Excel to CSV with the choices made explicit

You get the number, not the formula — a CSV has nowhere to put =SUM(B2:B40). Every formula cell in an .xlsx stores two things: the expression in an <f> element and the result of the last calculation in a <v> element, and the export uses that cached result. The consequence worth knowing is that if the workbook was written by a tool that never calculated — some export libraries write formulas with no cached value at all — those cells come out empty, and opening the file in Excel once and saving it fixes them. There is also a Formulas dropdown that writes the expression text instead, for when you are auditing a sheet rather than moving data.

Because Excel reads a .csv without a byte-order mark using the legacy system code page, not UTF-8, so Ştefan arrives as Åtefan. Tick UTF-8 byte-order mark and three bytes — EF BB BF — go in front of the file; Excel then reads it as UTF-8 on a double-click. Everything else you are likely to feed the file to, from pandas to psql to a Node script, is happier without the mark, which is why it is off by default.

By default the stored value is repeated into every cell of the span, because delimited text has no concept of a cell covering four columns. In the file itself a merged range is one value in the top-left cell plus a <mergeCell ref="A7:C7"/> record, and the other cells are genuinely empty — export them literally and a category column ends up with a value on the first row of each group and blanks underneath, which most downstream tools read as missing data rather than as repetition. Switch the Merged cells dropdown to leave the blanks in place when you want an exact transcription of the sheet.

Use commas unless the file is destined for a double-click in a European copy of Excel, in which case use semicolons. Excel splits a .csv on the Windows list separator, which is a semicolon in German, French, Spanish, Italian, Dutch and Polish locales, so a comma file opens as one column and a semicolon file opens correctly — and the reverse is true on an English machine. Tab-separated text is the neutral choice: no locale uses a tab as a list separator, and tabs almost never occur inside spreadsheet values.

That is the raw serial number, and it means the Dates dropdown is set to Excel serial or the cell had no date format attached. Excel stores a date as the count of days since 31 December 1899, with the time as a fraction of a day, and the only thing that makes 45678 look like a date on screen is the number format hanging off the cell. This page reads that format and converts on your behalf, so the default output is 2025-01-21 — pick date-only, US short or the serial itself if something downstream expects a different shape.

Hidden sheets appear in the picker with a 'hidden' label and export normally if you select one, and hidden rows are exported as ordinary rows. Nothing is filtered out silently, because a row hidden by an autofilter looks identical in the file to a row hidden deliberately, and quietly dropping either would change your totals. If a sheet should not leave the workbook, do not pick it — everything stays in this tab either way.

Because RFC 4180 section 2 specifies CRLF as the record separator, and Excel writes and expects it. Untick CRLF line endings for plain LF if the file is heading into a Unix pipeline, a Git repository or a database COPY command, where the stray carriage returns turn up as \r at the end of the last field. Both are read correctly by every CSV parser worth using, so this is about what receives the file, not about correctness.