First, Understand What Takes Up the File Size
PDF optimization isn't blind compression — you need to know where the bulk is first. Common proportions:
- Images: In scanned documents, images take up over 80%, the absolute bulk
- Fonts: Chinese PDFs with full font sets embedded can be tens of MB
- Metadata: Usually small, but accumulated can be a few KB to tens of KB
- Redundant objects: Discarded objects from merging and editing, 5%~15%
Targeted treatment is far more effective than one-size-fits-all heavy compression. When using the PDF Compression tool, first check the size change before and after optimization to determine which content type dominates.
Image Optimization (Most Effective)
Downsampling
Reduce high-resolution images to a reasonable DPI. 100150 DPI is enough for screen reading, 200300 DPI for printing. Scanned documents are often 300+ DPI — dropping one level immediately reduces size.
Re-encoding
- Photos use JPEG, quality 70~80 with no visible loss
- Line art and content with transparency use PNG
- Black-and-white scans use CCITT G4 for the smallest size
Merging Duplicate Images
The same logo embedded on every page is stored repeatedly. Optimization tools can deduplicate — store once and reference.
Font Optimization
Subsetting
Embed only the glyphs actually used in the document. A single Song typeface has tens of thousands of characters, but a document may use only hundreds — after subsetting, the font portion can drop from tens of MB to tens of KB. Essential for Chinese PDF optimization.
Remove Unused Fonts
Fonts declared in the document but not actually used should be cleaned up.
Cautious Embedding
If the document uses only common fonts (such as system-bundled ones), consider not embedding them and relying on the reader's local fonts. But cross-device display may be inconsistent — use with discretion.
Metadata and Redundancy Cleanup
Slim Down Metadata
Remove excess metadata left by editing software (author, modification records, software info), keeping only what's necessary. If you care about privacy, this step can also remove fields containing author information.
Clean Up Redundant Objects
After merging and multiple edits, unreferenced XObjects and discarded streams appear. A clean pass can save 5%~15%.
Compress Object Streams
Pack multiple objects into compressed streams — modern PDFs all support this, and tools do it by default.
Linearization (Fast Web Viewing)
Linearized PDFs let browsers display the first page while downloading, without waiting for the entire file to download. Suitable for PDFs placed on websites for online viewing. The size may not decrease, but the experience is better.
Optimization Strategies for Different Use Cases
| Use Case | Strategy | Expected Effect |
|---|---|---|
| Email attachment | Image downsampling 150 DPI + font subsetting | Reduce 50%~70% |
| Website download | Stronger compression 120 DPI + full optimization | Reduce 60%~80% |
| Archival storage | Light optimization, preserve clarity | Reduce 20%~40% |
| Print output | No downsampling, only font subsetting and redundancy cleanup | Reduce 10%~20% |
Optimization Considerations
Lossy Process
Downsampling and JPEG re-encoding are lossy — multiple optimizations accumulate loss. Keep the original file and optimize from the original each time.
Don't Over-Optimize
Over-compressing contracts, stamps, and signature documents may affect legal validity. Test one copy first to confirm clarity.
Compatibility
PDFs with extreme compression or new encodings (such as JPEG2000) may not open in older readers. Use standard encodings for external distribution.
Summary
PDF optimization is a "diagnose + treat" process: if images dominate, downsample and re-encode; if fonts dominate, subset; finally clean up redundancy. Choose the optimization level based on the use case — don't use the strongest level for all files. The DocsAll compression tool runs in the browser, so sensitive documents aren't uploaded. Combined with PDF Merge, "merge first, then optimize" works best.