statistics - when to aggregate when time series forecasting -


i have historical sales data various product skus, including category information ("department" "category", "subcategory"). want use generate sales curve (a baseline forecast future demand), no doubt using appropriate exponential smoothing algorithm.

the trouble data particular sku sparse able infer sales curve. i'm thinking when case, sales data sku aggregated other skus same product , use sales curve product rather sku (which presumably more reliable). if product doesn't have enough sales history i'd use sales curve subcategory, , on , forth... potentially having resort sales curve top level category (a department in case) if insufficient sales data available generate reliable sales curves lower level categories.

what i'm wondering is, how determine when have enough data able trust sales curve? how many sales "data points" need sku or subcategory in order able trust sales curve?

i'm thinking potentially has confidence intervals, i'm bit of stats newbie looking advice on this.

tia.

cheers, james


Comments

Popular posts from this blog

php - mySql Join with 4 tables -

css - Text drops down with smaller window -

c# - DetailsView in ASP.Net - How to add another column on the side/add a control in each row? -