AD HOC CLASSIFICATION

One common mistake in data warehouse analysis is the failure to plan for new classifications of data warehouse attributes. These classifications are not always known in advance, and it is not uncommon to see that the delivery of the data warehouse provides end users with a mechanism for identifying new classifications.

In our example for Guttbaum’s Grocery, we might see arbitrary, or ad hoc, groupings of data attributes. These ad hoc groupings of existing data attributes might be used to perform what-if analyses for decision support. Some examples of ad hoc classification might include:

* A “yuppie” (young urban professional)--This is an individual in age category two or three, with an income greater than $50,000 a year, who owns his or her home and has less than four children. Show me a breakdown by product category for all yuppie expenditures on non-food items.

* A “dink” (dual-income, no-kids)--This is a family unit where there are two wage earners with a combined yearly income greater than $60,000. Show me the buying habits of dinks for dairy products.

* A “cheapskate”--This is an individual who uses coupons for more than 50 percent of their purchases. Show me the buying habits of cheapskates for all non-coupon purchases.
As you can see, the ability of the data warehouse to develop arbitrary classifications can greatly improve the usefulness of a data warehouse. Note that these classifications do not always form a hierarchy, and that they may sometimes be required for the pre-calculation of aggregate values. For example, we may need our data warehouse to pre-summarize sales by product classes and brands for yuppies, dinks, and cheapskates.