forecasting-techniques

# Forecasting Techniques ## Metadata - **Name**: forecasting-techniques - **Description**: Multiple methods for projecting future values - **Triggers**: forecasting, projections, growth rate, CAGR, market prediction ## Instructions Apply forecasting techniques to project $ARGUMENTS into the future. Choose appropriate method based on data availability and context. ## Framework ### Three Main Approaches | Method | Data Required | Time Horizon | Precision | Best For | |----------|----------------|--------------|------------| | **Time Series Extrapolation** | 5-10 years of historical | Short-medium | High | Stable environments | | **Derived Demand** | Proxy variables, cross-correlation | Short-medium | Medium | Related markets | | **Expert Opinion** | Structured surveys | Any | Low | New products | ### 1. Time Series Extrapolation **Trend Analysis** - Simple growth rate: Compound annual growth (CAGR) - Linear regression: Straight line fit to historical data - Moving average: Smooths volatility, lags trends - Exponential smoothing: Recent trends weighted more heavily **Steps:** 1. Gather historical data (3+ years preferred) 2. Analyze patterns (cycles, seasonality, trends) 3. Choose model (CAGR, regression, etc.) 4. Apply to future periods 5. Validate against expert opinion **Example Output:** ``` Year | Historical | Projected | Growth Rate | |------|------------|------------|-------------| | 2023 | $100 M | - | - | | 2024 | $115 M | +15% | CAGR = 15% | | 2025 | $132 M | +15% | CAGR = 15% | | 2026 | $152 M | +15% | CAGR = 15% | | 2027 | $175 M | +15% | CAGR = 15% | ``` ### 2. Derived Demand **Proxy Methodology** - Identify proxy variable that correlates with demand - Use readily available data with reliable trend - Apply correlation coefficient - Adjust for unique factors **Examples:** - GDP growth as proxy for consumer spending - Housing starts as proxy for home goods - Demographics for category-specific demand **Steps:** 1. Identify correlation (r² should be > 0.5) 2. Gather proxy data 3. Apply coefficient 4. Adjust for local factors 5. Add confidence intervals ### 3. Expert Opinion **Structured Survey Method** - Multiple expert interviews - Weighted by expertise or track record - Delphi technique (iterative rounds) - Scenario-based questioning **Advantages:** - Captures qualitative insights - Accounts for disruptive changes - Incorporates expert judgment **Process:** 1. Define forecasting questions 2. Select experts (diverse backgrounds) 3. Conduct interviews (structured format) 4. Aggregate with weighting 5. Present scenarios (base, optimistic, pessimistic) 6. Review and iterate if needed ## Output Process 1. **Define scope** - What's being forecasted? 2. **Select method** - Based on data and time horizon 3. **Gather inputs** - Historical data, drivers, expert inputs 4. **Apply technique** - Run the chosen method 5. **Calculate projections** - For each year/period 6. **Validate** - Cross-check with other methods 7. **Add scenarios** - Best, base, worst case 8. **Document assumptions** - Clearly state all key inputs ## Output Format ``` ## Forecasting Analysis: [Subject] ### Forecast Methodology **Method Used:** [Time Series/Derived Demand/Expert Opinion] **Time Horizon:** [Years] **Base Year:** [Year] **Data Quality:** [High/Medium/Low] --- ### Projections | Metric | 2024 | 2025 | 2026 | 2027 | 2028 | CAGR | |--------|--------|--------|--------|--------|--------|------| | Revenue | $X M | $Y M | $Z M | $W M | $V M | % | | Growth | X% | Y% | Z% | W% | % | --- ### Key Drivers | Driver | Impact | Uncertainty | Scenario Impact | |--------|---------|-----------------|--------------| | [Driver 1] | High | Medium | [Description] | | [Driver 2] | Medium | Low | [Description] | | [Driver 3] | Low | High | [Description] | --- ### Scenarios | Scenario | 2028 Revenue | Probability | Key Assumptions | |----------|----------------|------------------|----------------| | **Base** | $X M | 50% | [Assumptions] | | **Optimistic** | $Y M | 30% | [Assumptions] | | **Pessimistic** | $W M | 70% | [Assumptions] | --- ### Confidence Intervals | Metric | Low | Base | High | Confidence | |--------|------|------|------|------|----------| | 2028 Revenue | $X ± Y% | $Z M | $W M | 80% | ``` ## Tips - Triangulate methods when possible - Use multiple methods for cross-validation - Be explicit about assumptions - don't hide them - Present confidence intervals for transparency - Consider mean reversion - growth rates tend toward averages - Validate with real outcomes when available - Document track record of forecasts - improve over time ## References - Makridakis, Spyros. *Business Forecasting*. 1998. - Armstrong, J. Scott. *Principles of Forecasting*. 2001. - Wikipedia. "Forecasting - Methods and Applications" (multiple sources)