Heavy Transformation from Flat Files - Better in Dataflow or Dataset?

[PowerUser123]

I have a lot of Excel, CSV, and text files stored in SharePoint which is my model's main data source. I perform heavy transformations on the data. In regards to performance, is it better to place all that transformation at the dataflow level and then simply import the dataflow into a dataset with little to none transformation happening there?

Or would it better to import the files with essentially no transformation at the dataflow level and then perform all of the transformation at the dataset level?

Or am I overthinking this and there is no real difference? 

Editing some data models directly hooked up to SharePoint can take forever when editing. Editing the same data when this data comes from a dataflow is seemless and easy. 

[arvindsingh802]

Dataflow is best place to do all your transformation.

If your files are big in size you can explore connecting as web, it will be lot faster
Quick Tips: Connecting to Excel Files Stored in SharePoint Online from Power BI Desktop (biinsight.c... 

Also you can do all your work in power BI desktop intially and then copy the M Code and paste in Dataflow

If this post helps, then please consider Accept it as the solution, Appreciate your Kudos!