I would not agree with the statement that a Star Schema is always best for Qlik. It depends on how you want the model to behave and what problem you are trying to solve.
In a typical database system, storing the Country field in a separate table would be most efficient for storage as it would eliminate redundant entries. Qlik stores data differently, so this redundancy issue is lessened or non-existent in some cases.
A star schema generally provides the best balance of storage and CPU usage at runtime. However, I would argue that in many cases, it's not worth spending time getting to a star.
My first rule is code for clarity and simplicity. Make the model easy to understand and use, and minimize the time you spend on creating and maintaining the script.
Considering your example. The separate Customer and Company tables are likely the way the data exists in your source system. Keeping it the same is easy to write the script and understand the model. What if I want a count of countries that purchased each product? Count(Country); easy. If I join Company to Customer I have to change to Count(DISTINCT Country); more complex.
We may be able to say that reducing the number of "hops" in a data model by going from snowflake to star would improve chart calc time by about 10%. If a chart now takes 50 milliseconds to calc and we reduce it to 45, is that important? Was it worth the effort? If I have 50 daily users with 30 clicks each, it it probably not significant.
-Rob
http://masterssummit.com
http://qlikviewcookbook.com
http://www.easyqlik.com
