Excel Save As CSV pitfalls and round-trip damage

Recognize zeros, dates and long IDs lost when a spreadsheet becomes a second export.

Save As CSV is a second export, not a lossless copy

Opening a CSV in Excel and choosing Save As CSV feels like housekeeping. It is often a new export written with Excel’s current cell types, regional list separator and display formats. Characters that survived the first download can be rewritten on the way back out: leading zeros drop, long identifiers become exponents, date-looking labels become locale date strings, and UTF-8 names may need a BOM the next consumer does not expect.

TableMender cannot open an existing workbook. This guide is about recognizing a round-trip injury in the CSV you already hold, and about preferring a text-preserving path before Excel’s grid becomes the source of truth. The teaching file below is invented. It shows values shaped like a post–Save As extract, not a live customer table.

product_id,ship_window,tracking_id,name
127,4/3/2026,1.23457E+16,Café tray
904,3/4/2026,1.5e-10,Naïve lamp

Compare that with an intact source that still had 00127, 03/04/2026, a seventeen-digit tracking ID and Café. Once zeros, slashes and code points are gone, local cleaning cannot reconstruct them. Treat Save As as a decision to re-encode the grid—not as a neutral “save.”

What commonly changes on the way out

A sequence that avoids overwriting the good file

1. Keep the vendor CSV read-only or copy it. Do not Save As over the only good download.
2. If you must inspect in Excel, use Data → From Text/CSV (wording varies by version) and set identifier columns to Text in the preview. Prefer that over double-click open when codes matter.
3. When the next consumer is Excel, convert the untouched CSV with CSV to Excel so fields default to Text. Download a new workbook; leave the CSV alone.
4. If a destination truly requires CSV, export from TableMender after you have previewed delimiters and encoding in CSV check & repair. Compare a few identifiers with the source before you delete anything.
5. If the file you received already looks like the damaged sample—lost zeros, exponents, locale dates—stop. Ask for a re-export. Local Save As will not heal it. See Repair versus re-export.

Build a text workbook from the original CSV →

How to recognize a round-trip file

Open the CSV in a text editor or load it in CSV repair with an explicit Comma or Semicolon setting. Look for product codes that no longer start with zeros, tracking fields that contain E+ or e-, ship windows whose day and month order matches Excel’s region rather than the source label, and names whose accents look like two Latin letters instead of one. Those are value injuries. Changing the delimiter will not put the digits back.

Also check the first bytes. Some Excel UTF-8 CSV exports begin with a BOM. A BOM can help Excel reopen UTF-8; it can also stick inside a field if a later process concatenates files carelessly. The BOM guide separates a useful leading signature from a signature saved as data.

Edge cases and boundary cases

Excel for the web, older perpetual Excel and other spreadsheet apps do not share one Save As dialect. A habit that works on your laptop can fail on a colleague’s machine. Test with a three-row synthetic file before you trust a large extract.

“CSV UTF-8” and “CSV (Comma delimited)” are different menu items in many Excel builds. The UTF-8 option may add a BOM; the legacy option may use a regional ANSI code page. Neither menu restores characters that the grid already lost.

Filtered views and hidden rows can change which records Excel writes out. If you Save As while a filter is active, confirm whether the CSV contains the full table or only the visible rows. TableMender always exports the scope you select in its own preview; it does not inherit Excel’s filter state because it never opens the workbook.

Merged cells, multiple sheets and pivot caches do not survive CSV. Only the active sheet’s grid values are candidates for export, and even those are subject to type conversion. This site will not reconstruct sheets from a flat file.

Decimal commas and thousands separators interact with the list separator. Saving a European workbook as “CSV” can produce semicolons between fields and commas inside amounts—or the reverse on another PC. Preview field counts before you merge that file with a comma-only extract.

Empty cells at the edge of a row may be dropped by some exporters, changing the apparent width. Uneven widths then block download in TableMender until you fix the source. That block is protective; padding empty fields by guesswork can misalign columns.

Where cleaning still helps—and where it does not

After you have a structurally valid CSV whose values are still correct, CSV clean can remove fully empty records or review duplicates. It will not expand 1.23457E+16 into the original digits, and it should not. Deduplicating on a truncated ID can hide distinct shipments that only differed in digits Excel already discarded.

If Save As introduced a stray header mid-file because two sheets were copied into one, use the merge preflight mindset: compare headers and widths before you stack more rows. See Check CSV files before you merge them.

Before and after: an intact file beside a rewritten one

These are two synthetic teaching files, not a recording of one Excel session. Use them to see the difference between characters that are still present and characters that are already gone.

Load each file separately in CSV to Excel with Comma, a header, and every column on Text. Clear the input between files. The contrast is the point: preserve the intact download. Do not Save As over it from a general-format grid, and do not ask CSV clean to dedupe the rewritten file — truncated ids can collide. Structural problems that still contain the original characters belong in CSV check & repair, which is the decision in Repair versus re-export.

Step-by-step troubleshooting

1. The only copy you have already shows 127 or E+. Stop. Ask the source system for a fresh text export. Another Save As writes the grid you have, not the digits you remember.
2. Excel still needs a workbook today. Convert the untouched CSV with CSV to Excel. Identifier columns stay Text. Download a new file; leave the CSV read-only.
3. A colleague’s “CSV UTF-8” and your “CSV (Comma delimited)” disagree. Those menu items are different dialects. Compare delimiter, BOM and a non-ASCII name. Neither menu restores a value the grid already replaced.
4. The CSV is shorter than the sheet. A filter or a hidden-row Save As may have written only visible records. This site never opens the workbook, so it cannot tell you which rows Excel omitted.
5. Accents look like two Latin letters and the file is valid UTF-8. That is saved mojibake, not a delimiter bug. See Mojibake and BOM.

Honest limits

TableMender does not import .xlsx, does not drive Excel’s Save As dialog and does not claim to match every regional CSV dialect Microsoft ships. Guides cite Microsoft documentation for conversion behavior; interface names differ by version. When the grid has already replaced the characters, the honest fix is a fresh export from the system of record—or a text workbook built from a CSV that was never opened as General-format cells.

← All guides