Dimensions 2

Well as I wrote that I was going off to think up some new W words I knew it was a bad idea!

Traditional dimensions in retail data warehousing have exciting names such as TIME, PRODUCT, CUSTOMER, STORE; see not a W in sight. Other data warehouse subject domains may have other dimensions in the core set of dimensions.

But other dimensions can be found when you look in more depth at the data. Take the humble product. In retail products often have a set of attributes such as colour, pack dimensions, storage requirements, original supplier etc. We need to consider each attribute and decide whether the attribute has a use in aggregations of fact data. I think it unlikely that we would want to aggregate items based on colour - what use is a comparison of the sales of blue items against red items?

But it is more useful to have aggregates based on original supplier, after all a lot of businesses need just that type of information to enable them to negotiate the best deals for buying in there goods, and some retailers, especially the large supermarkets make good money selling supplier sales data back to those suppliers. Here we need to think is this attribute best exposed an alternative hierarchy through the dimension or to make it a dimension its own right. I tend to go for the separate dimension for the simple reason that I can aggregate the product to a higher level and still maintain visibility of supplier. For example Proctor & Gamble make laundry products, beauty products and snack foods (and lots more besides) if our product dimension had a a supplier hierarchy we could only report total Proctor & Gamble sales but by making supplier a dimension we can access sales by sector for the supplier (P&G sales for pet foods for example), which makes the data much more useful and saleable