Why compressing a PDF sometimes makes it bigger

Run a PDF through a compressor and you occasionally get back a file that is larger than the one you started with. It looks absurd, and it is a common enough complaint that most tools quietly hide it. The reason is straightforward once you know what a compressor is actually doing. The first is lossless restructuring. A PDF is a collection of objects, and over a file’s life — edits, re-saves, incremental updates — those objects accumulate duplication and orphaned entries. Rewriting the file with object streams packs them more tightly. Nothing visual changes, text stays selectable, and the saving is usually modest: often 5 to 20 per cent, sometimes nothing at all on a file that was already written cleanly. The second is rasterisation. Each page is rendered to an image and that image is embedded in place of the original content. On a scanned document this is dramatic, because you are replacing a large photographic image with a smaller re-encoded one. On a text document it is a disaster: a page of crisp vector text that occupied a few kilobytes becomes a photograph of that text occupying several hundred. A tool that only rasterises will inflate any text-heavy PDF, every time. A tool that only restructures will report "no saving" on a file full of scans where the real win was available. Most tools pick one strategy and hope your file suits it. Try both and keep whichever is smaller — and if neither beats the original, return the original untouched rather than a larger file. That is what our compressor does: it attempts a lossless restructure, attempts a rasterised pass, compares all three sizes and gives you the smallest. On a scan-heavy file we have measured 8.8 MB down to 0.49 MB. On a clean text document it reports that the file is already about as small as it gets, and hands the original back. One trade-off is worth stating plainly: rasterising destroys the text layer. The page will look the same and be unsearchable, uncopyable and invisible to screen readers. If that matters, keep the larger file.

Tools

  1. Text-heavy PDF? Keep text selectable — expect a modest saving or none.
  2. Scanned pages or photographs? Allow rasterising — that is where the large wins are.
  3. If a tool offers no choice and the file grew, it rasterised a text document.