TodoBI - Business Intelligence, Big Data, ML y AI TodoBI - Business Intelligence, Big Data, ML y AI

Volviendo a las bases, Kimball´s What Not to Do

Kimball´s DW
Muchas veces nos perdemos en el día a día, en las novedades que van sacando los fabricantes, en los nuevos desarrollos, etc... y perdemos el foco en lo 'más importante'. En los buenos y críticos factores para construir un DW.
Recupero una entrada del 2004 de Kimball que nos da unas claves sobre los principales errores a la hora de construir un Data Warehouse. Imprescindible reelerlo!!
Mistake 12: Place text attributes in a fact table if you mean to use them as the basis of constraining and grouping.
Mistake 11: Limit the use of verbose descriptive attributes in dimensions to save space.
Mistake 10: Split hierarchies and hierarchy levels into multiple dimensions.
Mistake 9: Delay dealing with a slowly changing dimension (SCD).
Mistake 8: Use smart keys to join a dimension table to a fact table.
Mistake 7: Add dimensions to a fact table before declaring its grain.
Mistake 6: Declare that a dimensional model is “based on a specific report.”
Mistake 5: Mix facts of differing grain in the same fact table.
Mistake 4: Leave lowest-level atomic data in E/R format.
Mistake 3: Eschew aggregate fact tables and shrunken dimension tables when faced with query performance concerns
Mistake 2: Fail to conform facts across separate fact tables.
AND THE BIGGEST MISTAKE …
Mistake 1: Fail to conform dimensions across separate fact tables.