Merging and Calculated Columns in Power Query vs Calculated Columns

[newpbiuser01]

Hello,

I have a general question about creating calculated columns and merging tables in power query vs. dax. I have two tables - one with the sales data and another with payment information and need to merge a couple of columns from the payment to the sales table. Is it generally better to merge the tables in power query for faster performance (for tables with over 9 million rows) or do it in dax by creating a relationship between the tables? How about creating calculated columns (that need to be calculated at the row level and not aggregated so can't use a measure)?

I was under the impression that it was better to do it in power query so once the model was loaded, the report loading time and interaction time would be shorter. But the more I look into it, it sounds like dax would be the way to go? I'd really appreciate any insight! 

Thank you!

[v-henryk-mstf]

Hi @newpbiuser01 ,

Whether the advice given by @Stachu  has solved your confusion, if the problem has been solved you can mark the reply for the standard answer to help the other members find it more quickly. If not, please point it out.

Looking forward to your feedback.

Best Regards,
Henry

[Stachu · Accepted solution]

Star schema is the prefered approach for tabular models (i.e. DAX, so tables with relationships). That said I would recommend doing as much of the data preparation as possible in PowerQuery (with the cost of the model refresh being slightly longer), and avoid calculated columns at all (they come with a cost of using additional memory, and are not really necessary for the row level calculations - that's what the iterator functions like SUMX are for https://dax.guide/sumx/)

A very good article describing the differences in more detail:
https://www.sqlbi.com/articles/comparing-dax-calculated-columns-with-power-query-computed-columns/