Dear Statistics Gurus,
May be this is not the right forum for this question but trust some of you might admire science of forecasting as much as technology of DP
Hypothesis: Forecast based on weekly data points is more accurate than monthly data points
Science Says: Forecast in larger periodicities is more accurate than forecast in smaller periodicities. Well science actually says reducing variability improves predictability... which means aggregate forecasts are more accurate ( variability pooling)
I can test this on the system with the data I have but probably with a risk of not concluding anything useful... because each time series is so different. May be I end up more confused and hence confusing the planners even more.
Question1:
What business case / motivation (should) exists to recommended weekly forecasts to anyone if current basis is using month periodicities in past and future.
Question2:
What would be a good practice for someone who is a market leader in chips and cola with several years of presence in the market.
For simplicity sake
Assume forecast is carried out for individual SKU only, ignore geographical and organizational dimensions in aggregation.
Assume forecast using a model like seasonal trend model.
Assume the historical data is available and stored Weekly (I have real sales history by week)
I realize answers can vary but a couple of examples / experiences is probably what I am looking if not a statistic to test my hypothesis.
Thanks
Borat.