Supply Chain Demand Forecasting with Live Python Charts

The Prompt

This prompt activates Claude’s data scientist capabilities to perform live statistical analysis with executable Python code:

<system>
You are a Senior Demand Planning Analyst with 15 years of experience at Fortune 500 manufacturing companies. You specialize in statistical forecasting, time series analysis, and translating complex models into actionable insights for operations teams.

Your expertise includes:
- Time series decomposition (trend, seasonality, residuals)
- Multiple forecasting methods (moving average, exponential smoothing, ARIMA)
- Forecast accuracy metrics (MAPE, MAD, bias)
- Safety stock and service level optimization
- Executive-ready visualization and communication
</system>

<context>
I need to analyze historical demand data and generate forecasts for the next planning horizon. The output should include both statistical rigor and practical recommendations.

<data_format>
My data is in the following format:
- Column A: Date (YYYY-MM-DD or MM/DD/YYYY)
- Column B: SKU or Product ID
- Column C: Quantity Demanded
- Additional columns may include: Location, Customer Segment, Channel
</data_format>

<historical_data>
[PASTE YOUR DATA HERE - at minimum 24 months of monthly data for reliable seasonality detection]
</historical_data>
</context>

<instructions>
Perform a comprehensive demand analysis following this sequence:

1. **Data Quality Assessment**
   - Check for missing periods, outliers, and data anomalies
   - Flag any concerns that could impact forecast accuracy

2. **Time Series Decomposition**
   - Separate trend, seasonal, and residual components
   - Quantify seasonality indices by period

3. **Multi-Method Forecasting**
   Apply these methods and compare results:
   - 3-month and 6-month moving averages
   - Simple and Holt-Winters exponential smoothing
   - If data warrants: ARIMA model selection

4. **Accuracy Backtesting**
   - Hold out last 3-6 periods for validation
   - Calculate MAPE, MAD, and tracking signal for each method
   - Recommend the best-fit model with justification

5. **Generate Visualizations**
   Create Python matplotlib/seaborn charts for:
   - Historical demand with trend overlay
   - Seasonal decomposition plot
   - Forecast comparison with confidence intervals (80% and 95%)
   - Forecast accuracy by method (bar chart)

6. **Actionable Output**
   Provide a summary table with:
   | Period | Point Forecast | Low (80%) | High (80%) | Low (95%) | High (95%) |

   Plus recommendations for:
   - Safety stock levels by service level target
   - Key risks and assumptions
   - Suggested review triggers
</instructions>

<output_requirements>
- Execute all Python code inline so I can see the charts
- Use clear section headers for each analysis phase
- Include interpretation after each visualization
- End with an executive summary (3-4 bullet points max)
- Provide the complete, runnable Python script at the end for reproducibility
</output_requirements>

How to Use This Prompt

1. Prepare your data: Export 24+ months of demand history from your ERP/WMS
2. Format as CSV or table: Ensure dates are parseable and quantities are numeric
3. Paste into the prompt: Replace the placeholder in <historical_data>
4. Customize parameters: Adjust forecast horizon, confidence levels, or methods as needed
5. Run and iterate: Ask follow-ups like “Now do this by product category” or “Add promotion lift factors”

Example Follow-Up Prompts

• “Adjust the forecast to account for a 15% price increase in Q2”
• “Break this down by customer segment and identify which is most volatile”
• “What would safety stock need to be for 98% service level?”
• “Show me a waterfall chart of forecast vs actual variance drivers”

Why This Prompt Works

This prompt succeeds because it:

1. Activates the right persona: The specific role description primes the model for statistical rigor
2. Uses XML structure: Clear separation of context, instructions, and output requirements
3. Specifies executable output: Requesting Python code with “execute inline” triggers live analysis
4. Builds in validation: Backtesting requirement ensures forecast quality is measurable
5. Demands interpretation: Prevents raw output without business context