Sharing and Reusing Datasets

[Anonymous]

Hi,

My team is looking for a solution, but getting a bit lost in all the options and terminology.

Our main data source is an Oracle database which we pull from to Power BI Desktop. When we publish our reports they go to a Workspace with an on-premesis gateway. Lately we find we have too many different datasets trying to refresh at once. But many of them require similar or the same data, transformed in different ways. 

Is there a more economical way to, say, grab one big dataset and use it in different ways among other reports? We would require the ability to "Transform Data" from this communal dataset.

[parry2k · Accepted solution]

@Anonymous you cannot make changes to the Power BI dataset you are using either Live or Direct Query. Any change required to tge underline dataset, you have to do at the original Power BI dataset pbix file.

[parry2k]

@Anonymous yes it is a common practice and somewhat best practice too. You can create a golden dataset, and publish it to the service, for visualization. Connect Power BI desktop with the published dataset, and visualize the data, it is generally called a thin report because it doesn't contain data, just a reporting layer. 

[Anonymous]

Hi, thanks for your response. 
I have never heard of a golden dataset before. Searching for it brought me to this article which sounds like what I am looking for, however in practice I am having issues.

https://www.withum.com/resources/power-bi-data-model-governance-with-power-bi-golden-dataset-and-thi....

First of all, adding another dataset that I created it gets forced to be DirectQuery (to the dataset itself? not entirely clear).

Ok, that's fine I suppose. But when I go to Transform Data, Power Query shows that I have 0 Queries despite having just added this dataset (which is published in Workspace, updated with on-premesis gateway)

Is there some way to do this AND allow us to Transform Data? Or does that require a new dataset, pre-transformed? My concern is we want to reduce the number of datasets refreshing, and we may not always know ahead of time what transformations are required until we get asked to create a new report.