PDF compression shrinks the size of a PDF so it's easier to email, upload or store. It works by re-optimizing embedded images, removing redundant data and cleaning up the file's internal structure.
Done well, the document looks the same but weighs a fraction of the original. Very aggressive compression can soften scanned images, so there's a balance between size and quality.
Shrink a file with Compress PDF.
Where a PDF's size comes from
Text is almost free: a hundred pages of it makes a small file. Weight comes from embedded images, and specifically from scans. A page scanned at 600 DPI in colour is a photograph several thousand pixels across, and twenty of those produce a document of tens of megabytes. This is why compression transforms a scanned document and barely touches a text one.
What compression actually changes
Compressing a PDF means re-encoding the images inside it at a resolution and quality suited to a document rather than a photographic print, and cleaning up redundancy such as duplicated fonts and unused objects. The text, the page structure and the layout are untouched. A document that is genuinely mostly text has little to work on, so if such a file is unexpectedly large the cause is usually embedded font subsets or a handful of oversized images rather than the document as a whole.
How much to compress
There is a point where a scanned page stops being reliably legible, and it arrives sooner for documents than for photographs, because text lives in exactly the sharp edges that lossy compression damages first. If a document will be printed, signed, or relied on as a record, check the result at full zoom before sending it. If it is a receipt going into an expense system, compress hard.
Better than compressing: scan smaller
Most oversized PDFs are created rather than inherited. Scanning at 200-300 DPI instead of 600, and in greyscale rather than colour where colour carries no information, produces a document that never needed compressing. For text documents, 300 DPI greyscale is both perfectly legible and enough for OCR to work well.
Splitting is not compressing
A common confusion: splitting a 20 MB file into four does not give four 5 MB files. Each output carries whatever its own pages weigh, plus the fonts and resources those pages reference, and an embedded font subset can end up duplicated across every piece. Split when you need to send part of a document; compress when you need to send a lighter one.