History length for Autosource in RDF
One of the inputs to AutoSource is the 'today' date. Based on this date,
RDF will figure out the forecast start date by subtracting from this
date the forecast length. During forecast generation it will use
whatever sales history is available before the forecast start date.
Two years are recommended to make sure that all AutoES methods are
evaluated. This is because the Winters-based methods need two full years
to accurately predict seasonality.
However, even with less than two years of sales - counted from the forecast start date - the item/location will get a source level recommendation.
However, even with less than two years of sales - counted from the forecast start date - the item/location will get a source level recommendation.
0