The “Creative Performance Coach Prompt” describes how to use Polar MCP together with Claude to automatically analyze ad-creative performance data — especially from Facebook Ads — and generate clear, structured insights on what types of ads work best (by format, spend level, creative source, device, etc.). It lays out exactly what data to supply (ad-level spend, creatives, naming conventions, time period) and defines the sections of the output: top/bottom creatives by return, creative-format performance, fatigue analysis, creative-source breakdowns, product-level ad performance, device/placement performance, suggested new creative tests, and recommended budget reallocation.
Ensure you have the Polar MCP installed (see Using Polar MCP).
Create a new Project in Claude.
In the Instructions box paste the instructions from below.
Update the
USER INPUTSto specify details.To generate a report, start a new chat in the project with something like: “run the report”.
You can then:
ask follow-up questions to go deeper
tell Claude you want to amend the report
customise the instructions yourself for future use
Instructions
Put these in a Claude Project “Instructions” fields.
USER INPUTS:
- Analysis Period: {time_period} [default: last 45 days]
- Minimum Creative Spend: ${min_spend} [default: $50]
- Creative Naming Convention: {naming_convention} [to parse creative types]
PROMPT:
Pull Facebook Ads performance data at the ad level for {time_period}, analyzing creative performance patterns.
**Overall Creative Performance:**
Only include creatives with >${min_spend} spend
- Total creatives analyzed: X
- Total spend on creative: $X
- Overall creative ROAS: X
**Performance by Creative Format:**
- Video: X creatives, $X spend, X ROAS, X% CTR
- Image: X creatives, $X spend, X ROAS, X% CTR
- Carousel: X creatives, $X spend, X ROAS, X% CTR
- Collection: X creatives, $X spend, X ROAS, X% CTR
**Top 10 Performing Creatives:**
Ranked by ROAS (minimum {min_spend} spend):
1. [Creative ID/Name]: X ROAS, $X spend, X purchases, X% CTR
2. [Continue for top 10]
**Common Elements in Top Performers:**
{if naming_convention provided, parse for patterns}
- Hooks that work: [identify patterns]
- Visual themes: [products, lifestyle, UGC]
- Message themes: [benefit-focused, problem-solving, social proof]
- Length (for video): [average duration of top performers]
**Creative Fatigue Analysis:**
For creatives running >14 days:
- Week 1 performance: X ROAS
- Week 2 performance: X ROAS (% change)
- Week 3+ performance: X ROAS (% change)
- Optimal refresh frequency: Every X days
**Performance by Creative Source:**
{if identifiable in naming}
- UGC content: X ROAS, $X spend
- Studio/branded: X ROAS, $X spend
- Influencer: X ROAS, $X spend
- Graphics/designed: X ROAS, $X spend
**Product Performance in Ads:**
- Top 5 products by creative ROAS
- Bottom 5 products (consider removing from ads)
**Device-Specific Performance:**
- Mobile creative performance: X ROAS
- Desktop creative performance: X ROAS
- Placement optimization opportunities
**Creative Testing Insights:**
- Average test duration to significance: X days
- Minimum spend for reliable data: $X
- Win rate of new creatives: X%
**5 New Creative Concepts to Test:**
1. **Based on top performer elements:**
- Format: [video/image/carousel]
- Hook: [specific hook that works]
- Expected ROAS: X (based on similar creatives)
2. **Address fatigue in top spender:**
- Refresh [Creative ID] with [specific changes]
- Maintain [winning elements] but update [fatigued elements]
3. **Scale winning product:**
- Create 3 variations for [top product]
- Different angles: [list angles]
4. **New format test:**
- Test [format] which is underutilized
- Opportunity: $X spend reallocation
5. **Seasonal/timely angle:**
- Leverage [upcoming event/season]
- Historical performance of seasonal creatives: X ROAS
**Creative Budget Allocation:**
Current allocation:
- Top 20% of creatives get X% of budget
- Recommended reallocation:
- Increase top performers by $X
- Pause bottom 20% (saving $X)
- Test budget: $X (10% of total)
**Expected Impact:**
- Pausing underperformers: Save $X/month
- Scaling winners: +$X revenue/month
- Net impact: +X% ROAS improvement