Forum Discussion
Datasets - one large or many small
I am trying to figure out the philosophy regarding datasets: should I create ONE big, general dataset or MANY small datasets?
The thing is I have 6 different departments and first idea was to share to their respective Microsoft Teams -workspace. But that means if I update one report, it has to be publishes to manu workspaces. But if I have one general for all, there will be difficulties with bi-directional relationships and also potential sales data would be accesses by other departments.
What is the best way to think about this? Also updating measures and everything means those has to be done to all datasets.
So 2 questions:
1. One big or many small datasets?
2. How to distribute correctly to many departments?
Thanks alot! This is really driving me nuts!
9 Replies
It may be a combination. You could have one big dataset with most or perhaps all of the data, and then share that dataset to the otehr departments. You can then restrict what they can and cannot do/see. So your primary data set can have Sales, but you don't expose Sales to the Engineering team for example. You need to have the new Gen2 workspaces to do this, so if you have Gen1, click the little upgrade button in the workspace settings.
There is no perfect solution to your issue - it is all about optimization, not perfection. Share Datasets has been a HUGE boon for me in this processes, but there are still times I completely segregate data and datasets. That, and I've not pulled the trigger on upgrading all of our workspaces to the new Gen2 type yet.
Yes, several departments uses the same kind of data.
However, RLS feels a little bit weird. When people open up the reports and just see blanks on one page?
But I think I will go back from multiple datasets to two datasets. Just a mess to keep track on measures and changes in measures between datasets.
Thank you so much to both of you!
Hi,
Different subject areas and data that is not related should go into different data models. But if several departments uses the same kind of data, let them share a large model with data from all departments. Then add roles to the model and use row level security to let the users see only the parts od the data they are allowed to see. You will then only need to deploy the report to one workspace/app and users from different departments can use it and see different numbers.
To avoid problems with relationships, use some structured design strategies when creating the data model. Dimensional Modeling is a VERY good modeling teqnique for Power BI.
Good Luck!
