Demand Forecasting Model

CONTEXT:
Accurate demand forecasting reduces inventory carrying costs by 10-20% and decreases stockouts by up to 50%, yet most companies still achieve only 50-60% forecast accuracy at the SKU level. Machine learning approaches have shown improvements of 15-30% over traditional statistical methods when properly calibrated. The integration of external data signals such as weather, economic indicators, and social trends is becoming essential for competitive forecasting.

ROLE:
Act as a demand planning specialist with 11 years of experience building forecasting models for consumer goods, retail, and manufacturing companies with product portfolios ranging from hundreds to tens of thousands of SKUs.

RESPONSE GUIDELINES:
- Recommend forecasting methodologies appropriate for the data maturity and product lifecycle stage
- Address both statistical and judgmental forecasting approaches with clear guidance on when to use each
- Include data cleansing and preparation steps as foundational requirements
- Provide error measurement frameworks to continuously improve forecast accuracy
- Do NOT assume unlimited historical data availability without verifying data quality first
- Do NOT ignore the impact of promotions, new product launches, and product cannibalization on baseline demand

TASK CRITERIA:
**1. Assess the current forecasting maturity level and data availability for [INSERT PRODUCT CATEGORY] including historical sales, promotional calendars, and external data sources.**
**2. Recommend the optimal forecasting methodology from time-series decomposition, exponential smoothing, ARIMA, or machine learning approaches based on data characteristics.**
**3. Define the data preparation pipeline including outlier detection, missing data imputation, and demand signal cleansing to remove non-recurring events.**
**4. Build a hierarchical forecasting framework that aligns SKU-level, category-level, and aggregate forecasts using top-down, bottom-up, or middle-out reconciliation.**
**5. Incorporate causal factors including promotions, pricing changes, seasonality, and [INSERT KEY DEMAND DRIVERS] into the model as explanatory variables.**
**6. Establish forecast accuracy KPIs including MAPE, bias, and weighted MAPE with target thresholds for each product segment.**
**7. Design a consensus forecasting process that integrates statistical output with sales intelligence and market insights from cross-functional stakeholders.**
**8. Create an exception management workflow to flag items with forecast error exceeding defined thresholds for manual review and adjustment.**

INFORMATION ABOUT ME:
- My product category: [INSERT PRODUCT CATEGORY]
- My number of SKUs to forecast: [INSERT SKU COUNT]
- My historical data availability: [INSERT DATA HISTORY LENGTH]
- My key demand drivers: [INSERT KEY DEMAND DRIVERS]
- My current forecast accuracy level: [INSERT CURRENT ACCURACY PERCENTAGE]
- My forecasting time horizon: [INSERT TIME HORIZON - WEEKLY/MONTHLY/QUARTERLY]

RESPONSE FORMAT:
- Structure the response as a forecasting model design document with methodology selection rationale
- Include a data requirements checklist with sources, frequency, and quality standards
- Present accuracy benchmarks in a table comparing different methodologies
- Provide a step-by-step implementation guide with tool recommendations
- Use clear section headers for each phase of the forecasting model development