Forecasting Methods for Demand Plans

Application supports the following forecasting methods:

• Regression: The classical regression model is useful in identifying seasonal demands and casual-driven effects of holidays and price.

• Ridge Regression: Regression that safeguards against one or more causal factors getting dramatically larger affects than others. It is often similar to regression.

• Log Transformed Regression: Regression on a log transformed demand pattern. Useful to smooth out variance which can't be easily explained in demand. It is best suited for highly variable demand patterns.

• Holt Exponential Smoothing: Use this method for instances where the amount of data is limited, such as newly introduced products. It creates a level-driven forecast without seasonality or other causal factors.

• Croston Method for Sparse Demand: Use this method when a large amount of historical data is intermittent or spare. This method evaluates periodicity of demand.

• Regression for Sparse Demand: Useful for sparse demand where there are still some seasonal or causal driven impacts.

When you forecast using a demand plan:

• Each item-organization combination that has historical demand is analyzed separately.

• The analysis automatically removes any zero demand entries and fills the missing historical data.

• The analysis also identifies peaks and valleys in the history that are erroneous information or outliers.

• The forecasting process evaluates which of the predefined forecasting methods are most appropriate for analyzing the particular item-organization's historical demand and selects one or more forecasting methods.