Usually it’s a case where a data provider gives you data that is necessary for a business but _also_ includes data you don’t want or need.

Some process will ingest and index the valuable data and throw the original file into non-indexed storage in case it needs to be re-run.

Ideally you filter that data at ingest time but even that is building a capability that feels against the intent.

But let’s say you didn’t even know to build the filter. Now you need to go in and scan all those files to remove the personal data. A capability that requires you to read data you hadn’t prior.