Why some PDFs do not get much smaller
A PDF can be large for very different reasons. A 100-page text document may be small because text and vector lines take little space. A two-page scan may be huge because it contains high-resolution photographs of paper. Some PDFs have already been optimised by the software that made them, so another compression pass has little left to remove. That is normal, not a failed result.
The risky approach is to rebuild every file as a low-quality image. It may reduce a large scan, but it can also make small text fuzzy and remove selectable text. This tool compares candidates and keeps the original when it is already the smallest safe result. That guard is useful for exactly the frustrating case where a 'compressed' download turns out larger than the source.

