Data Architect … Data Analyst … Data Engineer…
In addition to the expected comparison of dataflow and dataset definitions in the context of Power BI, Reza gave a surprisingly in-depth overview of different architecture models, given the limited time available.
On the face of it, Power BI dataflows and datasets may seem interchangeable: they both involve getting and manipulating data, an understanding of your data structures are needed, and they both require coding. However, in the same way that M & DAX are not interchangeable — yes, DAX can be used to transform your data once its been loaded into the model, but really don’t do this — dataflows and data sets do different things within the data architecture.
The dataflow forms the ETL layer and feeds data into the data set — this is where data transformation and Power Query M is needed. The dataset is the modelling layer and informs the visualizations — this is where data modelling and DAX comes in. Dataflows and datasets need not be used in conjunction with one another, each is a solution to a different problem: dataflows resolve the issue where multiple versions of the same table are used in different .pbix files; datasets solve the problem of having multiple versions of the same model (DAX) in different .pbix files.
…Reporting and Formatting Expert
This presentation made it apparent that for larger scale environments and/or projects there are multiple Power BI sub-specialisms. A lone developer can see there is a clear division between data connection and transformation (Power Query M), and and data modeling and the visualisation itself (DAX and meaningful graphics).
In fact, there is a further division to be made between data modelling and report creation — the success of the latter requires firm foundations and clear technical understanding of data model. New updates coming to Power BI mean that the report creation aspect can be handed over to less technical super-users who may not fully understand the model but can utilise it effectively.
By using dataflows and datasets, data preparation, modelling and visualisation can be decoupled into a tripartite structure — in this way Power BI can be used as the back-end to provide data to visualisations in other reporting tools, such as Tableau, Excel, or other platforms with a live connection.
It also becomes even more apparent that as the Power BI offering continues to evolve, and as more businesses turn to this technology to give them insights into their data, the range of professional roles within the Power BI world will increase.
Featured Image Source: Dolls Hats, pinned by Maka R – Toronto (Pinterest)