Numbering the pages you have, not the pages you started with
Numbers are applied to the document as it stands when you press the button — after any pages you removed, added or reordered. That sounds obvious and it is where most numbering goes wrong: number first, then delete a page, and every number after it is a lie that will not be noticed until someone reads a cross-reference.
Because the whole workspace is one document, the sensible order is to get the pages right and number last. Removing a page and re-numbering is two clicks, not two round trips through separate tools.
Starting somewhere other than one
Two settings cover the cases that actually come up. Skip the first page leaves a cover or title page unmarked and starts the count on the page after it — the usual choice for a report. Start at sets the first number, which matters when the document is a chapter or section of something bigger and its first page is really page 47.
They combine: skip the cover and start at 47 and the second sheet is numbered 47. When the format includes a total, the total counts the pages that were actually numbered, so "Page 47 of 58" stays internally consistent rather than mixing a section number with a document total.
Where to put them
Bottom centre is the safe default and reads as neutral. Bottom right suits documents that will be printed single-sided and flicked through, because the corner is where a thumb naturally lands. Top corners suit documents that will be bound at the bottom or hole-punched along the bottom edge.
If the document will be printed double-sided and bound, be aware that a fixed left or right position lands on the outer edge on one side and the inner gutter on the other. There is no mirrored option here yet; bottom centre avoids the problem entirely and is what most people should choose.
The number is drawn at a small margin from the page edge, sized to the page. On a mixed-size document — a report with a landscape table dropped in — each page gets its number positioned against its own dimensions rather than a single fixed coordinate that would fall in the wrong place.
What this does to the file
The numbers become part of the page content, exactly like a watermark. They print, they survive conversion to images, and they cannot be toggled off by a reader. The text of your document is untouched — nothing is re-encoded, no images are recompressed, and the file grows by a few hundred bytes.
If a page already has numbers printed on it from the original layout, these are added on top rather than replacing them, which produces two numbers. There is no way to detect and remove existing numbering reliably, because to the file they are simply text like any other. If the original has numbers you do not want, the honest answer is to fix the source document.