Claus Madsen

Business Intelligence Consultant and Data Platform Architect. Passionate about data and storytelling, bike riding and rum tasting.

[Claus Madsen @clausm73]

@GBrueckl @bob_duffy In databases we have schemas for this. Even schemaless items have schema (dbo). We can assign ownership to this.
In Fabric the workspace should be considered the default schema and workspace admins its owners.
I just don’t get that each object is owned by the developer?

[Claus Madsen @clausm73]

@chandeep2786 You made a dimension table. A bridge table is when you need to traverse between dimensions (sometimes referred to as factless fact table).

[Claus Madsen @clausm73]

@MMarie Yes! Please please PLEASE. Voted!

[Claus Madsen @clausm73]

@GBrueckl @mim_djo Interesting thought: Would the whole idea of a SaaS platform not be that it chooses the most effective processing engine by itself based on workload / payload?

[Claus Madsen @clausm73]

@wynhopkins Voted!

[Claus Madsen @clausm73]

@PowerBIguy Exactly. Making reports doesn’t mean you have the proper business context for making them as the request often is “just build something similar to this old report”. No use case, no business requirement.
Generative AI could probably ”fill in the blanks” for such metadata.

[Claus Madsen @clausm73]

@PowerBIguy That would only lead to garbage values. Instead implement some kind of review process of metadata, fx using Purview and Delegate responsibility.

[Claus Madsen @clausm73]

@AgulloBernat I totally agree and have consistently names mine “Extract”, “Transform” and “Load” (ETL). Often with some additional folders like “Parameters”, “Functions”, “Tests”.

[Claus Madsen @clausm73]

@VicVijayakumar Seems like recruiters or other resume readers need to learn how to apply AI agents on top of resumés if they don’t fancy creative layouts.

[Claus Madsen @clausm73]

@KratosBi @cannydata Calculated columns are a way to move certain calculations from being done at query time to processing time. I have seen cases where in-memory processing power and time of a calculated column is vastly superior to having Power Query or even a SQL engine trying to do the same.

[Claus Madsen @clausm73]

@Ko_Ver @JohannesVink @MrAndyCutler @denglishbi Currency should definitely have its own dimension to properly handle currency conversion in calculation groups across fact tables with different grain (transaction, budget, forecast, etc.)

[Claus Madsen @clausm73]

@Ko_Ver @MrAndyCutler @denglishbi I know, but if I have a single column in a fact table changing values for “almost” every row in the table I won’t bother making a junk dimension -> I leave it in fact as a degenerate dimension or let’s call it “drillthrough detail column” 😉

[Claus Madsen @clausm73]

@kurtbuhler @MrAndyCutler @Ko_Ver @denglishbi A degenerate dimension?

https://t.co/HN2PD3gvZr

[Claus Madsen @clausm73]

@tommartens68 Well conceptually you still need the data warehouse. But yes you can make it as a data lakehouse, but you still need to prep and structure your data (a dedicated copy with targeted purpose) as with the data warehouse.

[Claus Madsen @clausm73]

@GilbertQue So if I somehow figure out the recipient address I can increase your Fabric cost? 😉

[Claus Madsen @clausm73]

@aiolicashout @matsonj No need to install Python. You can use it as a cloud service depending on Excel version. https://t.co/0sUhWPuAs3

[Claus Madsen @clausm73]

@GBrueckl @KratosBi Awesome!

[Claus Madsen @clausm73]

@KerryKolosko There is always a trade off. Agility is not always preferred if you have different use cases and user personas for the data. Very few have the DAX gifts needed to pull data properly from semantic model for ML work (even with semantic link) -> most data prep needs to be upstream.

[Claus Madsen @clausm73]

@CJMajka Parameters for budgeting and forecast scenarios - maybe even writeback features. AI for commenting numbers as financial reporting is highly standardized and here you can actually get very decent insights if accounts are properly named.

[Claus Madsen @clausm73]

@MrAndyCutler Why not just name it CoPAIlot and be done with it 😁