Are the Kimball and Inmon Data Warehouse models dead? Some industry advisors are saying they are no longer relevant in today’s work. I think they still have a place.
In the drive to provide better business value from Data Analytics projects and meet the changing needs and tools the business have (Power pivot for example) the delivery of a solitary DataMart is certainly now changed, both in the way it can be built and how data is being provisioned.
How data is provisioned for the bsuienss is changing and current Analytics architectures provide for a Data Enablement platform that contain DataMarts but more importantly several other sets of data also including:
- Sandboxes – Data sets where business users can stand-up small sets of data form business critical questions where the time to deliver a DataMart is simply too long
- Sets of data that users can bring into tools like Microsoft PowerPoint to further enhance thinking and analysis. Think of exchange rates and CPI type data.
- Hadoop stores for less structured data and large volumes.
- Traditional DataMarts to provide trusted, audited, quality, cleansed data that can be relied upon and used for the businesses core reporting
- Sets of data for statistical analysis as traditional DataMarts do not support this effectively.
The way data is feed into the DataMart is also up for grabs. Inmon style EDW or CIF architectures are slow to build and evolve, Kimball stage schemas cannot manage massive volumes/complexity meaning some organisations are using innovative methods to get the data ready to then be fed into both the Data Mart and other sets of data in the platform. http://merolimited.wpengine.com/blog/data-enablement-platforms/Value pair based databases and even Hadoop style stores can provide support for massive volumes and rapidly changing environments in the place of the Kimball or Inmon architectures. Really the approach is about thinking of innovative methods to meet whatever business requirements’ you have, rather than forcing the business requirements into a predefined architecture.
So the DataMart is clearly not dead and has a valuable place in any Data Enablement platform as the core trusted set of data, however is this rapidly changing and value driven world it must now share the limelight with other sets of data that are more targeted towards different business needs.
Having recently delivered a platform of this nature I personally I think the idea of a data enablement platform is a great step forward.