CSV to Excel that keeps 007 as 007

Because Excel decided the value was a number, and any number over 11 digits gets displayed in scientific notation in a default-width column. A 17-digit order id like 90071992547409931 becomes 9.00720E+16 on screen and — worse — the underlying value is rounded to 9007199254740993000, because a spreadsheet stores every number as a 64-bit double and only 15 significant digits survive. This page checks digit length before it writes anything: any column with values past 15 digits is written as text, so the id you exported is the id you get back.

Yes — a column containing values like 007 or 02138 is written as text automatically, and the panel says so before you download. There is no way to keep a leading zero on a numeric cell, because 007 and 7 are the same number and the padding lives only in the display format. Writing the column as text is the only lossless answer, and it is what the Text Import Wizard has always asked you to pick by hand.

Because Excel splits a double-clicked .csv on the list separator from your Windows regional settings, not on the comma. On a machine set to German, French, Spanish or Dutch that separator is a semicolon, so a comma-separated file arrives as one long string per row. Building an .xlsx sidesteps the question entirely — the cells are already cells, so regional settings have nothing left to guess at.

Nothing in the file can tell you, which is why this page asks instead of guessing. When a slashed-date column contains any value with a first or second component above 12 the order is proved and the panel says which, and when every value is ambiguous a day-first / month-first switch appears in the status bar. Whatever you pick, the cell is written as ISO 8601 text — 2026-04-03 — which sorts correctly and can never be re-read the other way round.

Excel's autocorrect reads them as abbreviated dates on import and rewrites them as 2-Sep and 1-Mar, and once saved the original symbol is gone. It is not a rare curiosity: a 2016 survey found the error in roughly a fifth of published genomics papers with Excel supplements, and in August 2020 the HUGO Gene Nomenclature Committee renamed 27 human genes — SEPT2 became SEPTIN2 — because it was easier than fixing the spreadsheets. Nothing on this page autocorrects anything; a text column is written exactly as it arrived, character for character.

Because the ZIP is written store-only, with no compression, and every text cell carries its own XML wrapper. A 5,000-row, 8-column table produces a 2.31 MB package in about 21 milliseconds; the same XML deflated the way Excel writes it would be roughly a tenth of that. Open the file in Excel or LibreOffice and save it once and it shrinks to a normal size — the trade is a simpler, verifiable writer against a file that is temporarily fat.

No. The workbook is assembled by JavaScript in this tab and lands in your downloads folder as a finished .xlsx, so nothing about the conversion needs Microsoft Office, an account or a network connection. The resulting file opens in Excel 2007 and later, LibreOffice Calc, Google Sheets, Apple Numbers, pandas and anything else that reads OOXML.