PDF compression does not reduce quality in a single step. It applies a series of techniques, each of which trades some amount of quality for some amount of size reduction. Understanding what each technique removes allows you to choose the right compression level for each document rather than applying the maximum setting and hoping for the best.
What is actually in a large PDF
PDF file size is dominated by images. A PDF with only text is typically small — a hundred-page report with no graphics might be 500 KB. The same report with one high-resolution photograph per page might be 50 MB. The photographs are stored at their original resolution inside the PDF, even if they will never be printed larger than postcard size. The simplest compression technique is reducing that image resolution to something appropriate for the intended use.
Text in PDFs is stored as font data plus character references. If a PDF embeds the full font file for a font that only uses twelve characters, the unused characters represent wasted space. Font subsetting — removing unused characters from the embedded font — can reduce size without affecting appearance at all. Most good compressors perform font subsetting as a standard step.
PDF files accumulate structure overhead over time. When a document is edited and resaved repeatedly, the PDF format appends changes rather than rewriting the whole file. This produces revisions that contain the current version plus every previous edit. Linearizing or rebuilding the PDF removes this accumulated overhead and can reduce file size significantly on heavily-revised documents.
The three compression levels and what they remove
Light compression applies font subsetting, removes embedded metadata that is not needed for display, and optionally linearizes the file. Images are not downsampled. The quality of visual content is unchanged. This level typically reduces file size by 10 to 30 percent and is appropriate for any document where the PDF will be printed, edited further, or must maintain exact visual fidelity.
Standard compression adds image downsampling to an appropriate screen-viewing resolution — typically 150 DPI. This is sufficient for any document that will be read on screen, sent by email, or uploaded to a web system. Print-quality images at 300 DPI or higher are downsampled to screen quality. For most professional documents, standard compression produces a file that looks identical on screen while being 40 to 70 percent smaller.
Aggressive compression reduces images further to low-resolution display quality and applies lossy compression to any color images. The resulting PDF looks acceptable on small screens but shows visible degradation when zoomed or printed. This level is appropriate for archiving documents you need to access but rarely read in detail, or for documents where the only requirement is that text remains readable.
Documents that do not compress well
PDFs that are already compressed compress poorly a second time. Lossless compression algorithms reach diminishing returns quickly — the second pass on an already-compressed file often achieves less than five percent additional reduction. If a PDF is already at a reasonable size, additional compression may not be worth the quality tradeoff.
Scanned document PDFs compress differently than PDFs created from digital sources. A scanned page is a photograph of paper. At any given resolution, the file size is determined by the scan resolution and the amount of visual noise in the image. Scans of clean white paper with black text compress well with lossless methods. Scans of aged or slightly yellow paper contain color variation that compresses less efficiently.
Preserving quality where it matters
Before compressing, identify which elements in the document must be preserved at full quality. Photographs in a portfolio require high fidelity. Charts and graphs in a report require legible labels. Signatures and official seals must remain clear for verification. If these elements exist in the document, use light compression or set a minimum image resolution that preserves their legibility.
If only part of the document contains quality-critical content, consider splitting the PDF into sections, compressing the text-heavy sections aggressively, and keeping the image-heavy sections at light compression, then merging the results. This approach is more work than a single-pass compression but allows different quality settings for different content types within one document.
Compression results in measured terms
Compression results vary enough by content type that quoting a single percentage is misleading. The figures below are typical ranges measured on documents of common types using standard compression settings.
A 10-page text-only PDF created from Word — a contract, a memo, an academic paper without figures — typically reduces from its source size by 5 to 15 percent. The source PDF is already efficient because text is a compact form of content. The reduction comes primarily from font subsetting and metadata cleanup, not image processing.
A 20-page corporate report with embedded photographs and charts typically reduces by 40 to 60 percent on standard compression. The original images were stored at print resolution; the compressed version stores them at screen resolution, which is sufficient for the document's normal use. The text content is unchanged.
A scanned document — a contract photographed page by page and saved as PDF — varies most. A scan saved at 600 DPI in color compresses to roughly 30 percent of its original size when downsampled to 200 DPI grayscale, which remains legible for text content. Scans already saved at a lower resolution have less compression headroom.
A presentation deck exported from PowerPoint or Keynote with image-heavy slides typically reduces by 50 to 70 percent. Presentations often use large images sized for high-resolution displays; downsampling for PDF distribution reduces file size substantially with no visible effect at typical viewing zoom levels.
When not to compress a PDF
Not every PDF should be compressed. Some categories of document benefit from staying at their original quality.
Legal documents with signature blocks should remain at original quality. Compression artifacts on signatures can affect their visual identity, and for documents that may need to be verified by signature comparison, compression introduces unnecessary visual change. The file size savings on a typical signed document are small relative to the cost of having to re-scan or re-sign if the compressed signature is rejected.
Archival documents intended for permanent storage should retain their highest available quality. The marginal cost of storage is low; the cost of having degraded the only remaining copy of an important document is irreversible. Compress copies intended for distribution, but keep the archival original at full quality.
Documents prepared for high-resolution printing should not be compressed in advance. Printers can downsample to their output resolution at print time. Compressing first removes information that the printer might have used to produce a sharper output, particularly for line art and detailed graphics where the source vector data is more flexible than a compressed raster version.
If a recipient may need to enlarge or zoom into the document — technical drawings, maps, detailed photographs — keep the source quality intact. Compression that looks fine at 100 percent zoom often shows degradation at 200 or 300 percent. The recipient may need that detail.
Compressing in Filum
Filum's PDF compression runs entirely in your browser — the file is never uploaded. The compression applies the standard techniques: font subsetting, metadata removal, and image downsampling in a single pass. Choose from three quality levels depending on how much size reduction you need against how much visual fidelity you want to keep.
For email-size requirements, standard compression typically reduces a 10 MB document to under 3 MB. Aggressive compression can reduce the same document to under 1 MB, with a visible reduction in image quality on zoom. The tool displays both the original and compressed file sizes before you download, so you can compare the result before keeping it.