I was at Qlik Qonnections last week and attended hic presentation on under the hood, which was a fantastic presentation.
It got me thinking about the Qlik data model design, my question is:
'If it really necessary to design a dimensional model and load this data into Qlik? If I was to add the dimensions to the fact table and the resultant table was not too wide would this be a more efficient data model? How wide is considered too wide?
The star schema dimensional model provides a nice balance of RAM usage and runtime performance. A single flattened table or even an extreme snowflake will produce the correct results. For typical small to medium apps any of the models will provide acceptable performance and resource usage.
In some cases a single table can provide better response time, but at the cost of significantly more RAM. It's usually not worth the tradeoff in my experience.