I'd like some peoples views on the ascertain that synthetic keys should be left in your model when generated from interval match actions.
I've developed Qlikview for a number of years and had it drilled into me that synthetic keys cause performance issues and should always be removed.
However, recently I've heard how these should be left in your model (allegedly) as they are optimal. This feels completely alien to me, particularly as it is a simple task to take them out via an additional step in the load.
Can anyone explain to me the logic of the 'optimalisation' for leaving these synthetic keys in a model please?
I don't think there is anything special about the synthetic key with regard to the creation by the IntervalMatch() function, so you might be interested in some general discussions about synthetic key performance.
