Preparing a PDF that will actually arrive by email

An attachment that bounces costs more than the minutes spent avoiding it — it usually gets re-sent in a hurry, and hurried sending is when the wrong file goes out.

Know the real limit

The number you have in mind is probably the recipient's inbox limit, but the message has to pass your sending server, any gateway in between, and their server. The strictest one wins, and you cannot see it.

Attachments are also encoded for transport, which inflates them by roughly a third. A 20 MB PDF arrives as about 27 MB of message. Against a 25 MB cap it fails, and the bounce message will be unhelpful.

Treat 10 MB as safe and 20 MB as risky.

Find out why it is large

Compressing blindly is the wrong first move. Check the document with PDF Info and think about what is actually in it.

Text-only documents are never large. A hundred pages of text and vector graphics is typically well under 5 MB. If a text document is 40 MB, something specific is wrong — usually one embedded high-resolution image, or a logo placed as an uncompressed bitmap on every page.

Scans are large by nature. Every page is a photograph. At 300 DPI in greyscale, expect 100–300 KB per page; in colour, three times that. Fifty colour pages is comfortably over 30 MB and there is nothing anomalous about it.

Mixed documents — a report with a scanned appendix — are usually dominated by the scanned part.

The distinction matters because the fixes differ. A single oversized image is worth locating and replacing. A genuinely large scan needs either compression or a different delivery method.

Compress, then look at the result

Compress PDF re-encodes embedded images at reduced quality; text and vector content are unaffected. On a scan-heavy document, reductions of 50–80% are common. On a text document, the saving will be small because there was nothing to squeeze.

Always open the compressed file and check the pages that matter at 100% zoom. Compression artefacts are most visible on small text within images and on fine lines in diagrams — which is exactly the content people need to read. If the result is unreadable, the document was not the problem.

If the output is larger than the input, keep the original.

When compression is not enough

Send only what is needed. People routinely attach a 60-page report when the recipient asked about one appendix. Extract Pages takes seconds and produces a better email.

Split it deliberately. Split PDF into logical parts — not "part 1 of 4" arbitrary chunks, which annoy the recipient — and say clearly in the message what each part is.

Use a link. Above about 25 MB, a shared link is the honest answer. Check who can access it: a link that works for the recipient without signing in also works for anyone who is forwarded it. For a confidential document, that is a worse outcome than a bounced attachment.

Rebuild from source. If you exported the PDF yourself, exporting again with sensible image settings usually beats compressing a bad export. Print-quality settings embed images at 300 DPI or more; for a document that will only be read on screen, that is several times what is needed.

Before you send

Whatever route you take, the pre-send checks still apply — and they apply to the compressed or split copy, not just the original. Re-processing a file can change what is in it, and the version you send is the version that matters.

One last habit worth having: attach the file first, then write the message. It is a small thing, and it eliminates the most common email mistake there is.