Header matching is exact
Product ID and product id are different headers, and a trailing space also matters. If a requested column is not found, it is ignored. Copying the header directly from the first CSV row is the quickest way to avoid invisible spelling differences.
Repeating a header does not duplicate the column
If the order field says id,name,id, the tool uses that source column once. This tool rearranges existing columns; it is not a column-cloning tool.
“Keep unspecified columns” changes the job
When enabled, requested columns come first and all remaining columns follow in their original order. When disabled, only valid requested columns are output, which also makes this useful for dropping columns. If no valid column remains, the tool stops instead of exporting an empty CSV.
Rows are not sorted
Only column positions change. Row order and cell values stay the same. If your goal is to sort records by price, date, or name, that is a row-sorting job instead.
A practical example
For a partner import that requires sku,title,price,stock, enter those four headers and turn off “keep unspecified” if extra fields are forbidden. Check both the header and two or three rows before upload.
If you want the deeper picture