How PDF compression works at this target
PDFs are containers — the file size comes from three sources: embedded images (usually 70-95% of total), embedded fonts (5-10%), and vector page content (rest). Compressing to a smaller target attacks the images first: recompressing each embedded image with JPEG at lower quality, then downsampling from 300 DPI (print quality) to 150 DPI (screen quality) or 96 DPI (mobile quality). A scanned 100-page PDF at 20 MB can hit 2 MB via this route without touching page count or layout. Text-only PDFs are already near their floor and rarely compress by more than 15-20%.