# Study Notes: Business Forecasting ## Overview of Business Forecasting Business forecasting involves predicting future values or trends to aid decision-making. Forecasts can be short-term, medium-term, or long-term, depending on the context and objectives. Effective forecasting requires careful consideration of the following key questions: ### Key Questions for Forecasting - **Why is a forecast needed?** - Predict sales, manage stock, plan operations, expand into new markets, or support decision-making. - **Who will use the forecast?** - Stakeholders such as managers, investors, or supply chain teams. - **What data is available?** - Assess the quality, accessibility, and relevance of the data. - **What is the time period of the forecast?** - Define the "forecast horizon" (short, medium, or long-term). - **How much accuracy is needed?** - Determine whether a rough guide or precise planning is required. - **What will the forecast cost?** - Consider costs in terms of time, money, expertise, and software. Always align the forecasting method with the objective and available resources. --- ## Objectives of Forecasting Forecasting serves two primary purposes: 1. **Produce predictions of future values** - Methods: Time series extrapolation (e.g., Naïve, Smoothing, ARIMA, LSTMs). 2. **Establish causal relationships** - Methods: Regression-type models to identify factors predictive of the variable of interest. --- ## Characteristics of Data The choice of forecasting method depends on the characteristics of the data: ### Key Data Features - **Seasonality** - Use methods like Seasonal ARIMA, Holt-Winters, or seasonal naïve models. - **Random component dominates trend** - Single Exponential Smoothing works best, assuming trends and turning points are unpredictable. - **Trend dominant** - Use methods that model and project trends, such as ARIMA, Holt’s method, or regression. - **Increasing variance** - Apply multiplicative models (e.g., Holt-Winters) or transform data (e.g., logarithmic or square root) to stabilize variance. For seasonal data, a retailer might use Holt-Winters to forecast holiday sales based on past seasonal patterns. --- ## Number and Frequency of Forecasts Forecasting requirements vary based on the number and frequency of forecasts: - **Large number of items, frequent basis** - Example: Inventory forecasts (daily, weekly, monthly). - Methods should be simple, automated, and include feedback systems. - **Small number of items, infrequent basis** - Example: Yearly sales forecasts. - Methods can be more sophisticated, with qualitative approaches playing a larger role. For frequent forecasts, automation reduces effort and ensures consistency. --- ## Adequacy of Quantitative Models Quantitative models should meet the following criteria: - Parsimonious (simple yet effective). - Residuals should resemble white noise and be approximately Normally distributed. - Parameters must be statistically significant. - Forecasting errors should be "small enough" for practical use. Many students overlook residual analysis, which is critical for validating model adequacy. --- ## Forecasting Timescales Forecasting timescales depend on the type of decision being made: | **Timescale** | **Type of Decision** | **Examples** | |------------------|----------------------------|---------------------------------------| | **Short-term** | Operating decisions | Inventory control, production planning | | **Medium-term** | Tactical decisions | Leasing equipment, employment changes | | **Long-term** | Strategic decisions | R&D, mergers, product changes | Timescales are context-specific. Adapt the forecast horizon to your organization's needs. --- ## Quantitative Methods Quantitative methods rely on historical data and assume past relationships will hold in the future. They are best suited for shorter-term forecasts. ### Common Quantitative Methods | **Method** | **Description** | **Range** | **Min Data Requirement** | |--------------------------|-----------------------------------------------|-----------------|--------------------------| | **Naïve models** | Simple user-constructed models | Short | Low | | **Decomposition** | Investigates data structure | Short to Medium | Low to Medium | | **Moving average** | Smooths time series data | Short | Low | | **Exponential smoothing**| Weighted smoothing models | Short to Medium | Low | | **ARIMA** | Iterative approach for time series modeling | Short to Medium | Medium/Large | | **Regression** | Time series or explanatory forecasting | Short to Medium | Medium/Large | | **LSTMs** | Neural network models for complex patterns | Short to Medium | Very Large | Use ARIMA for data with complex patterns and relationships, but ensure sufficient historical data is available. --- ## Long-Term Forecasting Long-term forecasting requires different approaches due to the unpredictability of historical patterns over extended periods. Methods include: - **Judgmental forecasting** - Relies on expert opinions and qualitative methods. - **Scenario planning** - Develops narratives for plausible futures (optimistic, pessimistic, and most likely scenarios). Scenario planning might consider the impact of emerging technologies on market demand over the next decade. --- ## Dangers of Extrapolation Extrapolation can lead to inaccurate forecasts, especially for longer-term predictions. For example: - **IBM’s Forecast Failure** - IBM predicted exponential growth in sales and profits, ignoring competitive pressures and market changes. This led to hiring 100,000 new staff unnecessarily. Never assume the business environment will remain static during the forecast period. --- ## Judgemental Forecasting Judgmental forecasting is used when data is scarce or irrelevant. Steps include: 1. **Select judges** - Choose experts with relevant knowledge. 2. **Pose questions** - Frame questions to elicit meaningful forecasts. 3. **Obtain forecasts** - Use methods like interviews, questionnaires, or Delphi panels. ### Basic Judgmental Methods - Personal interviews - Telephone interviews - Mail questionnaires - Delphi panels (reduce bias) - Focus groups - Role-playing scenarios Judgmental forecasts are generally less accurate than statistical methods but can be useful in data-scarce situations. --- ## Lead Indicators Lead indicators signal future events and are useful for forecasting. Examples include: - **Baby boom** - Signals increased demand for housing, cars, and household goods 15-25 years later. - **Housing market** - Expands before economic growth and contracts before recessions. - **Machine orders** - Indicates business confidence and capital spending. Rising machine orders might signal an upcoming economic expansion. --- ## Analogies Analogies use historical data from similar situations to make forecasts. They are useful for: - **Short-term**: Effects of special events or competitive actions. - **Medium-term**: Predicting recessions. - **Long-term**: Sales of new products. Ensure the analogy is justified and relevant to the current context. --- ## Scenario Writing Scenario writing involves creating narrative forecasts for long-term predictions. Scenarios consider factors like: - New technologies - Population changes - Consumer preferences ### Types of Scenarios - **Most likely scenario** - **Optimistic scenario** - **Pessimistic scenario** Scenario writing encourages long-range thinking and prepares management for environmental changes. --- ## Summary: Choosing a Forecasting Method When selecting a forecasting method, consider: - **Objective**: What are you trying to achieve? - **Data availability**: Is there sufficient and relevant data? - **Frequency**: How often are forecasts needed? - **Timescale**: Short, medium, or long-term? Combine quantitative methods with judgmental approaches for comprehensive forecasts. --- ## Exam Preparation Tips - **Timing**: Practice managing your time during the exam. - **Question order**: Answer questions in the order that suits you best. - **Clarity**: Write clearly and concisely. - **Stay calm**: You’ve covered all necessary material in class. Use the formula sheet provided in the exam to save time and avoid errors.