Make a file small enough to e-mail, upload to a portal or attach to a form, without destroying the text.
Compressing a PDF re-encodes the parts that take the most space — almost always photographs, scans and images — while leaving text as text. That is why a 20 MB scanned document can drop to 1 MB with nothing visible lost on screen, while a 2 MB PDF made of typed text barely shrinks at all: there is nothing left to squeeze.
A light setting keeps images sharp enough for printing and is what you want for a contract going to a lawyer. An aggressive setting is for the portal that refuses anything above 5 MB and where the file only needs to be readable on screen. If the compressed result ever comes out larger than the original, keep the original: the tool will tell you and hand back the smaller of the two.
A file that is large because it holds 300 pages of dense vector plans will not become small. In that case split the document, or send only the pages the recipient needs.
No. Text stays as text; only images are re-encoded.
It probably contains little image data — text and vector graphics are already compact.
Yes, and that is where the gains are largest, since a scan is essentially a photograph per page.