Forecasting with Prophet
Best practice forecasting starts with the configuration of suitable forecasting methods and continues with consistent review and resolution of Forecast Alerts.
For time-phased models that bring in supply from outside vendors, the best practice is to use Installed Base or Auto-Select of statistical forecast methods.
- The Installed Base Demand forecast method calculates demand forecasts based on an estimate of month-to-month changes to
contracts. Installed Base forecasting provides more accurate predictions because it is based on historical and current contract data
while also capturing future contracts to allow for proactive planning of expected increases or decreases. Further, this method is
better equipped to adjust to demand spikes and enhances the lifecycle management of materials.
- The Auto-Select method uses an algorithm to determine the statistical equation that best fits the historical data and considers
forecasting methods such as Simple Moving Average, Croston’s Intermittent Demand, Automatic Exponential Smoothing, and BoxJenkins.
Forecasting with Auto-Select is highly configurable, allowing customers to set specific parameters within the methods such
as default number of historical months or outlier corrections settings that smooth spikes and dips.
For re-order point models setting optimal target stock levels at field sites, best practice is to employ rate-based methods to use the number of contracts and calculated demand rates to determine the forecast. With rate-based forecasting, projected demand is based on the number of contracts and the average number of occurrences per contract in either a single region or globally. Additionally, Prophet has developed forecasting tools specific to technician planning that allows an aggregate forecast at a team level to be allocated to individual technicians on that team. Technician forecast allocation distributes forecast to preferred technicians based on training, current inventory, forecast, and demand history of previous service calls to improve min stability and allocate forecast to the best fit technicians.
Once an appropriate forecasting method has been configured, Forecast Alerts allow planners to manage by exception so that data falling outside of pre-defined limits generates an alert for the designated planner. Forecast Alerts measure large deviations from an archived set of benchmark forecasts and alert parameters and thresholds are adjustable to match criteria unique to the customer’s environment. For planners that manage hundreds or thousands of products and materials, alerts help organize and prioritize planning activities.