A three-page document arrives by email at 41 MB. It contains a letter, a table and a signature. Nothing about it looks expensive. Someone opens it, sees ordinary text, and reasonably concludes that PDFs are simply bloated.
They are not. A PDF is a fairly efficient container, and a well-made three-page text document is comfortably under 200 KB. When one is two hundred times that size, something specific went wrong, and it is almost always one of a small number of things. Knowing which one is the difference between compressing the same file every month and never producing an oversized file again.
What a PDF is made of
A PDF is a collection of objects: page descriptions, fonts, images, and metadata, wrapped in a structure that lets a reader find any page without reading the whole file.
Text is astonishingly cheap. A character is a couple of bytes plus a reference to a font. The complete text of a hundred-page book is under half a megabyte. If your document is large, the text is not why — a fact worth internalising, because it means dropping the font size or removing paragraphs will not help.
Everything expensive falls into three categories.
Cause one: images at camera resolution
This is the overwhelming majority of cases.
A modern phone camera produces images of twelve megapixels or more. As a JPEG that might be 3 to 5 MB. Paste four such photographs into a document and you have a 20 MB file before you have written a word.
The critical detail is that most software does not resize the image when you insert it. Word, Google Docs, LibreOffice and PowerPoint all display it at whatever size you dragged it to on the page, while storing every original pixel. A photograph shown as a two-inch square in your document is still a 4000-pixel-wide image in the file. On screen you see the small version; the file carries the large one.
Then you export to PDF, and the PDF faithfully embeds what the document contained.
The scale of the waste is worth spelling out. That two-inch square, printed at 300 DPI, needs 600 pixels across. You embedded 4000. You are storing roughly forty-four times more image data than the page can ever display.
The fix, at the source: resize images before inserting them. Almost every word processor has a “compress pictures” option — in Word it is under Picture Format, and it will offer to downsample everything in the document to a target resolution in one action. Choosing 220 DPI there routinely takes a 30 MB document to under 3 MB with no visible change at all. It is the single most effective thing you can do about file size, and it takes one click.
Cause two: scanning at the wrong settings
Scanners default conservatively, and the defaults are frequently wrong for what you are actually scanning.
The two settings that matter are resolution and colour mode, and their effects multiply.
Scanning a text document at 600 DPI in full colour produces roughly sixteen times the data of the same page at 300 DPI in greyscale. For printed black text on white paper, the extra detail captures paper texture and scanner noise. It records nothing a reader will ever use.
Sensible defaults for a text document are 300 DPI and greyscale, or even black-and-white if the scanner does a good job of it. Use 600 DPI only for photographs, fine artwork, or documents where a signature’s ink texture genuinely matters. Use colour only when the colour means something.
If the scan already exists at bad settings, you cannot recover the wasted space losslessly — but rasterising at a sensible resolution essentially rescans it from the pixels you have, and typically recovers 80 to 95 per cent of the size.
Cause three: fonts, and the “embed all characters” trap
A PDF embeds the fonts it uses so that the document looks identical everywhere. This is the whole reason the format exists and it is worth the space.
The question is how much gets embedded. A subset contains only the characters the document actually uses — usually 20 to 60 KB for a typical page of Latin text. A full embed contains the entire font: every character, every weight, sometimes 500 KB or more per family. For a font with comprehensive language coverage, several megabytes is possible.
Most software subsets by default. But an “embed all characters” or “embed for editing” option exists in many export dialogues, and turning it on for a document using four font families can add several megabytes to a text-only file. If you have a large PDF with no images at all, this is almost certainly the explanation.
The fix is at export time: leave subsetting enabled, and reduce the number of distinct font families in the document. Three families is a lot for any document; two is plenty.
Cause four: incremental saves
PDF supports appending changes rather than rewriting the file. Each save adds a new revision on top of the old one, and the old data stays where it was.
This is a genuine feature — it makes saving fast and preserves an audit trail — but for a document that has been through twenty rounds of comments, it means the file contains twenty versions of itself. Only the last one is displayed. All of them occupy space.
This is why a document that has been edited for months can be several times larger than the same content freshly exported, with no images and no unusual fonts. It is also the case where structural compression shines: rewriting the file discards every superseded revision at once, with no change to what anyone sees. Reductions of 40 per cent or more are common on heavily edited documents.
A word of caution: those old revisions can contain text that was later deleted. If a document has been through redaction or editing before circulation, a fresh export is safer than an incremental save, because it leaves nothing behind to recover.
Diagnosing your own file in one minute
You do not need specialist software.
Count the pages and divide. Take the file size and divide by the page count. Under 100 KB per page is a clean text document. Around 300 KB to 1 MB per page suggests a reasonable scan. Above 2 MB per page means images at camera resolution, almost without exception.
Try to select text. If words highlight, it is a digital document and the problem is images, fonts or revisions. If nothing highlights, it is a scan and the problem is resolution and colour mode.
Look at the pages with images. If a page with a photograph is twenty times larger than a page of plain text, you have found your answer and it is cause one.
Check the producer field. The metadata names the software that created the file. Export paths differ enormously in quality — some office suites and “print to PDF” drivers are notably worse at image handling than others, and knowing which produced the file tells you where to change the settings.
Producing smaller files from the start
Compression is a repair. These are the habits that mean you rarely need it.
Resize images before inserting them. The highest-value habit by a wide margin. If your word processor offers “compress pictures”, use it before every export.
Scan at 300 DPI in greyscale unless you have a specific reason not to.
Export rather than incrementally save when a document is finished and about to be circulated.
Use the right export preset. Most PDF exporters offer something like “standard”, “minimum size” and “print”. The default is often the print preset, which preserves image resolution you do not need for a document that will be read on screen.
Photograph documents in document mode. Most phone camera apps have one. It flattens perspective, crops to the page, converts to something close to greyscale, and produces files a fraction of the size of a raw photograph — while looking considerably better.
Question the page count. The cheapest byte is the one you never included. Sending the relevant three pages rather than the whole eighteen-page statement is smaller, faster and more considerate than any compression setting.
If you have an oversized file right now, Compress PDF will deal with it — but check first whether deleting empty pages gets you there losslessly. It often does.