SUO_MGT3059 W2 L3 Forecasts And Errors

Forecasts and Errors Can We Predict The Future? Let’s use the following table to demonstrate how forecasts differ on the basis of the selected technique. Using each technique, you’ll estimate the demand for Week 10. Week Demand 1 960 2 1,340 3 1,790 4 1,500 5 1,220 6 1,710 7 1,140 8 1,030 9 1,560 Naïve Approach : This is the easiest forecasting technique. The forecast for the demand in the next period is equal to the sales in the previous period. Therefore, the forecast for the demand in Week 10 is 1,560. Moving Averages : This technique is slightly more difficult. You need to use the historical demand over a specific number of periods in order to estimate the demand in the next period. If you’re using a three-week moving average, the forecast for the demand in Week 10 will be equal to the sum of the demand in the previous three weeks divided by 3. (1,140 + 1,030 + 1,560)/3 = 1,243.33 ~ 1,244 units Note : In forecasting, you always need to round off (represented by ~) because you can’t produce partial units such as 1,243.33. Rounding is usually done in the upward direction, as you want to be sure to able to plan on enough inventory (or capacity) for the level of demand that you are anticipating. Exponential Smoothing : This technique is a bit more complicated than the moving average forecast. Exponential smoothing uses a smoothing factor, α (the Greek symbol alpha), having a value between 0 and 1 to weigh the difference between the demand and forecast for the last period. For example, if the forecast for Week 9 is 1,294 ([1,710 + 1,140 + 1,030]/3) using a three week moving average and α = 0.15, the forecast for Week 10 will be:

Forecasts and Errors Can You Predict The Future? 1,294 + 0.15(1,560 – 1,294) = 1,294 + 0.15(266) = 1,294 + 39.9 = 1,333.9 ~ 1,334 Now we turn our attention to measures that check the accuracy of a forecast. As stated in the video lecture, three of the most commonly used measures of forecast accuracy are MAD , MSE , and MAPE . The following table shows the forecast and actual demand data for a product for five weeks. On the basis of this information, we wish to calculate the MAD, MSE, and MAPE, in order to determine how effective the forecasts have been. Note that the “Error Squared” column in the table is simply generated by multiplying the value in the “Deviation” column by itself (“squaring” it). Also, note that the “Absolute Percent Error” column is calculated by dividing the value in the “Absolute Deviation” column by the value in the “Actual” column, and expressing it as a percentage (by multiplying it by 100). Week Forecast Actual Deviation Absolute Deviation Error Squared Absolute Percent Error 1 10 12.4 2.4 2.4 5.76 19.35% 2 10 8.2 -1.8 1.8 3.24 21.95% 3 10 11.2 1.2 1.2 1.44 10.71% 4 10 9.7 -0.3 0.3 0.09 3.09% 5 10 10.7 0.7 0.7 0.49 6.54% Sum 6.4 11.2 0.616552 MAD = Sum of absolute deviation / n = 6.4/5 = 1.28 MSE = Sum of errors squared / n = 11.02/5 = 2.204 MAPE = (100 × Absolute percent error) / n = (100 × 0.616552)/5 = 12.33% Therefore: MAD MSE MAPE 1.28 2.204 12.33% What do these values tell you? Is a MAD of 1.28 good or bad? This is difficult to say. Organizations may compare such values among several forecasting techniques (like those discussed above) in order to determine what is “good enough” as a forecast. The MAD can change based on the magnitude of the data. The MAD measures the dispersion of the observed values from the expected value. You organization must make a judgment as to whether or not your forecast is doing an adequate job in terms of the MAD value. The MSE measures the squared differences between the actual demand and the forecast demand. Like the MAD, the MSE can change based Page 2 of 3 Operations Management ©2017 South University