Data Mart Models Data models enterprise data models data warehouse models industry data models generic data models best practice data models Erwin Allfusion data models HP Oracle EDS

Data marts represent both a significant business opportunity and potential liability depending upon their architecture and implementation.

When integrated with the overall goals of the organization consistent with the data architecture and the data warehouse, data marts transfer the reporting responsibility to the business units, which have are most responsive to their own requirements.

Data marts built in stand-alone fashion external to the organization's data architecture and data warehouse planning represent another source of data that is unplanned, needs to reconciled with other corporate data and requires additional effort.

There is every reason to avoid another set of inconsistent data in terms of costs ("70% of system costs are maintenance"), redundancy and overall inefficiency.

The easiest way to manage data marts is to design them to be consistent with the overall data architecture and so that data that can be extracted from the data warehouse into the data mart(s). In that way the reporting and manipulation of the data is transferred to those having the most need and the source of clean, dependable data remains the data warehouse and available to all.

ADRM Data Mart models are constructed from the ADRM Subject Area Models, which themselves are derived from the Enterprise Model to insure consistent, integrated data across each data mart within an overall enterprise data architecture.

The ADRM Data Warehouse provides a blueprint of the source data that will be used to load the data marts. The data mart models provide a blueprint of the data that will be held in those data marts. It's a mechanism to keep things in synch, identify any variances and differentiate efforts and responsibilities.

Of course, there are several strategies depending upon the actual Configuration of applications, data warehouse and data marts in operation. Data Mart models can be used independently, integrated with other models or linked together to construct a full data warehouse in either star schema or 3NF format.

Where star schema data marts are utilized, ADRM Business Area and Data Warehouse models are used to design the facts and more importantly the dimensions. The dimensions of the each star must be consistently designed and applied in order for the stars to be conforming or able to be used for common access of data. Stars are deceptively easy to build but there is little real value to the organization in building stars whose data cannot be shared due to poor design.

A top-down design derived from an overall data architecture of enterprise, business area and data warehouse models will pay significant dividends and result in a well-managed IT infrastructure.