When a PDF refuses to shrink, it is usually because the bulk is in one specific place. Work through this list from the top; the first two items solve most cases.
Nothing beats fixing the source. 300 dpi colour is the default on most scanners and is overkill for anything read on a screen. Dropping to 200 dpi roughly halves the size; switching colour to greyscale can cut it again.
Rasterising each page at a chosen JPEG quality and rebuilding the PDF is the single most effective post-hoc method for scanned documents. It works because it throws away the image data the eye cannot use.
Scanners love producing blank backs and cover sheets. Deleting ten blank pages from a fifty-page scan is an instant 20% reduction with zero quality cost.
PDFs from design tools sometimes carry dozens of layers, each stored separately. Flattening collapses them into one.
Subsetting embeds only the glyphs actually used rather than the complete typeface. On a document using six fonts this can save a megabyte.
An image masked to show a small area still stores the whole picture. Cropping discards the hidden pixels for real.
A photo saved as PNG can be five to ten times larger than the same image as JPEG at 85%. If a page contains a photo, that photo should not be PNG.
Embedded spreadsheets, original source files and full edit histories are invisible but heavy. Stripping metadata is usually a one-click operation.
If the recipient can accept several files, splitting avoids aggressive compression altogether — often the better trade when quality matters.
Sometimes the right answer is not to shrink the file at all. Upload it once and share a URL; email attachments are a poor way to move anything above about 10 MB.
If your PDF viewer can show document properties, look at the total size of images versus fonts versus embedded files. That single number tells you which of the ten items above is worth your time.
Try it now: Compress PDF · Split PDF