Guides / Why Excel files get corrupted

Why Excel files get corrupted: the mechanics, the causes, and the habits that prevent it

Updated August 29, 2026

Corruption feels random. The file was fine at 5pm and broken at 9am. It isn't random. Nearly every corrupt workbook traces back to one of a short list of causes, and most of those are preventable. Understanding one piece of internal anatomy explains almost all of it.

The anatomy: a zip with its table of contents at the end

A modern Excel file (.xlsx) is a ZIP archive containing XML files: one per worksheet, a shared table holding all the workbook's text (xl/sharedStrings.xml), styles, and the relationship files that wire everything together. Two structural facts drive almost every corruption story:

• The ZIP's index (its central directory) lives at the very end of the file. A write that stops early leaves an archive whose contents are present but whose index never got written. Excel reads the index first, so a truncated file is rejected outright, even when every row of data survived.
• One workbook, many internal parts. Damage to a single part has outsized effects: a broken shared-strings table blanks every text cell in every sheet at once, and one malformed worksheet can trigger “unreadable content” for the whole file.

When Excel saves, it writes the new version out and swaps it into place. Interrupt that dance at the wrong moment and the file on disk is neither the old version nor the new one.

The common causes

1. Interrupted saves. Excel crashes mid-save, Windows force-restarts for updates, the laptop battery dies, someone kills a frozen Excel from Task Manager. The save's final rename never happens, or happens onto a half-written archive.
2. Saving directly to network shares and VPNs. A save over SMB or a VPN is a long conversation over a link that can hiccup. A dropped connection mid-write is an interrupted save you never see; the file quietly lands torn.
3. USB sticks pulled too soon, and cheap flash. Operating systems buffer writes; yank the stick before the flush and the tail of the file never arrives. Worn or counterfeit flash drives also silently return bad data. This pairs with data-recovery damage. See the USB recovery guide.
4. Cloud-sync conflicts. Two machines (or one machine and a phone) editing the same synced file, a sync client uploading while Excel is mid-save, or an interrupted partial upload can produce a file stitched from two versions.
5. Failing storage. Bad sectors and dying SSD cells corrupt whatever happens to occupy them. If several files break around the same time, suspect the drive, not Excel. Back everything up now.
6. Very large, very old, heavily-abused workbooks. Giant monoliths accumulate risk: enormous saves take longer (bigger interruption window), crash more, and often carry years of internal cruft. The workbook that everyone edits, every day, for five years, is the workbook that eventually breaks.
7. Transfer mangling. Rarer today, but legacy FTP in ASCII mode, over-eager email gateways, and misconfigured download proxies can alter bytes in transit. If a downloaded file won't open, re-download before repairing. See the extension-not-valid guide.

Prevention that actually works

• Use versioned storage. Keep working files in OneDrive, SharePoint or Google Drive. Every one of them keeps version history, turning most corruption events into a two-click restore. This is the single highest-value habit.
• Turn AutoRecover on and short. File → Options → Save: check AutoRecover, set the interval low (a few minutes), and note the AutoRecover file location. That folder has rescued a lot of mornings.
• Don't live-edit off USB sticks or network shares. Copy the file to the local disk, work there, copy back when done. Local saves are fast and far less likely to be interrupted.
• Eject removable drives properly. Boring, effective. The seconds saved by yanking the stick are occasionally very expensive.
• Split the monolith. If a workbook has grown huge and slow, break it up or archive old sheets. Smaller saves, smaller blast radius.
• Let one writer win. Co-edit through the cloud versions of Office/Sheets where merging is designed in, not by two people saving the same .xlsx on a share.

Frequently asked

Related guides

• “File format or extension is not valid”When the container is the casualty: the full fix ladder.
• “We found unreadable content”When one internal part is torn and Excel offers to amputate.
• Corrupted after USB or data recoveryThe deep dive on flash drives, undeletes and truncation.
• XLSX repair tools: free vs paidIf prevention is too late, the honest map of every fix.