Datamarts for reporting

The team today released our "Version 1" of our reporting datamart.


What is a data mart? (I hear you cry)
The principle is that rather than having reports for management running of our production analysis platform, we create a data set that is bespokely created to answer our managements reporting needs.


In short, a datamart is an abstract view of your database, with aggregates and summarisations that are required for repetative tasks, already built into its design.


Why you would need a data mart 
Benefits:
1. Fast response times to often complex reports
2. Quicker time to implement new reports by expanding an already existing model
3. Complete control of the business rules from a central location
4. One system to optimise for reporting purposes
5. Decreased impact of reports on your production environment


Slight drawbacks 
1. If pre-aggregated data is needed at multiple aggregation levels, all these levels would have to exist exist in the data mart
2. The data mart is only as good as the thought put into it


Considerations for design 
1. What is the purpose of the data mart? - How "Little" data does it require (be brutal)
2. The reports that it will generate, what data will it need?
3. How timely are these reports needing to be - this will feed into how often it will have to be produced
4. Is a cube is SSAS more suited? (How often does your data model change)
5. What technologies will sit on top of this platform? SSRS, qlickview, etc?


I realise this is not very detailed, but its a starter for 10.


Details will come in the next few days of what it looks like, and how it has been put to use...


If you want to talk about it in more detail:
You can tweet me at @tristix or email me at tristix@gmail.com