First determine what makes the PDF large
A scanned 200-page PDF full of photographs needs a different strategy from a mostly-text PDF with embedded fonts or a design file containing many high-resolution images. Compression is not one magic percentage slider.
Before changing the file, record its current size, page count, whether text is selectable, and whether image detail or print quality matters.
Four common size-reduction methods
The best method depends on the document's purpose. A screen-only copy can tolerate more image reduction than a print master, archival record, signed form, or evidence file.
| Method | What changes | Main risk |
|---|---|---|
| Image recompression | JPEG/WebP-like image quality/encoding | Visible artifacts, text in scans becomes fuzzy |
| Image downsampling | Pixel dimensions/resolution | Fine print and diagrams can lose detail |
| Object/font cleanup | Unused or duplicated resources | Usually safer but savings vary |
| Rebuild/print-to-PDF | Document structure is recreated | Can flatten links, forms, tags, layers, accessibility and metadata |
Use a two-copy workflow
- 1Keep the original
Never overwrite the only copy of a signed, legal, source, or print-quality PDF.
- 2Create a moderate compressed copy
Reduce images first when they dominate the file.
- 3Compare representative pages
Zoom into small text, scanned signatures, diagrams and photos.
- 4Test functionality
Check links, bookmarks, forms, selectable text and signatures that matter to the use case.
- 5Check the upload requirement
Stop compressing once the file is below the portal limit with acceptable quality.
Scanned PDFs need special care
A scan is often just one or more page images. Aggressive downsampling can make small printed text unreadable, especially when the source was already low resolution. If OCR text is present, verify that compression does not remove or misalign the text layer.
Do not rewrite signed or evidentiary PDFs casually
Many digital-signature schemes protect the exact document bytes or revisions. Recompressing, printing, or rebuilding the PDF can invalidate the signature or change evidentiary characteristics. Keep the signed original and confirm whether a separate smaller copy is acceptable.
Judge compression by use, not only percentage
A 70% size reduction is not automatically better than 30%. Compare readability and required features. For a portal upload, the right answer is usually the smallest file that remains clearly readable and functionally correct—not the smallest file the software can produce.
Scenario: a 28 MB report must fit a 10 MB portal limit
The report is searchable text with six high-resolution photos and several vector diagrams. Rasterizing every page may get the file under 10 MB, but it would destroy selectable text and soften the diagrams. A better pass preserves text/vectors and targets only oversized images.
After saving, reopen the derivative and test small labels, search/select, links and page count. Keep the 28 MB master separately; the 9.6 MB portal copy is a delivery derivative, not a replacement master.
28 MB / 62 pages / searchable text.
Six oversized photos dominate bytes.
Preserve text and vector content.
Search, links, diagrams, small text.
9.6 MB derivative; master retained.
Quality checks after compression
| Content | What to inspect |
|---|---|
| Scanned small text | Characters remain distinct at normal zoom/print size |
| Charts/line drawings | Thin lines and labels are not blurred |
| Photos | No obvious blocks/ringing in gradients/texture |
| Searchable text | Search/select still works |
| Forms/links/bookmarks | Required interactive features remain |
| Digital signatures | Understand whether rewrite invalidated the signature |
The same 20 MB file size can need completely different fixes
PDF A is a 12-page brochure with full-page photos. PDF B is a 300-page text report with embedded fonts and a few diagrams. PDF C is a scan where every page is a high-resolution image. Their byte counts can be similar, but the safe reduction strategy is not.
Photo-heavy files often respond to image recompression/downsampling; scans need readability/OCR care; mostly-text/vector PDFs may have much less safe image reduction available. Diagnose before applying a generic ‘strong compression’ preset.
| PDF type | Likely dominant bytes | First thing to inspect |
|---|---|---|
| Photo brochure | High-resolution photos | Placed image dimensions/quality |
| Scanned document | One raster image per page | Small text readability and OCR |
| Digital report | Fonts/vectors/images | Embedded resources before rasterizing |
| Signed form | Existing signed byte structure | Whether any rewrite is allowed |
Keep a preservation master and create a delivery copy when goals conflict
A print-quality or signed source and a small web-upload copy serve different purposes. Preserve the authoritative/high-quality original, then create an explicitly named derivative for the portal or email limit. That avoids turning a one-time size constraint into permanent quality loss.
If a portal limit cannot be met without unreadable text or broken functionality, solve the constraint rather than destroying the document: rescan inefficient pages, remove legitimately unneeded pages, split attachments if the destination allows it, or use an approved alternative transfer method.