August 13, 2018 at 7:07 am
I have 2 dimensions - 1 measure group.
Dim1 : Customer
Dim2 : Group
The grain of the measure group is customer and there is a 1-m relationship between Group and Customer. The measure table has GroupKey and CustomerKey.
In order to keep the aggregations working I think I need a referenced dimension in this case. i.e. to group --> customer -> measure table.
I understand reference dimensions result in slow queries and materialising them does increase performance without too much storage.
Group is a tiny dim and customer is a large dim.
This is making me think referenced dim is the way to go.
Is my thinking correct?
August 13, 2018 at 9:48 pm
Customer dims get big, you need something to roll-up customers before showing as an attribute, no-one likes 20,000 columns in a dashboard or report.
I look for things to group by, like Geography, maybe Country, State, or Region, or demographic data like Male/Female, Age-bucket, etc.
Sounds like you have a good natural candidate with Group, I would add GroupKey to your CustomerDim, and when you present your CustomerDim, expose GroupName as a top-level attribute (hierarchy), then allow drill-through to State maybe, and lastly individual Customers.
Hope that helps,
ChrisBecker.blog
Chris Becker bcsdata.net
Viewing 2 posts - 1 through 1 (of 1 total)
You must be logged in to reply to this topic. Login to reply