Practical guidance
Fix unreadable characters in a product CSV
Names such as Café, Áo dài, and 東京 should survive a file round trip. Garbled text often means the bytes were decoded using the wrong character encoding.
Separate a display problem from damaged text
Try the original bytes with the encoding used by the exporting application. Work from the original file: saving a garbled view can turn a temporary decoding problem into permanent text changes. A replacement character, displayed as a diamond or question mark, needs special care.
- Keep the original download untouched.
- Select UTF-8 first when the source export is documented as UTF-8.
- If that fails, identify the export encoding from the source application rather than guessing from the product language.
- Re-export as UTF-8 after checking representative accented and non-Latin text.
Check the first header too
A byte-order mark can appear before the first header in some exports. It is file metadata, not part of the word Title or SKU. A parser should distinguish that marker from actual header text and keep an audit trail of normalization.
Worked example
Synthetic Unicode values that should remain intact
SKU,Title
00021,Café cup
00022,Áo dài
00023,東京 notebookDo not transliterate these names or delete accents to satisfy a parser.
What this check cannot establish
If the source system already replaced characters or saved mojibake, selecting a different encoding may not restore the original names. Obtain a clean export.
Check your file
Start with a copy. Review changes before downloading your import file.
Open the CSV tool