Which PDFs actually shrink, and which never will

Compression does almost all of its work on the images inside a PDF. It re-encodes them at a resolution and quality suited to reading rather than printing, and leaves the text alone. So the useful question about any PDF is simple: how much of it is pictures?

Scans are entirely pictures. Every page is a photograph of a sheet of paper, often captured at 600 DPI when 150 would read identically on a screen, and these are the files that fall by 80 or 90 percent. A phone photograph turned into a PDF behaves the same way, because a modern camera produces far more detail than a document needs.

A PDF exported from Word, LaTeX or a web page is mostly text instructions and vector lines, which are already compact. Running it through compression may save a few percent, and sometimes nothing at all. That is not a failure of the tool; there is nothing redundant left to remove. Compressing it a second time only costs quality.

One more thing worth knowing: files exported from Photoshop or Illustrator can carry a hidden editable copy of the original artwork, which is never used to display the page and can be most of the file size. Our compressor strips that out, which is why some design PDFs collapse dramatically while looking untouched.