The most sensible approach is to look at the actual past demands. Some demands show some kind of trend or cycles, which could be used for our advantage and to forecast more accurately.
The common behaviors of the demand is as following:
- Stationary: here the demand show a smooth pattern where no increase or decrease in the demand.
- Linear: an steady increase or decrease in the demand
- Nonlinear: Where the demand takes a weird increasing or decreasing slopes.
- Trends:
- Seasonal: Where the demand is repeated after a certain period
- Cycle: this is easily detected graphically where the demand repeats in each cycle.
- Random: The most annoying type. it maybe meaningless to forecast such kind of behavior. However, the industrial engineer could still simplify the behavior and remove outliers from consideration.
Each one of these cases has its own way to forecast.
