Re-render pages at a resolution and quality you choose to cut file size — with a clear explanation of exactly what the trade-off costs you.
Almost always: images. A scanned page at 600 dpi is several megabytes on its own, and phone-photographed documents are worse. Text and vector graphics are negligible by comparison — a hundred pages of pure text is often smaller than one photograph.
That means compression is only worth doing when the file is image-heavy. Knowing which case you are in saves you from flattening a searchable contract for a 3% saving.
96–120 dpi reads fine on screen and hits email limits. 150 dpi is the practical default: clearly legible, prints acceptably. 200–300 dpi is print quality and shrinks much less. Below about 96 dpi small type starts to break up.
If the document is mostly text and simply feels big, the size is usually in one or two embedded images. Extracting the pages with PDF to JPG, compressing them with the image compressor, and rebuilding with JPG to PDF gives you finer control than a blanket re-render.
Each page is re-rendered at the resolution you pick and stored as a JPEG. For scans and image-heavy documents that is a dramatic saving, because the original page images are usually far larger than they need to be.
Text stops being selectable and searchable, because the page becomes a picture. That is a real cost — if your document is mostly text, do not compress it this way.
150 dpi at 75% quality is a good default for documents that will be read on screen or printed casually. Use 200–300 dpi for anything going to a print shop, and 96–120 dpi when the file must fit an email limit.
Because there was nothing to squeeze: text and vector drawings are already tiny. Rasterising them can even make the file larger. Compression pays off on scans and photographs, not on generated documents.
Not from the compressed copy, because each page has become a JPEG image. Keep the original PDF when search, copying, accessibility or later editing matters.