topic Best practice - Dimensional modeling in QlikView

Best practice - Dimensional modeling

Wed, 07 Oct 2009 20:12:05 GMT

I got the attached relational database (only a part of the big database) with several circular references. My client wants me to do a demonstration of what Qlikview can do with his data so I don't have any real limitations yet. However I don't want to limit my possibilities for the future and therefore I would like to keep all possible relations.

What would be the best way to do the Qlikview structure when loading the data? How do I get a good star schema? What tables should I join? What else should I do?
TblUser, tblCompany and tblBranch are the key values for the client but tblCountry and tblAdresss is also of interest.

Any suggestions?

Best practice - Dimensional modeling

Wed, 07 Oct 2009 21:22:08 GMT

Are those all of your tables? All of those look like dimension tables and they seemed to be joined alright.

Star schemas are based around a fact table, which then links off to your dimension tables. What are the facts in your data model? Do you have sales or revenue or what are you measuring?

Your fact table can then contain foreign keys to commonly linked dimension tables. I'd imagine your fact table would contain User_Code (or User_Id) and maybe Company_Code. It looks like a few of those other dimension tables come from User or Company data.

The key with QlikView is to get your data loaded and linked and then start working with it. If you're just doing a software demo, then you probably don't need to worry about reorganizing your data. QlikView is going to be able to handle it as is. Once you have a better idea of how you plan to use the data and what information is important, that may help you in designing your data set.

Best practice - Dimensional modeling

Thu, 08 Oct 2009 16:44:13 GMT

I like your discussion in the article. As always there is a question of "If", "but" and "depending on" when selecting the right solution. It's just a matter of finding the right solution in each individual case. Thanks for charing you knowledge.
/Odd