Creative Ops

LinkedIn ad creative testing and iteration engine. Closes the loop performance marketers burn the most cycles on: generating fresh creative variants across multiple angles, structuring statistically valid A/B tests against realistic audience sizes, and learning from results so the next iteration is sharper than the last - not just different. Pairs naturally with Sales Pipeline (audience signals from intent scoring) and Revenue Intelligence (which messaging wins deals downstream).

When to Use

• Test new creative for [audience]
• Generate LinkedIn ad variants for [campaign]
• Design an A/B test for our [value prop] ads
• Why did our last LinkedIn test fail to declare a winner?
• Refresh ads - we're seeing creative fatigue
• Score our LinkedIn ad performance

Workflow

1. Brief Intake - target audience description, value prop, optional current best-performing creative, audience size estimate, weekly budget
2. 3-Angle Variant Generation - produce 6 creatives:
   • 2 × Problem/Solution (lead with pain → product as resolution)
   • 2 × Social Proof (lead with customer outcome → product as the cause)
   • 2 × Contrarian (challenge conventional wisdom → product as the new way) Each variant includes: headline, intro line (first 150 chars), CTA, visual brief, and the single hypothesis it tests
3. Test Structure Design:
   • Map each variant to a budget split (typically equal allocation unless a control needs protection)
   • Compute MDE for the audience size at 80% power, α = 0.05
   • Pick stopping rule (Sequential / Bayesian / Fixed Horizon) based on weekly impressions
4. Pre-Mortem - list the top 3 failure modes with mitigations (e.g., "Audience too narrow → MDE > 30% → relax interest filters or extend test 2 weeks")
5. Live Monitoring Plan - daily KPIs to watch (CTR, CPL, frequency), early-stop triggers, and which variant gets cut first if budget reallocation is needed
6. Post-Test Diagnostic:
   • Winning variant + statistical confidence (posterior probability if Bayesian, p-value if frequentist)
   • Angle-level signal: did one of the 3 strategic angles dominate? If yes, that's a learning, not just a win
   • Segment cuts: did the win hold across job titles / company sizes, or was it driven by one segment?
   • Next-iteration recommendation: which winning element to lock in and which dimension to test next
7. Promote to Playbook - winning creative + full context archived so future iterations build on prior learnings instead of starting from scratch

Reference

Full methodology lives in README.md in this folder: configuration, scoring formulas, output formats, integration points, and worked examples. Read it before producing scored output.