Start with the type of document
A digital text PDF and a photographed document store information differently. Structural optimization can compact PDF objects without rasterizing text, but it does not promise large image savings. Scan mode is more aggressive because it renders page images and compresses those images.
Read the tradeoffs before flattening
The scan-compressed copy loses searchable text, form fields, links, accessibility tags, layers, and digital signatures. Keep the original when those features matter. Scan mode handles up to twenty pages; this limit is about processing scope, not a promise that twenty pages fit a small byte allowance.
A target is a measured outcome
The saved PDF includes internal document structure as well as page content. At extremely small targets, that overhead alone can consume much of the allowance. The tool reports an unmet target rather than claiming success, and retains the original when attempted optimization makes the file larger.