I’m exploring best practices in designing data pipelines and want to understand how different teams handle computationally intensive transformations. Some advocate for doing it early during ETL to keep models clean and fast, while others prefer flexibility and defer transformations to the modelling stage. Curious to hear what’s worked for others and why.

I’m exploring best practices in designing data pipelines and want to understand how different teams handle computationally intensive transformations. Some advocate for doing it early during ETL to keep models clean and fast, while others prefer flexibility and defer transformations to the modelling stage. Curious to hear what’s worked for others and why.