|
Dataface, Inc. |
|
Software and services without fluff or bit-twiddling! |
|
Dataface, Inc. |
|
Software and services without fluff or bit-twiddling! |
|
Data Warehousing and Business Intelligence—Our Advice |
|
Management A senior executive acting as Business Sponsor must guide the project so that it does not degenerate into I.T. speculation about what the business needs. A Steering Committee should be formed to include all the principal potential beneficiaries. There must be a clear purpose for the project and good examples of the value it is supposed to provide to the business. The project must show some results of value within 3-6 months or it will drift and lose support.
The Planning Process Requirements gathering, by interviewing business managers and if appropriate, organizing facilitated sessions, to identify areas of ignorance/uncertainty or missed opportunities. Documenting and prioritizing the needs expressed. Dividing the project into iterations of limited scope and identifying the payback of successive iterations. Identifying the location and accessibility of existing information, which may extend beyond operational systems to spreadsheets, other desktop files, and even shop-floor instrumentation. Building ETL processes consumes the majority of the technical resources and time. Plan must provide for training business users and allow time to transition them to the data warehouse.
Why a data warehouse has to be a dedicated system Data is non-volatile and saved for longer periods than in operational systems Data model fits business structure and integrates data from all operational systems
Critical Success Factors ETL tool should provide success audits, performance logs, and auto-restart capabilities ETL must be metadata driven and support simple and complex transformations Identify data quality issues and need for any specialized data cleansing tool Set conventions that help simplify data warehouse implementation. support and use Database design should follow a de-normalized, star schema/snowflake design standard Capture data state relationships in historical data Use standard business terms as database attribute names Transform natural keys into surrogate keys Implement hierarchies that support the organizational relationships of your data Capture dimension and hierarchy change history Use intelligent default values for missing or corrupt data Build summary views of the detail data for common queries Capture data at its most atomic level Expose business rules applied to summarized data Support drill down from summarized views into detail data Be prepared to provide standard reports and queries and nurture users Use enterprise or federated data warehouse design Support generation of “specialized” data marts tuned to department/functional needs Be prepared to support data warehouse access by operational systems |
