I compressed twice and it barely changed. Is the tool broken?

No. PDF structural optimisation can only remove redundancy once. A second pass has nothing left to work on, so a near-identical result is the correct outcome, not a failure.

How much smaller should a PDF get?

It depends entirely on content. A text-heavy PDF exported for print might drop 60–80%. A scanned PDF typically moves 2–10%, because almost all of its bytes are image data that browser-side tools do not re-encode.

Does splitting a PDF reduce the total size?

No — it distributes the same bytes across more files. That helps when the portal enforces a per-document limit, and does nothing when it requires a single combined upload. Check which applies before splitting.

Will converting to greyscale definitely shrink my PDF?

For scanned colour documents, substantially — often a third to a half. For text-based PDFs with no images, barely at all, since there is little colour data to remove in the first place.

Can I just rename a .jpg to .pdf to get under the limit?

No. Portals validate the file's internal format, not its extension, and a renamed file is rejected as corrupt. It also does not change the size, so it solves nothing even if it were accepted.

The limit is 500KB and my file is 502KB. What now?

Delete a blank page, or re-export at slightly lower quality. At a gap that small almost any single change closes it — including cropping a few millimetres of scanner margin off the page edges.

PDF Still Too Large After Compression?

Updated 12 September 2026

Compression ran, the file did get smaller, and it is still over the limit. This is a different problem from a PDF that refuses to shrink at all, and it has a different fix. The useful first move is not to compress again — it is to measure how far over you are, because the gap between your file and the limit tells you which of four quite different solutions you need.

Measure the gap before you do anything else

Divide your current file size by the limit. Under about 1.2× and you are in trimming distance — small changes will close it. Between 1.2× and 3× you need to change something structural, like page count or colour depth. Above 3× and no amount of optimisation will help; the document has to be produced differently or the images re-encoded. People waste the most time when they treat a 10× problem as a trimming problem, running the same file through compressor after compressor and getting a percent back each time.

Under 1.2× over: trim the edges

At this range the fastest win is usually removing pages you were never asked for. Portals typically want one certificate, and people upload the full scanned booklet including blank backs and the cover. Deleting blank pages alone often closes a gap this size. After that, check for an embedded thumbnail or attachments, and if the PDF was exported from a design tool, re-export it as 'smallest file size' or 'web' rather than 'print' — that single dropdown routinely accounts for more than every compression pass combined.

1.2× to 3× over: change the colour and the page count

This is the range where converting a colour scan to greyscale does real work, because it removes two of the three colour channels and typically takes a third to a half off the file. If the document is dark text on white paper, nothing of value is lost. Combine that with splitting the document — if the portal has a per-document limit rather than a per-application one, three files of 400KB each may sail through where one 1.2MB file is rejected. Check whether the upload page offers multiple fields before assuming you need one combined PDF.

More than 3× over: the file has to be remade

A 12MB PDF against a 500KB cap is not a compression problem. Either the pages were scanned at a resolution far beyond what the document needs, or it is a photo-heavy export that should never have been a PDF. Rescan at 200 DPI greyscale, or if you cannot rescan, accept that you need an image-aware server-side compressor that re-encodes the embedded page images — that is a genuinely different class of tool from anything that runs in a browser tab, and there is no browser-side trick that substitutes for it.

Do not compress the same file twice

Structural optimisation is idempotent. Once redundant objects, duplicate fonts and uncompressed streams are gone, a second pass finds nothing left to remove and returns a near-identical file. Running a PDF through three different compressors in sequence produces the same result as running it through the best one of them once. If the first pass did not get you there, the answer is upstream — resolution, colour, page count — not another pass.

Frequently asked questions

I compressed twice and it barely changed. Is the tool broken?: No. PDF structural optimisation can only remove redundancy once. A second pass has nothing left to work on, so a near-identical result is the correct outcome, not a failure.
How much smaller should a PDF get?: It depends entirely on content. A text-heavy PDF exported for print might drop 60–80%. A scanned PDF typically moves 2–10%, because almost all of its bytes are image data that browser-side tools do not re-encode.
Does splitting a PDF reduce the total size?: No — it distributes the same bytes across more files. That helps when the portal enforces a per-document limit, and does nothing when it requires a single combined upload. Check which applies before splitting.
Will converting to greyscale definitely shrink my PDF?: For scanned colour documents, substantially — often a third to a half. For text-based PDFs with no images, barely at all, since there is little colour data to remove in the first place.
Can I just rename a .jpg to .pdf to get under the limit?: No. Portals validate the file's internal format, not its extension, and a renamed file is rejected as corrupt. It also does not change the size, so it solves nothing even if it were accepted.
The limit is 500KB and my file is 502KB. What now?: Delete a blank page, or re-export at slightly lower quality. At a gap that small almost any single change closes it — including cropping a few millimetres of scanner margin off the page edges.