A 24-page supplier contract, scanned on an office multifunction device, landed in my inbox last month at 61 MB. The same document exported straight from the word processor that produced it would have been under 1 MB. Nothing in it had changed. It had simply stopped being text and become 24 photographs of text, saved at whatever resolution the scanner had been set to years earlier and never questioned since.
That is the shape of nearly every oversized PDF: one bad default, repeated on every page. Which is encouraging, because the fix is usually one decision as well, and in most cases it costs you nothing you would ever notice on screen.
Where the megabytes actually go
Before you compress anything, work out what you are carrying. Fat PDFs are fat for a small number of reasons, and they need different treatments.
- Scanned pages stored at full resolution. A colour scan of an A4 page at 600 DPI is roughly 35 megapixels. Twenty of those and you are past 50 MB before anyone has written a word.
- Phone photos placed inside the page. A 4000-pixel-wide camera shot dropped into a column 15 cm wide carries three or four times more detail than the page can ever show.
- Complete font files embedded instead of subsets. A full family with several weights can add several megabytes for the sake of the eighty characters you actually typed.
- The same logo re-embedded on every page. Some exporters store one shared copy, others a fresh one per page, turning a 400 KB header graphic into 40 MB across a hundred pages.
- Saved editing history. PDF supports incremental saving, so a file edited twenty times can still contain deleted paragraphs, cropped-away image data and old versions of every annotation.
The distinction matters because the remedies do not overlap. A text-only report sitting at 40 MB has a font or revision problem, and resampling images will achieve nothing at all. A scanned contract at 60 MB is pure image data, and font tricks are irrelevant. If you are not sure which you have, run it through the free PDF compressor on 1mail.lt and compare the before and after: a file that drops by ninety percent was image-heavy, one that barely moves was never about images in the first place.
What "losing quality" actually means
People hear compression and imagine a uniform blur. In practice a PDF tool does three separate things and only one of them is visible. Resampling images to a lower resolution is the one you can see. Subsetting fonts, so only the glyphs you used are stored, changes nothing on screen, and discarding unused objects and old revisions is equally invisible. Take every invisible saving first, then decide how far to push the images.
For the visible part, the useful number is around 150 DPI. That is comfortably sharp on any screen, holds up on an office laser printer, and is the point where most documents stop looking obviously compressed. Reserve 300 DPI for material going to a commercial printer or for photographs that are the point of the document rather than illustrations in it. Below about 110 DPI, thin serifs and hairlines in diagrams start to break up, and once they break there is no getting them back.
Scans deserve a second decision: colour. If the original is black ink on white paper with a blue signature, the colour channel is carrying almost no information and roughly triples the size for it. Converting those pages to grayscale, or to a clean bitonal scan where the paper is genuinely white, often halves a file with no perceptible loss. Keep colour where colour is meaning: a highlighted clause, a coloured chart, a stamp that has to be recognisable.
If the weight comes from photographs you added yourself, deal with them before they go into the document rather than after. Shrinking a picture to the width it will actually appear at is far cleaner than letting a PDF tool guess later, and the image resizer and compressor at img.1mail.lt does it in the browser in a couple of seconds.
The same logic applies to any picture you publish anywhere, and the trade-offs are worth understanding once; the guide to compressing an image without visible quality loss goes through them with examples.
When two files beat one crushed file
There is a point where compression stops being the right question. A 90-page tender response with a scanned annexe does not want to be a 4 MB blur; it wants to be a 1.2 MB main document plus a separate annexe that the three people who need it can open on request. Splitting also survives forwarding better, and the same browser-based PDF toolkit extracts a page range without re-encoding anything that stays behind.
The ceilings that force the issue are lower than most people think. Gmail refuses attachments over 25 MB, and that limit counts the message after encoding, so the practical cut-off is nearer 20 MB. Most Outlook and Exchange setups sit between 20 and 25 MB, and plenty of corporate gateways are tightened to 10 MB without telling anyone. Near those numbers, a link beats an attachment: it does not sit in fifteen mailboxes forever, and you can replace the document without sending a correction.
Splitting and rejoining are the same skill in reverse. When you do need one file out of many, the walkthrough on merging PDF files without installing software covers the ordering details that trip people up.
The thirty-second check before you send it
Open the result at 100 percent zoom and go straight to the worst page, not the cover. The smallest body text, a footnote, an axis label on a chart, a stamp or a signature: those tell you whether the resampling went too far. A cover page with a big headline looks fine at almost any setting and teaches you nothing. If the document has an OCR text layer, try searching for a word you know is in it. Aggressive tools sometimes flatten pages back to plain images, which silently destroys search, copy-paste and screen-reader access, and a slightly larger file that is still searchable beats a small stack of pictures.
Know when to stop, too. Compressing an already-compressed PDF a second time is close to pure loss: the images have already been through a lossy encoder, and running them through another one degrades them again while barely moving the byte count. If a file only drops from 5.4 MB to 5.1 MB, that is the file telling you it is done. Go back to the original source instead and export it properly, or split it.
One last note on the documents that matter most. Payslips, contracts, ID scans and medical letters are exactly the files people upload to the first search result they find. Anything processed in the browser never leaves your device, which is the reason to prefer that kind of tool for personal data. And when a service demands an email address before it will hand your own file back, a disposable inbox from 1mail.lt keeps that transaction out of your real mailbox.
Questions people ask about shrinking PDFs
How small can a scanned PDF realistically get?
A typical office scan drops by 80 to 95 percent, so 60 MB becoming 3 to 8 MB is normal. The savings come from resampling to about 150 DPI and, where appropriate, dropping to grayscale. If your file barely moves, it was never image-heavy, and the weight is in fonts, embedded attachments or saved revisions.
Will compressing a PDF break the signatures or form fields?
A digital signature validates the exact bytes that were signed, so any re-save invalidates it. Compress first, then sign. Interactive form fields usually survive normal optimisation, but tools that flatten pages to images will turn the fields into a picture of fields, so test one copy before you send the batch.
Is 150 DPI really enough for printing?
For everyday office printing on a laser or inkjet, yes, and most readers cannot tell the difference from 300 without a magnifier. Where 150 is not enough is commercial print production, fine line art, and photographs meant to be examined closely. When in doubt, print one page at each setting and compare the smallest text.
