The opening step is translating what you want into something countable. Fields are filled only where a source supports them. Delivery includes a plain statement of what the file will not do.
Format handover
Delivery in the structure your systems already accept, tested on a sample so nothing needs reworking afterwards.
Repeat control
Removing duplicates within the file and against anything previously supplied to you.
Segment splitting
Dividing a universe into parts that can be approached differently, with counts given per part.
Ask what the file will not do; that answer is the useful one.