Shared datasets and dataflows - need advice

[Anonymous]

Hello Community  -  Our company is getting more and more on board with Power Bi and looking to roll it out across the company.   We are in the process of setting up our Power Bi  Apps  navigation structure and I have some questions that probably apply to lots of folks out there in a similar situation.  

Over the last year, I have created a number of reports in use by the company and as such, have about three data models that are highly customized  (custom measures, custom columns, or custom power query transformations in the pbix file).   For example, one of those data models is for a Logistics report, another is specifc to a Sales and CRM report.   The datasets for each of these are fueled by common data flows set up by our sysadmin.   

Our sysadmin also has a created his own dataset that he views as the "golden" dataset.   The problem is that it lacks many of the customizations of the data models I have built.   It also includes a huge amount of dataflows (nearly 60) many of which are irrelevant to the reports I have created.   Accounting ledger information for example which has no relevance to our sales team (and thus no need to bring it in to the report).  As such, I do not want to use his dataset for my reports, not to mention the challenge involved in converting all of my reports to use his dataset.  

The other challenge is that I make updates to my reports quite frequently - small changes aesthetically or requests come in to add a new visual etc.    Or if Power Bi adds a new feature that I want to incorporate, this also requires me to republish any report I have made changes to.   

With that background, here are some questions I have:  

1.  Does it make sense to set up the App structure to service particular groups?    In other words, create a specific workspace for the Accounting and Finance team, another workspace (App) for sales and marketing, another for Operations, etc  (and note, some of the reports specific to those teams may also need to be shared cross-departmentally).   

2.  Per the above, would it make sense to have unique datasets that service the specific reporting needs of each group?   I know ideally you want to limit this, but we have a high need for customization (especially the use of calculated columns) in our reporting.   So perhaps using composite models is the way to go?  

We are really just beginning our journey on how best to set this up in terms of user navigation, row level security, data model proliferation, etc.    Thanks!

[v-deddai1-msft · Accepted solution]

Hi @Anonymous ,

>>The other challenge is that I make updates to my reports quite frequently - small changes aesthetically or requests come in to add a new visual etc.    Or if Power Bi adds a new feature that I want to incorporate, this also requires me to republish any report I have made changes to.   

If you just need to add some visual or change the layout of visuals , you can just edit report in power bi service. But if you need more designs like creating new calculated columns, measures even use  the new feature, you need to do it in power bi desktop and re-publish again.

>>1.  Does it make sense to set up the App structure to service particular groups?    In other words, create a specific workspace for the Accounting and Finance team, another workspace (App) for sales and marketing, another for Operations, etc  (and note, some of the reports specific to those teams may also need to be shared cross-departmentally).   

You can creat workspace  for each teams. you can create different datasets for each workspace. And you can share reports in your organizations or external users.

>>Per the above, would it make sense to have unique datasets that service the specific reporting needs of each group?   I know ideally you want to limit this, but we have a high need for customization (especially the use of calculated columns) in our reporting.   So perhaps using composite models is the way to go?  

You can create different datasets for different teams  to limit them using data belongs to them. RLS may not suit in the case. It only works for viewer, and users with build permission of dataset will see all the data in the dataset.

If this post helps, then please consider Accept it as the solution to help the other members find it more quickly.

Best Regards,

Dedmon Dai