Start with an explicit table structure
CSV needs a header row with unique, non-empty names. Choose comma, semicolon or tab to match the file rather than assuming every CSV uses a comma. Quoted delimiters and line breaks belong inside their cells. Supply UTF-8 text up to 1 MB, with no more than 10,000 rows and 100,000 data cells. XLSX is not text: export it to UTF-8 CSV first.
Preserve identifiers as strings
CSV-to-JSON creates string values for every field. For example, 001 remains "001", and a long tracking number is not converted through floating-point arithmetic. A value that looks like a date stays text. This cannot restore zeros that your spreadsheet already removed before creating the CSV; fix the source first.
Use flat, consistent JSON records
JSON-to-CSV accepts an array of flat objects with exactly the same key set. Key order may differ; output follows the first record’s keys. Nested data, duplicate keys and missing properties stop conversion. A number token such as 1e3 is emitted as that original token text, avoiding arithmetic rounding, but CSV itself does not preserve JSON numeric types.
Review intentional representation changes
Null becomes an empty CSV cell, so a later conversion cannot distinguish null from an empty string. Formula-like values and headers are prefixed with an apostrophe in the CSV export; the preview shows original values and reports how many were protected. Import spreadsheet columns as text to avoid renewed type guessing. Do not treat CSV as a type-preserving JSON backup.
Check the file, not just the preview
The table preview is limited to 50 rows for readability, while a valid download contains every record. Check row count, column names and a few difficult values in the exported file. Use the CSV comparison tool for a later repeated export, or keep the original JSON beside the converted copy. The workspace does not upload or save your source after you leave.