What this Skill helps with

Ad Creative Performance Predictor compares proposed creative concepts with patterns in supplied historical results. It produces a reasoned scorecard for choosing what to test, without presenting the score as a guarantee of future performance.

Who this Skill is for

It is intended for paid-social teams, creative strategists, designers, and founders deciding which concepts should enter a test plan.

What you provide

  • Historical creative results with spend, impressions, clicks, and outcomes
  • Creative tags such as hook, format, message, offer, visual style, and CTA
  • Images, scripts, or descriptions for proposed concepts
  • The campaign objective, audience, platform, and placement
  • Minimum data thresholds and business constraints

What it produces

  • A scorecard for each proposed concept
  • Similar historical examples and the signals they share
  • Confidence notes based on data quantity and consistency
  • Risks, missing inputs, and reasons a comparison may be weak
  • A ranked test shortlist with suggested control concepts

How the Skill works

The workflow groups past creatives by tagged elements, compares their results within similar campaign conditions, and looks for repeated patterns. It then scores new concepts against those patterns while separating evidence from subjective judgment.

Installation and first use

  1. Download and add the Skill to Claude.
  2. Attach a cleaned historical export and its tagging guide.
  3. Describe the proposed concepts and campaign context.
  4. Ask Claude to flag sparse or inconsistent data before scoring.
  5. Use the scorecard to form a test plan, then judge concepts with live results.

Example workflow and expected output

Provide three months of Meta creative results tagged by hook, offer, format, and audience, then add six proposed concepts. Ask for a comparison table containing evidence match, score, confidence, risk, and recommended test order.

Limits and review guidance

Historical association does not prove that a creative element caused performance. Platform delivery, audience, seasonality, spend, and offer changes can distort comparisons. Treat scores as test-planning evidence, not forecasts.

Frequently asked questions

Does the Skill guarantee which ad will win?

No. It ranks concepts using supplied historical patterns. A controlled live test is still required.

Can it work without tagged creative data?

It can help create tags, but confidence will be lower until the historical creatives are classified consistently.

Does it upload ads to a platform?

No. It produces a review and test shortlist only.