Building the Data Warehouse

Скачать в pdf «Building the Data Warehouse»
Comparing Internal Data to External Data

One of the most useful things to do with external data is to compare it to internal data over a period of time. The comparison allows management a unique perspective. For instance, being able to contrast immediate and personal activities and trends against global activities and trends allows an executive to have insights that simply are not possible elsewhere. Figure 8.8 shows such a comparison.

When the comparison between external and internal data is made, the assumption is that the comparison is made on a common key. Any other assumption and the comparison between external and internal data loses much of its usefulness. Unfortunately, actually achieving a common-key basis between external and internal data is not easy.

To understand the difficulty, consider two cases. In one case, the commodity being sold is a large, expensive item, such as a car or a television set. For a meaningful comparison, sale by actual outlet needs to be measured. The actual

industry sales (in billions)

corporate sales (in millions)

sales by dealer is the basis for comparison. Unfortunately, the key structure used for dealers by the external source of data is not the same key structure used by internal systems. Either the external source must be converted to the key structure of the internal source or vice versa. Such a conversion is a nontrivial task.

Now consider the measurement of sales of a high-volume, low-cost item such as colas. The internal sales figures of the company reflect the sale of colas. But the external sales data has mixed the sales of colas with the sales of other beverages such as beer. Making a comparison between the two types of sales data will lead to some very misleading conclusions. For a meaningful comparison, there needs to be a “cleansing” of the external sales data to include only colas. In fact, if at all possible, colas only of the variety produced and sold by the bottler should be included. Not only should beer be removed from the external sales data, but noncompeting cola types should be removed as well.

Скачать в pdf «Building the Data Warehouse»