If a requirement can be specified it can be sized, and sizing happens first. Everything removed is logged, so the removals can be inspected. Delivery includes a plain statement of what the file will not do.
Change-only refresh
On a repeat build, reporting what actually moved instead of re-delivering the whole file.
Record identifiers
Giving every row a stable key so a later refresh can be matched against this one.
Intake review
Reading an existing file and reporting what is wrong with it before any work is proposed.
Describe the exclusions as well as the targets; both shape what is possible.