What is taking up the space?
In a typical PDF, text and vector graphics are a few tens of kilobytes per page. A photo or a scan is measured in megabytes. A 20 MB report is almost always five photographs, or 30 scanned pages, not the words.
That is why "reduce the quality slightly" is such an effective lever: dropping a 12-megapixel photo to the 1,500 pixels a page can actually show removes 90% of its data and changes nothing you can see on screen or in print.
What does "without losing quality" really mean?
Strictly, any JPEG re-encode loses something. In practice the question is whether you can see it. Text is untouched — it stays as vector glyphs — so headings and body copy remain razor sharp at any zoom. Images lose detail you were never seeing at page size.
After a compression here, the result card shows the two pages under a divider with a measured similarity score for the page you are looking at. That is the honest version of the claim: not "no loss", but "here is what changed, on your file".
How do you hit a specific limit, like 1 MB?
Type the number. The tool works towards the target by lowering image resolution and quality in steps, checking the size after each pass, and stops as soon as it is under. A target it cannot reach with images alone falls back to flattening pages, which always gets there — at the cost of selectable text.
Common targets exist because someone else set them: Gmail attachments at 25 MB, many government portals at 2 MB or 5 MB, university submission systems at 10 MB.
When should you not compress?
Documents for printing at A3 or larger, photographs meant to be examined rather than read, and anything being archived as a master copy. Keep the original: compress a copy for sending.
A PDF that is already small and all text has nothing to gain — if a 400 KB text document will not shrink, that is because there is nothing in it to shrink, and a ZIP is the better way to get a few more percent.