An Office file larger than its content: what the extra size can be, and how to find it

Last updated: 25 October 2026

Short answer: when a Word, Excel or PowerPoint file is much larger than its text, pictures and tables seem to need, the extra size is often something you cannot see on the page: a whole file embedded as an object, the copy of the source data a pivot table keeps, the uncropped original of a picture or the data kept to undo picture edits, embedded fonts, or parts nothing in the file uses any more. A large file is not proof of hidden data, and a small one is not proof there is none; the size is a clue to where to look. Open a copy as a ZIP, sort the parts by size and see which folder the large ones are in.

What can make an Office file larger than its content

Embedded objects. A spreadsheet, document or PDF inserted as an object is stored whole inside the file, in a folder named embeddings, even when the page shows only its first screen or an icon. Microsoft's page Linked objects and embedded objects explains how this differs from a linked object. See custom XML and embedded objects.
Pivot caches. A pivot table keeps its own copy of the data it summarises. Microsoft's page Reduce the file size of your Excel spreadsheets says "you can reduce the file size by not saving the pivot table source data cache with the file". The copy can hold rows that are no longer on any sheet; see the pivot cache.
Picture editing data and cropped areas. Microsoft's page Reduce the file size of your PowerPoint presentations says: "By default, when you edit an image, the data from the original is retained (to ensure the image can be restored)." Cropping hides part of a picture without removing it. The page on reducing the size of Word documents describes Delete cropped areas of pictures, and Discard editing data, which "removes stored data that's used to restore the image to its original state after it's been edited". Until you use them, what you cropped away is still in the file.
Embedded fonts. The same Word page says "embedded fonts will increase your document file size". A font is not hidden data about you, but it can explain the size.
Tracked changes. Deleted text stays in the file as a tracked deletion until the change is accepted or rejected; see accept or reject tracked changes. Text is small, so this rarely explains a large file on its own.
Custom XML. Other software can store its own data inside the file, in a folder named customXml.
Parts nothing uses. Every part of an Office file should be reached through the package's relationships. A program that does not clean up when something is deleted can leave a picture or another part behind that no relationship leads to. We found no Microsoft page that says when Office does this; treat it as something to look for, not something to expect.

Find the large parts

Make a copy of the file and rename the copy from .docx, .xlsx or .pptx to .zip.
List what is inside with each part's size: open the .zip in your file manager and show the size column, or run unzip -l copy.zip in a terminal.
Sort by size. In files Office writes, the large parts are usually in these folders; the names can vary, and [Content_Types].xml and the relationships say what each part is:
- word/media, xl/media, ppt/media: pictures, audio and video. A picture much larger than it appears may be uncropped or keep its original.
- word/embeddings, xl/embeddings, ppt/embeddings: embedded objects. A file ending in .xlsx or .docx in there is a whole document; copy it out and open it.
- xl/pivotCache: pivot caches; the pivotCacheRecords parts hold the rows.
- customXml: data from other software.
- word/fonts, ppt/fonts: embedded fonts.
To see whether a part is still used, search the .rels files in the _rels folders for its file name. If none names it, it is probably left over. A full check follows the relationships from the package's root and resolves each relative path, which is what ScrubDrop does.

Parts and relationships are defined in ECMA-376 Part 2, Open Packaging Conventions (ECMA-376).

What Document Inspector does with them

For Excel and PowerPoint, Microsoft's Document Inspector page says it finds embedded files and objects but cannot remove them; the Excel section says it "can't remove these items for you because it might cause the workbook to not work properly". For Excel, it also finds pivot tables and other BI features that "may have cached data that's saved with the workbook", without removing them. The page does not describe the same check for embedded objects in Word. It removes custom XML data. The page does not mention file size, cropped areas or picture editing data; those are handled by Compress Pictures and the image size and quality options on the pages above. See what Document Inspector misses.

Making the file smaller in Office

Compress the pictures with Delete cropped areas of pictures selected, and select Discard editing data. Neither can be undone, so keep the original.
In Excel, clear Save source data with file in the pivot table's options, or paste the pivot table as values if the recipient does not need it.
Replace an embedded object with a picture or a link if the recipient does not need the whole file.
Embed only the characters used, or no fonts, if the recipient has them.

What ScrubDrop does

ScrubDrop checks a supported .docx, .xlsx or .pptx file in your browser; the file is not uploaded. For free, it lists every supported kind of hidden data it finds, including embedded objects, pivot caches, custom XML and picture metadata. It does not show the size of each part; use the ZIP listing above for that.

Free, one file at a time: it removes the properties (with the thumbnail), comments and tracked changes, which it accepts.
ScrubDrop Pro also removes embedded objects, pivot tables with their caches, custom XML and picture metadata. In Word and PowerPoint, an embedded object's picture stays where the object was; in Excel, removing the object can take its preview picture with it, unless a drawing that stays still uses that picture; check the result. The values a pivot table showed stay in the cells. Where a pivot chart, a GETPIVOTDATA formula or a slicer uses the pivot data, it refuses the workbook rather than guess.
In every version it rebuilds the file from what the relationships reach, so parts nothing uses, and parts of a kind it does not keep, are not written; the result says how many parts were dropped.
It does not compress or re-encode pictures and does not delete cropped areas: picture metadata such as Exif is taken out, and the pixels stay. Embedded fonts are kept, because they draw the text.

Questions

Does a large Office file mean it has hidden data?

Not necessarily. Pictures, video and embedded fonts make files large with nothing hidden. The size tells you where to look: list the parts by size and see whether the large ones are embedded objects, pivot caches or pictures larger than they appear.

Does Document Inspector make the file smaller?

Microsoft's page does not say so. In Excel and PowerPoint it finds embedded objects without removing them, and in Excel pivot caches too, and cropped areas and picture editing data are handled by Compress Pictures and the image size and quality options instead.

Does ScrubDrop make the file smaller?

Only by what it removes. It does not compress pictures or delete cropped areas. Removing embedded objects or pivot caches, which is ScrubDrop Pro, can make a file much smaller; removing properties and comments rarely changes the size much.

Sources: the Microsoft pages linked on this page were checked on 25 October 2026. ScrubDrop is not affiliated with Microsoft.

See what is in the file before you send it, in your browser. Free: one file at a time; every supported kind of hidden data it finds is listed, and properties, comments and tracked changes are removed. Check a file.