Why keeping transformations centralized in a dbt monorepo is ideal (even for data science)

[gwenwindflower]

Especially for data science!

In my time consulting I've had many people ask me about when and how to split up their dbt repo. The answer is never! :)

Okay, I jest a little bit, there are certainly circumstances that call for splitting up your dbt project into multiple codebases, but the reasons for doing this are generally constraints that force you to do so, not advantages that can be gained. dbt is about defining a single source of truth, which naturally becomes more difficult when you have multiple sources.