When a compression tool takes 40 MB down to 38 MB, the problem is not compression. It is what is inside the file.
Diagnose before you compress
Thirty seconds of looking tells you which of the four causes you have, and therefore whether a compressor will help at all.
Divide the file size by the page count. A text page costs a few kilobytes; a scanned page at 300 DPI costs 1-3 MB; a full-page photograph costs 2-5 MB.
| Size per page | Almost certainly | Will compression help? |
|---|---|---|
| Under 100 KB | Normal text document | No — already near its floor |
| 100 KB – 500 KB | Text with some graphics | A little |
| 500 KB – 3 MB | Scanned pages or photos | Yes, substantially |
| Over 3 MB | High-DPI scans or print-resolution images | Yes, dramatically |
| Over 1 MB but looks like plain text | Structural problem | No — fix it upstream |
That last row is the interesting one, and it is where most frustration lives. A file that looks like text and weighs like photographs has something hidden in it.
The four causes, and which ones a compressor can touch
- Image resolution. A scanner set to 600 DPI when 200 would do produces nine times the data. Compressors fix this well — it is essentially all they do.
- Image encoding. Photographs stored as PNG or as raw bitmaps inside the PDF can be ten times larger than the JPEG equivalent. Compressors fix this too.
- Unsubset fonts. Embedding a complete font family costs several hundred kilobytes per weight, when the document may use forty characters of it. Compressors generally leave fonts alone.
- Incremental save history. PDFs can be edited by appending changes rather than rewriting, so a heavily revised file carries every previous version inside it. Compressors do not strip this; only a full rewrite does.
The split is clean: the first two are image problems, and every compression tool is an image tool. The last two are structural, and the only reliable fix is to produce the file again properly.
The hidden full-page background
One specific cause deserves its own section because it is common, invisible, and produces exactly the "text document that weighs 40 MB" symptom.
Templates — letterheads, certificates, branded report layouts, some invoice generators — frequently implement their design as a full-page image placed behind the text. Every page then carries a multi-megabyte photograph you cannot see as a photograph, because it looks like a border and a logo.
You can spot it by trying to select text over the decorated area, or by noticing that every page costs the same suspiciously large amount regardless of how much text is on it.
Compression does help here, because it is an image after all. But the real fix is in the template: a letterhead drawn with vector shapes and real fonts costs kilobytes instead of megabytes, and stays sharp at any zoom.
Flattening is not compressing
These get confused constantly, and they do different things.
Flattening merges interactive layers — form fields, annotations, comments, signature widgets — into the static page content. The form stops being fillable and the comments stop being editable. It usually makes the file slightly smaller, because the interactive structures go away, but that is a side effect rather than the point.
Compressing re-encodes images at lower resolution and quality. It does not touch interactivity at all.
If your goal is size, compress. If your goal is to stop someone editing the form fields or to make sure the annotations print, flatten. If a 2 MB form becomes 1.9 MB after flattening, that is the expected result, not a failure.
When re-exporting beats compressing
For the structural causes, going back to the source is not a workaround — it is the actual fix, and it is usually faster than fighting a compressor.
Re-exporting from Word, Google Docs, InDesign or whatever produced the file subsets the fonts, discards revision history, and writes a clean object structure. A 40 MB file with an editing history commonly comes back as 800 KB with no quality loss whatsoever, because nothing was thrown away — the waste was never content.
- From Word: File → Save as PDF, and choose "Minimum size" if the document is for screen reading.
- From Google Docs: File → Download → PDF. Exports are always clean.
- From a scanner: rescan at 200-300 DPI rather than 600, in greyscale if there is no colour on the page.
- From a design tool: export for screen rather than for press, and check that image downsampling is enabled.
If you do not have the source and the file is structurally bloated, the last resort is to render the pages with PDF to JPG at 150 DPI and rebuild. That discards the text layer, so only do it when the file needs to be viewable rather than searchable.
Frequently asked questions
Why is my PDF so large when it is only text?
Usually one of three things: a full-page background image behind the text from a letterhead template, embedded fonts that were not subset, or retained revision history from repeated saves. Compression barely helps with any of them. Re-export the document from its source instead.
Why did compressing barely reduce my PDF?
Because compression tools work on images, and your file is not large because of images. If the size is structural — fonts, save history, object bloat — a compressor has nothing to act on. Re-exporting from the original document usually fixes it completely.
Does flattening a PDF make it smaller?
Slightly, because the interactive form and annotation structures are merged into the page content and discarded. But that is a side effect. Flattening exists to stop fields being edited, not to reduce size — expect a few percent, not a few times.
How large should a normal PDF be?
Under 100 KB per page for a text document, 100-500 KB per page with some graphics, and 500 KB to 3 MB per page for scans or photographs. Anything much over 1 MB per page that looks like plain text has a structural problem.
Does scanning in colour make the file bigger?
Yes, roughly three times bigger than greyscale for the same resolution, because colour carries three channels instead of one. If the original page is black ink on white paper, scanning in greyscale or black-and-white is the single easiest saving available.
Can I remove the revision history from a PDF?
Not with a compressor. You need a tool that fully rewrites the file structure rather than appending to it, or — more reliably — re-export from the original document. Most PDF editors offer a "save as" that rewrites rather than an incremental save.