Compression

Compression is what keeps PDF file sizes manageable, and a single document usually mixes several methods because it mixes several kinds of content. Text and vector drawing instructions compress losslessly with Flate (the same Deflate algorithm behind ZIP), so every character comes back exactly as it went in.

Images are the heavy part, and that is where the real choices are made. Photographs are typically stored with lossy JPEG, where a quality setting trades fine detail for smaller bytes; flat graphics and scanned text pages do better with lossless or specialised schemes. Beyond per-image encoding, a PDF can shrink by downsampling images that are stored at a far higher resolution than they will ever be displayed at, by removing unused objects, and by subsetting embedded fonts so only the glyphs actually used are kept.

Getting a good result is a balance: compress too hard and text turns fuzzy or photos show blocky artefacts. reader.me runs the whole optimisation pass in your browser, so a confidential report can be slimmed down for email without the original ever being uploaded to a compression service.