Demonstrates: How to restructure an existing prompt using CRAFTER framework
Version: 0.2
Pattern: Meta-Prompt Enhancement
Use Case: Transforming a vague marketing prompt into structured CRAFTER format
The Challenge
A marketing manager has this prompt:
You're a data analyst. Look at this sales data and find insights.
Problems with this prompt:
- ❌ Vague action (“find insights” — what kind?)
- ❌ No format specification (table? paragraphs? bullets?)
- ❌ Missing context (what product? what timeframe?)
- ❌ No audience specified (who will read these insights?)
- ❌ No examples of what “good insights” look like
- ❌ No iteration guidance
Result: Unpredictable output quality, hard to reuse, difficult to debug.
Mode A Process: Enhancement
Step 1: Analyze Original Intent
Core goal: Extract actionable insights from sales data
Implied context:
- Marketing team context
- Need for decision-making support
- Time-sensitive business need
Missing elements:
- Specific analytical steps
- Output format
- Target audience and communication style
- Examples of quality insights
- Refinement guidance
Step 2: Map to CRAFTER
What’s already there:
- R (partial): “data analyst” → We can expand this
- A (vague): “find insights” → Needs specific steps
What needs to be added:
- C: Business context, data constraints
- F: Output structure
- T: Who will use this + appropriate tone
- E: Example insights
- R: How to iterate
Step 3: Fill Gaps & Restructure
Now we restructure into proper CRAFTER format…
Context
You’re analyzing Q3 2024 sales data for an e-commerce company selling outdoor gear. The data includes:
- 50K transactions across 3 product categories
- Geographic data (North America only)
- Customer segments: new vs returning
- Marketing channel attribution
The marketing team needs insights to plan Q4 campaign strategy. Budget decisions are due next week.
Role
You are a Data Analyst specializing in e-commerce analytics and customer behavior patterns. Your expertise includes:
- Cohort analysis and customer segmentation
- Marketing attribution modeling
- Identifying trends and anomalies in sales data
- Translating data patterns into actionable business recommendations
Action
Follow these steps:
- Identify top 3 trends in the sales data
- Look for patterns by: product category, customer segment, geography
- Quantify each trend with specific metrics
- Analyze anomalies or surprises
- Compare Q3 2024 vs Q3 2023
- Flag any unexpected patterns (positive or negative)
- Generate strategic recommendations
- For each trend, suggest one specific marketing action
- Estimate potential impact (high/medium/low)
- Include confidence level for each recommendation
- Validate insights
- Check that each insight is actionable
- Ensure recommendations align with budget constraints
Structure your analysis as a Markdown report with three sections:
1. Key Trends (300 words max)
- Bullet list with 3 trends
- Each trend includes: metric, change %, and implication
2. Anomalies & Surprises (200 words max)
- 2-3 unexpected patterns
- Brief explanation of why each matters
3. Recommendations Table
| Recommendation |
Target Segment |
Est. Impact |
Confidence |
Priority |
| [Action] |
[Who] |
[H/M/L] |
[H/M/L] |
[1-3] |
Target & Tone
Target: Marketing managers and VP Marketing (action-oriented professionals, moderate data literacy, need quick decisions for budget planning)
Tone:
- Direct and scannable — lead with key takeaways
- Use business language, not statistical jargon
- Quantify insights when possible (percentages, dollar amounts)
- Provide clear next steps with priority levels
- Assume moderate comfort with data (explain complex methods briefly)
Examples
Example Insight #1
Input data pattern: New customer acquisition increased 35% in August, primarily from Instagram ads (45% of new customers vs 20% in Q2)
Output insight:
**Trend:** Instagram driving new customer growth
- New customers from Instagram: +35% vs Q2 2024
- Conversion rate: 3.2% (above platform average of 2.1%)
- Average order value: $85 (consistent with other channels)
**Implication:** Instagram ads are working efficiently for acquisition
**Recommendation:** Increase Instagram ad budget by 25% in Q4, focusing on
similar creative formats (user-generated content, product demonstrations).
Est. Impact: HIGH. Confidence: HIGH. Priority: 1
Example Insight #2
Input data pattern: Returning customers in camping category dropped 15% in September
Output insight:
**Anomaly:** Returning customer drop in camping category
- Q3 2023 returning customers: 12K
- Q3 2024 returning customers: 10.2K (-15%)
- Most pronounced in September (-22%)
**Implication:** May indicate product quality issues, inventory problems,
or competitive pressure
**Recommendation:** Conduct customer survey with Sept camping purchasers.
Investigate product reviews and competitor pricing. Est. Impact: MEDIUM.
Confidence: MEDIUM. Priority: 2
Refining
If the user requests changes:
Framework: CoachSteff’s CRAFTER (SuperPrompt Framework v0.2)
Pattern Used: Meta-Prompt Enhancement (Mode A)
License: CC-BY 4.0 — Attribution: Steff Vanhaverbeke (coachsteff.live)
📊 Changes Made
Added Components
Context (C):
- Business environment specified (e-commerce outdoor gear)
- Data scope defined (50K transactions, Q3 2024, North America)
- Decision timeline added (budget decisions next week)
- Stakeholder urgency clarified
Action (A):
- Broke “find insights” into 4 concrete steps
- Added specific deliverables for each step
- Included validation criteria
Format (F):
- Markdown report with 3 sections
- Word count limits for scannability
- Recommendation table structure specified
Target & Tone (T):
- Identified specific audience (marketing managers, VP Marketing)
- Defined their characteristics (action-oriented, moderate data literacy)
- Specified communication approach (direct, quantified, business language)
Examples (E):
- Two complete input→output examples
- Showed how to structure insights
- Demonstrated recommendation format
Refining (R):
- Four common adjustment scenarios
- Specific guidance for each refinement type
Enhanced Components
Role (R):
- Expanded “data analyst” to include specific expertise areas
- Added e-commerce and customer behavior specialization
- Clarified what analytical lens to apply
Preserved Elements
Original intent: Extract insights from sales data to inform marketing decisions
Original role: Data analyst perspective (expanded but not changed)
User’s domain: E-commerce and marketing context maintained
Key Learning: Before vs After
Before (Original Prompt)
- 1 component partially specified (Role)
- 13 words
- Vague output expectations
- Hard to reuse or debug
- Unpredictable quality
After (Enhanced Prompt)
- 7 components fully specified
- 650+ words
- Clear output structure
- Reusable across similar analyses
- Predictable, high-quality results
- Includes validation and iteration guidance
Time investment: 10 minutes to enhance
Quality improvement: 5-10x more predictable output
Reusability: Can now be used for any Q3/Q4 sales analysis with minor adjustments
When to Use This Pattern
Use Mode A (Meta-Prompt Enhancement) when:
- ✅ You have an existing prompt that works “okay” but could be better
- ✅ Output quality is inconsistent
- ✅ You’re getting different results each time you run the prompt
- ✅ You want to make a prompt reusable
- ✅ You need to debug why a prompt isn’t working
- ✅ You’re onboarding someone and want to standardize prompts
Don’t use this pattern when:
- ❌ Starting from scratch (use Mode B: Superprompt Creation instead)
- ❌ The original prompt is fundamentally wrong (redesign is better than enhancement)
Next Steps:
- Try this enhancement pattern with your own prompts
- Use the self-test checklist to verify all 7 components are present
- Compare output quality before and after enhancement
- See example-mode-b-creation.md for building prompts from scratch
Questions or feedback?
Repository: https://github.com/CoachSteff/superprompt
Author: Steff Vanhaverbeke (@CoachSteff)