I finally got access to Granola's automation suite last week. Excited to streamline my data prep, I enabled every single automation rule I could find—data validation, auto-joins, outlier correction, the works.
Big mistake. My clean sales dataset spawned hundreds of new "enriched" tables overnight. It auto-joined to unrelated marketing schemas based on similarly named columns, replaced all my NULLs with the column average (even for IDs!), and created so many derived columns I couldn't trace the logic. Spent the whole week just untangling it and figuring out what changed. The power is scary. Anyone have a strategy for rolling these out slowly? Maybe a testing sandbox?
Oof, I've been there with Granola. That auto-join on similar column names is a particular trap - it once linked my customer "region" codes to a legacy "region" table of office locations. Total chaos.
My strategy now is to treat it like a new team member. Enable ONE automation at a time, on a single dataset copy, and check its "work" before moving on. I also make heavy use of the "exclude columns" list in the rule settings to block things like ID fields from averaging.
It's a slower start, but you learn what each rule actually does. A proper sandbox feature would be a lifesaver though. Did you find a way to revert everything, or was it manual cleanup?