results_dir }}'))
   results.mkdir(parents=True, exist_ok=True)
   required = ['product_name', 'product_url', 'campaign_objective', 'icp', 'monthly_budget', 'landing_pages', 'business_type']
   missing = [name for name in required if not os.environ.get(name.upper())]
   report = {'version':'1.0.0','stages':[{'name':'setup','passed':not missing,'message':'Missing: '+', '.join(missing) if missing else 'Required inputs present'}]}
   (results / 'validation_report.json').write_text(json.dumps(report, indent=2))
   if missing:
       raise SystemExit(1)
   PY
   

2. 2
   Step 2

   Intake and Objective Selection

   ▶

   Collect or confirm the campaign intake:

3. 3
   Step 3

   Research Audiences and Competitors

   ▶

   Run bounded research for targeting and messaging. Use max 3 rounds: one round for product/category targeting, one for ICP-specific interests or behaviors, and one for competitor Meta Ad Library findings when competitor names are available.

4. 4
   Step 4

   Build Campaign Architecture

   ▶

   Design the campaign tree with enough consolidation to exit learning phase where possible.

5. 5
   Step 5

   Generate Copy Framework and Budget Plan

   ▶

   For each ad set, generate 3-5 copy variants using these angles: pain, outcome, social proof, contrarian, and product-led. Include placement-specific guidance for Feed, Stories, Reels, Right Column, and Audience Network.

6. 6
   Step 6

   Write Outputs

   ▶

   Write `meta_campaign_plan.md` with this structure:

7. 7
   Step 7

   Evaluate Outputs

   ▶

   Run programmatic checks and update `validation_report.json`:

8. 8
   Step 8

   Iterate on Errors (max 3 rounds)

   ▶

   If validation fails or finds partial coverage, perform up to max 3 rounds of targeted fixes. Fix missing files first, then missing sections, then weak assumptions or unsupported recommendations. Stop early once all required checks pass.