The better understanding the model has of the dynamics of the business and the causal drivers the better demand forecasts are possible, normally the model gets better as more demand drivers are provided such as
- macroeconomic factors,
- covid restrictions,
- pricing etc.
When one has the data for all the drivers one needs to consider, one also needs to decide the granularity of the data one wants to use. In general, the lower the time scale the more stochastic sales and demand tend to be, so depending on the specific application one must find correct data granularity, should one predict hourly, daily, or weekly demand?