A forecast method that is designed for intermittent and erratic demand, characterized by infrequent events or transactions, which are not necessarily unit-sized. It calculates the estimated interval between demands and the estimated amount of that demand. Croston forecast method requires a minimum of 12 months of demand history. If less than 12 months of demand history are available, then the application will use single exponential smoothing forecast method and post a Review Board record. The Croston method breaks the forecasting process into two components:

These two components are assumed to be independent in Croston method and are generally forecasted using exponential smoothing method. In addition, it is also assumed that there is no seasonality.

Croston is a smoothing algorithm. The initial data set is taken as a whole in the initialization step to produce a forecast size, forecast interval, and forecast standard deviation. Subsequent data smooths the recalculated values. The intermediate values for demand size (set by the crostons_alpha global setting), interval standard deviation, and demand size standard deviation are updated when the forecast is recalculated after each non-zero demand period.