Using Qlikview to Build Star

[#1]

I've been asked to explain why or why not and how one might build a star topology of data in QVD sets without the use of an ETL tool such as Expressor, Data Stage or Informatica.  Question is, has anyone attempted such an endeavor? If so, how well did it work out in terms of maintenance ease, extensibility, performance and data quality?

[johnw · Accepted solution]

Probably half the QlikView data models out there are star schema built without using ETL tools, and the majority of the rest are probably snowflake schema.  As for why we use star schema, it's hard to be clear why we do.  QlikView can process any number of data structures, and they definitely don't have to be star or snowflake schema.  But the forum consensus seems to be that star and snowflake schemas just work best, are the most QlikView-friendly.  The required denormalization isn't an issue because it's only a reporting tool so all synchronization works one way, and because QlikView's data compression keeps the additional data from actually taking additional space.  The model itself is then pretty clean to look at, and pretty safe from loops, which are something that must be avoided in QlikView.

[#3]

John,

Thank you for your reply.   I dropped the investigation last October, and began it again recently.  I'd like to see how the half Qlikview data models star schema based were built without a robust ETL tool.  Are there examples available somewhere?

The issues uncovered thus far concerning Qlikview itself is its inability to manage type 2 dimensions with surrogate keys.  This is not a criticism, but rather a statement of the Qlikview technology focus as it appears - perhaps correctly so - to be focused on the associative capabilities it does very well in presentation analytics.

So now I am investigating Qlikview Expressor3.9, and I am finding similar issues with it.  At present I don't see Expressor in competition with ETL tools such as DataStage or Informatica, but rather another tool for managing and extending Qlikview datasets initially created from a star topology created by other ETL technologies.

What I find missing is a knowledge/experience base on using Qlikview Expressor to accomplish a managable and extendable star topology.

Any information links appreciated that may help appropriately position and deploy the combination of these tools - DataStage, Qlikview Expressor, and Qlikview itself.

Tom