90 likes | 276 Views
ISQA 451 Business Forecasting. Dr. Alan Raedels C.P.M. Summer 2010 Week 6. Problems of Forecasting Intermittent Demand . Traditional forecasting methods ignore problem of frequent zeroes. Wrongly assumes normal distribution for demand. Judgment adjustments not feasible.
E N D
ISQA 451 Business Forecasting Dr. Alan Raedels C.P.M. Summer 2010 Week 6
Problems of Forecasting Intermittent Demand • Traditional forecasting methods ignore problem of frequent zeroes. • Wrongly assumes normal distribution for demand. • Judgment adjustments not feasible. • Not practical with a large number of items
Benefits of Bootstrap Method • Avoids the need for unrealistic assumptions • Re-samples historical demand data to create simulated data sets that are statistically similar to observed data. • No need to perform complicated time series models.
Sources of Forecast Error • The equation is miss specified. • Variables are missing • Variables are entered improperly. • Incorrect values are assumed for missing data. • Data series are revised. • The structure of the underlying process has changed.
Forecasting Methodology Determine the: • Variables to be predicted. • Principle use of the forecast • Time frame • Data adequacy • Acceptable error range • Symmetric or asymmetric loss function • Conditional or unconditional forecasts • Top-down or bottom-up forecast
Forecasting Methodology • Structural or time-series model • Identify the main drivers • Use of judgment • Determining the sample period error • Alternative methods of forecasting • Scenario buildup and intervention analysis • Accuracy of independent variables • Selling to top management or other clients
General Approach • Graphical explanation of the key drivers affecting the variable to be forecast. • Presentation of the equation used for forecasting. • Discussion of the macroeconomic outlook. • Discussion of other key factors. • Presentation of forecast for next two years. • Discussion of other variables that could be relevant. • Alternative scenarios and their probabilities. • Presentation of alternative forecast for next two years/