Documents arrive out of order more often than they should. A feeder pulled two sheets at once, so page 7 is missing and there is a duplicate of page 6. The appendix ended up before the conclusion. Someone scanned the stack from the bottom.
The instinct is to split the file, discard what you do not want and merge the rest back. That works, and it is three tools and a lot of clicking for something that should be one operation.
Doing it in place
The whole document stays in view, so you can see what you are doing rather than reasoning about page ranges in your head. It runs in your browser, so nothing is uploaded.
Where this beats split-and-merge
Deleting scattered pages. Removing pages 3, 11 and 24 from a 30-page document by splitting means extracting four separate ranges and merging them in order. In a thumbnail view it is three clicks.
Reordering. Moving the appendix from the end to the front is a drag. With split-and-merge it means extracting two ranges and merging them in the reverse order.
Anything you need to see to decide. If the problem is "some of these pages are duplicates and I need to work out which", thumbnails answer that question directly. Page ranges do not.
Split and merge still win for one case: extracting a large continuous range from a very long document. Pulling pages 200-350 out of a 900-page file is a single range entry, whereas scrolling a thumbnail grid of 900 pages is tedious.
Working out what is wrong first
Before reorganising, diagnose the pattern. Scanner faults are usually systematic, and recognising the pattern is faster than fixing pages one by one.
Whole document reversed — the stack was fed from the wrong end. Reverse the entire order rather than dragging pages individually.
Every second page blank — duplex scanning of single-sided originals. Delete all the blanks, which alternate predictably.
Pairs swapped throughout — a duplex feeder scanning back-then-front. Systematic, so fix it as a pattern.
One or two pages out of place — a genuine misfeed. Drag those individually.
If the pattern is systematic, it is also worth fixing the scanner setting so it does not recur.
Check before you download
Two things people miss:
Page count. If you deleted three pages from a 40-page document, confirm the result has 37. An accidental extra deletion is easy to miss in a thumbnail grid and hard to notice later.
Landscape pages. Rotated pages render as landscape thumbnails and are easy to mistake for pages that need reordering. Check the content rather than the shape.
What organising drops
Same caveats as merging, since the underlying operation is similar: bookmarks, internal links and form field data may not survive. Page content, text and images come through unchanged.
For scanned documents — which is most of what needs reorganising — none of that applies, because a scan has no bookmarks or form fields to lose.
Deleting pages does not always shrink the file
Removing pages removes their content, so the file usually gets smaller. But not always proportionally: shared resources like embedded fonts and reused images stay if any remaining page still references them.
Deleting half the pages of a text document typically saves less than half the size. If you need a specific size, compress afterwards rather than expecting deletion alone to get you there.