PDF compression guide
If two PDFs of the same letter can differ in size by a thousand times, something fundamental is going on. In our own testing a born-digital text PDF was 5 KB while a scanned colour version of similar content was over 7 MB. Understanding why is the single most useful thing you can know about PDF size, because it tells you whether compression will help a little or a lot.
A born-digital PDF is one created straight from software - a word processor, spreadsheet, design app or a 'print to PDF' option. Its pages are stored as text characters, font definitions and vector shapes: compact instructions that a viewer redraws crisply at any zoom. Because there are no pixels to store, these files are small from the start and barely respond to compression. Ours was 5 KB, and the compressor left it untouched because it was already under every target.
A scanned PDF is the opposite. Each page is a photograph of paper - a grid of pixels - so the file stores millions of coloured dots rather than a few instructions. A full-colour page scanned at 300 DPI holds an enormous amount of image data, which is why our three-page scan started at 7.16 MB. There is no text in there for the computer to read; it is a picture that happens to show words.
Compression works almost entirely on images, so a scan has a great deal it can remove and a born-digital file has almost nothing. That cuts both ways. The scan compressed from 7 MB down to 450 KB at our 500KB target - a huge saving - while the text PDF could not get smaller because it was already tiny. If your file barely shrinks when you compress it, that is usually a sign it is already lean, not that the tool failed.
Whenever you control how a document is made, prefer exporting over scanning. Save or 'print to PDF' directly from the application that created it, rather than printing it and scanning the paper back in. The result is smaller, sharper and has selectable text. This single habit avoids the size problem entirely for anything you author yourself.
Sometimes a scan is unavoidable - a signed form, an ID, an old paper record. Then the levers are the familiar ones: scan in greyscale when colour is not essential, use a sensible resolution rather than the maximum, crop empty borders, and compress to the largest target your portal accepts. Our scan stayed perfectly readable at 500KB, so you rarely need to push a scanned document into the strictest tiers.
Because scanning turns text into images. The scan stores every page as a grid of pixels, which is far heavier than the compact text and vectors in a born-digital PDF.
Usually only a little. Text and vector PDFs are already compact, so there is little image data to remove - in our test a 5 KB text PDF could not shrink further.
Use 'Save as PDF' or 'Print to PDF' directly from the program that created the document, instead of printing it and scanning the paper copy.