Why SIMPLE
Most commercial forecasting systems have a “best fit” feature which identifies the algorithm that can most closely replicate historical values. Other systems analyze demand history and categorize similar patterns into groups so as to assign the most appropriate algorithm for each group. These approaches are essentially applying a discrete set of solutions to data that is infinitely variable.
The future of statistical time series forecasting will be continuous adaptation, in which algorithm parameters are automatically optimized in response to all possible demand inputs. By evaluating demand history and adjusting its parameters accordingly, the SIMPLE forecast algorithm provides a critical milestone in this evolution.
SIMPLE only requires the user to answer one question: should the forecast be conservative, moderate, or aggressive? General guidelines for selecting a forecast strategy are outlined below.
Conservative
Typically produces better accuracy for:
Near-term forecasting (0 to 5 months forward)
Products with sporadic demand history
Later life-cycle products
Strategy is to avoid over-forecasting
Moderate (default)
Typically produces better accuracy for:
Medium-term forecasting (6 to 11 months forward)
Products with variable demand history
Mature products
Strategy is to minimize forecast bias (neither under-forecasting nor over-forecasting)
Aggressive
Typically produces better accuracy for:
Long-term forecasting (12+ months forward)
Products with consistent demand history
Earlier life-cycle products
Strategy is to avoid under-forecasting