The Informed Company

Chapter SixteenData Mart Implementation

Data warehouses have numerous schemas that can become difficult to navigate. One department alone such as marketing may have tens if not over a hundred tables; combined with all the tables for sales, finance, HR, production, and customer support, it can get overwhelming. Moving from a data warehouse to data marts makes it easier for users to find the data they need for their business process or use case.

Data marts can be created in three steps.

Views on the Data Warehouse

Create new views by selecting the relevant data from the data warehouse (Figure 16.1) to build out each data mart:

Figure 16.1

CREATE VIEW mart_view
SELECT
  * 
FROM 
  DataWarehouse.View

Most if not all of the cleaning should have occurred when going from the lake to the warehouse; if there is a cleanliness issue, address it with modeling in the warehouse stage.

Additional modeling to create aggregations for performance reasons is fine, but remember it does limit analysis. Also, combining data to make it even easier to analyze is common in marts. For combining data we still recommend using the wide table approach.

Segment Tables

Determine how to split the data into different data marts.

Common organizational ways include:

• Department
• Product line
• Use case
• Access level
• Security considerations

Examples of splitting by business process:

• Lead to cash
• Product analytics ...