Architecture advices for new Power Bi reports

[MattRGX]

Hello everyone,

I have to set up a new report available to the clients of the company I work for. This will probably be a report embedded within a web application.

Today, the entities within the web application are structured as follows:

• Client, which is the highest level
• Agency, which can be multiple within a single client
• Employee, which can be multiple within an agency and will be the end-users of the report.

Regarding data confidentiality, I need to have a fairly fine granularity. A client cannot see another client's data. Within the same client, an agency cannot see another agency's data.

The report I have to produce will be the same for all clients (same visuals etc...), with no special cases. Only the data within this report will be specific to each client or agency.

My main question is whether I should create a unique report for each client (knowing that initially there won't be many clients, but over time, the number of clients will increase). For example, there could be a report with a main dataset containing all the measures, tables, etc., and multiple reports that are all identical but specific to each client (in terms of data) and sourced from the main report dataset (which will contain data of all clients)

Or, my other solution would be to build a single, unified report and implement RLS to filter the display of data based on the clients/agencies/employees using the report.

I think the first method would be more complicated to maintain, and I lean more towards the second solution. But I don't know if RLS can provide a fine enough level to achieve my data confidentiality goal.

Thanks in advance for your help

[lbendlin · Accepted solution]

A properly implemented RLS will provide you with the data security.  It needs to be combined with your outer authentication tool in the "App owns data" embedded scenario.

Separate semantic models may mean more effort but they give you flexibility once your customers start to ask for customizations.

[MattRGX]

Thanks a lot for your answer. We will not do any customizations for any customers.

I've seen also that it is possible to have multiple identical reports (and therefore datasets) for each customer but when we do modification to only deploy one report that will spread the modifications to all other reports. May be with Pipelines ?

Is it really possible to do that ?

thanks in advance for your help