[vortep]

I left my tenure at Google with this same impression. Data is constantly being copied, denormalized, cached and transformed. It is true that efforts are made to leave an audit trail that could in theory be followed from initial source data to all derivatives. However, this is practically infeasible and certainly not doable with a single engineer’s worth of knowledge about the system.

[TillE]

It must be trivial to find where a copy of data lives. That just needs basic tools to trace a request. I can understand that finding every copy of some data is non-trivial, though.

[vortep]

Good point. I was thinking of the task of deleting all of a user's data upon request.