A blank quantity is not a zero. This fictional stationery CSV shows a small cleanup that preserves missing information.
Agree the transformations
First retain the source. For this example, trim leading/trailing spaces and lowercase item labels because the fictional owner has agreed that “Notebook” and “notebook” identify the same item. Case changes are not universally safe: product codes can be case-sensitive.
Remove only the duplicates you can explain
After the agreed label normalization, two notebook rows have the same quantity and unit price. This sample removes one exact duplicate. In a real transaction export, identical rows can represent separate sales; require an owner-approved rule before removing them.
Flag missing values
The folder quantity stays blank. The output adds a review note rather than guessing a count or calculating an unsupported total.
Original fictional CSV · Cleaned fictional CSV