Forecasting Efficient operations require that managers match supply to demand. As they lack a crystal ball, they must rely on forecasting to help predict future demand. Forecasts then drive decisions regarding purchasing, production, and logistics - the key activities in the order fulfillment process. Since most forecasting methods rely on historical data to predict future behavior and are our "best guess," they are almost always wrong. Even so, in the absence of perfect information, good forecasts can improve the quality of our decision-making. Three types of simple forecasts-simple moving average, weighted moving average, and exponential smoothing-are described below. - Simple Moving Average: Uses the average of recent time periods to estimate the next period's demand. Using more time periods, increasing stability. Using fewer time periods, increases responsiveness. - Weighted Moving Average: More recent data may better reflect future behavior. Managers acknowledge this by weighting recent time periods more highly. Managerial judgment (and measurement of forecast error) is used to identify the number of periods and set appropriate weights. Weights must add up to 1.0. - Exponential Smoothing: Sometimes unexpected, and largely random, spikes in demand may occur. To avoid being overly influenced by these spikes, managers may use exponential smoothing, which weights the last period's demand with the last period's forecast. Manager choose a smoothing constant based on whether they have more faith in the actual demand or the previous period's forecast. The formula for exponential smoothing is as follows: Forecast t+1= Actual Demand +(1) Forecast Because forecasts tend to be wrong, it is important to measure how wrong-and then to make_adiustments. to improve the forecasting process. Depending on the industry, forecast errors are often 3080%. Two basic approaches -mean squared error and mean absolute error-are described below. - Mean Squared Error: Is the average of all the squared errors. The result is not very intuitive. - Mean Absolute Deviation: Is the average of absolute values of the difference between the actual and forecast values. Taking the absolute value prevents high and low forecasts from canceling each other out. Problems: Using the numbers in the chart below, calculate the following: Your Answers 1. Three-period moving average: 2. Four-period weighted moving average (weights: p1=.1,p2=.2,p3=.3,p4=.4 ): 3. Smoothed forecast for Period 6 (use an =.3 ): 4. Mean Squared Error for Periods 1-5: 5. Mean Absolute Deviation for Periods 1-5: 6. Average percent error for Periods 1-5