Lena Fischer

I compare visually, because everything else is a proxy. A file that halves in size and turns a table into grey mush has not succeeded. Compression operates at several layers and only one of them is usually under your control. The file structure can be reorganised and streams rewritten, which saves space without changing a single pixel. Then there are the images, and here the loss happens. A scan stored as a bitmap can be downsampled or recompressed and will look worse. A vector drawing recompressed badly can break. That is why the sensible workflow differs by document. A text document should be recompressed at the structure level and its images left alone. A scan is almost entirely image, so the image settings decide the outcome. Resolution is the setting that matters most for scans. Text in a scan needs enough resolution to keep its edges crisp, and going below that produces the blotchy, unreadable result that makes a file feel worse despite being smaller. Grayscale conversion is the cheapest real saving on a monochrome document, and it costs nothing visually. It costs a great deal on anything with colour, so it belongs behind a check rather than a default. My pages end on a target rather than a number. The useful question is what resolution the page will be viewed at, and compressing for that beats compressing for a smallest-possible file every time. It also helps to know what the file is for before choosing a target. A file going into an email attachment and a file being uploaded to a document store have different budgets, and optimising for the smaller one usually makes the larger one worse.