You ran the file through a compressor, waited, and downloaded a PDF that is barely smaller than the one you started with — 24 MB became 22. The tool didn’t fail. It had nothing to work with. What compression can remove depends entirely on what is inside your file, and the fastest route to a genuinely smaller PDF is a one-minute look at where the bytes actually live. That look tells you whether the answer is a stronger setting, a different tool entirely, or accepting that this particular file is already as small as it can honestly be.
First, do the arithmetic
Divide the file size by the page count. That single number narrows the diagnosis more than anything else:
Under ~50 KB per page — a born-digital, mostly-text document. It is already close to its floor, because the text inside a PDF is stored compressed from the moment it is written. No setting will turn a 300 KB contract into a 30 KB one.
100 KB to 1 MB per page — embedded images are deciding the size: photos, logos, charts exported at print resolution. This is where a compressor earns its keep, typically cutting the file by half or more.
Over 1 MB per page — almost certainly a scan, a photographed document, or artwork at full resolution. The biggest wins live here, because each page is one large picture. A quick confirmation: try to select the text with your cursor. If nothing highlights, every page is a photo of a page — and image settings are the whole game.
The five reasons a compressor changes nothing
1. The file is mostly text. PDF stores text and page structure compressed by default. Running a compressor over it is compressing something twice — the second pass finds almost nothing, which is why that 24 MB → 22 MB result usually belongs to a different cause below, and a 300 KB text file simply stays a 300 KB text file.
2. The images are already JPEGs at the quality you asked for. Most PDF images arrive as JPEGs. Re-encoding a JPEG at a similar quality produces a file of roughly the same size — you only lose a little sharpness for your trouble. Real savings come from downsampling: storing the image at fewer pixels, not just re-squeezing the same pixels.
3. The setting was too polite. A “print quality” preset keeps images at up to 300 dpi. On a 300 dpi scan that is a no-op by design — the tool did exactly what it was told, which was: change nothing. For screen reading and email, 150 dpi is the sensible target; our Compress PDF calls these levels Printer (300 dpi), Ebook (150 dpi) and Screen (72 dpi).
4. Fonts add up quietly. Every embedded font family costs tens to hundreds of kilobytes. A dozen of them can outweigh the text itself. A rebuild that subsets fonts — keeping only the characters actually used — recovers some of it, but this is a modest lever compared with images.
5. The bulk isn’t page content at all. PDFs can carry attachments, page thumbnails, form data, hidden layers — and, most often, their own editing history. Many editors append every save instead of rewriting the file, so the old versions ride along invisibly. Quality settings never touch any of this; only a full rebuild does. (Ream’s compressor rewrites the entire file, so this junk is cleared at any quality level — occasionally that alone is the whole saving.)
Match the fix to the file
| What you observe | What it means | What actually helps |
|---|---|---|
| Can’t select any text; over 1 MB per page | Scanned or photographed pages | Compress at Ebook or Screen; add grayscale if colour isn’t needed |
| Sharp photos on otherwise normal pages | Camera images at full resolution | Compress at Ebook (150 dpi) — typically half the size or better |
| Mostly text, already under ~50 KB per page | Nothing meaningful left to remove | Honestly: nothing. Send it as it is |
| File grew after editing and re-saving | Appended save history and leftovers | Any full rebuild clears it — Compress at Printer keeps quality untouched |
| Only one section is actually needed | The file is long, not heavy | Extract the pages you need, or delete the rest |
What the compressor actually does — and why running it twice is pointless
Ream’s Compress PDF rebuilds your document from scratch with Ghostscript, the same engine print professionals use. Images are resampled to the target resolution and re-encoded; fonts are subsetted; the file structure is rewritten cleanly, which discards accumulated history and unused objects. Text and vector graphics pass through untouched — they were never the problem. You choose the trade-off: Printer (300 dpi, visually lossless for most uses), Ebook (150 dpi, the everyday choice), or Screen (72 dpi, smallest, for quick previews).
One pass is the whole story. Compression is a one-way ratchet: the first pass at a given level removes the pixels above that level, and a second pass at the same level finds nothing left — it re-encodes the same data, shaving a percent or two while softening the image further. If the first result is still too large, don’t repeat it — step down a level, or change strategy.
When smaller means fewer pages, not smaller ones
If only the signed page matters, sending it alone beats compressing all forty. Extract Pages pulls out just the range you need, Delete Pages removes what nobody asked for, and Split PDF breaks one heavy file into sendable parts — all three run entirely in your browser, so those files never leave your device. For the specific ceilings that started this search — a portal that wants 2 MB, an inbox that bounces at 25 — we’ve written a dedicated guide to hitting an exact size target and one for email limits in particular.
Is it private, and is it free?
Everything here is free, with no sign-up and no watermark. Page tools (extract, delete, split) run in your browser — the file never leaves your device. Compress and Grayscale need real Ghostscript, so those two run in our hardened server sandbox: the file travels over HTTPS, is processed in an isolated container with no network access, and is deleted the moment your download is ready. Nothing is stored, logged or used for anything else — the details are on our privacy overview.
Frequently asked questions
Why is my PDF still 20 MB after compressing? Almost always one of two things: the images are already at the quality your setting targets, or the setting preserved print resolution on a print-resolution scan. Do the size-per-page arithmetic above, then try the Ebook level — it downsamples to 150 dpi, which is where the large savings live.
How small can a scanned page realistically get? A readable A4 text page at 150 dpi grayscale lands around 100–300 KB. Pushing far below that costs legibility. If someone demands 50 KB per scanned page, what they really need is a born-digital file — or fewer pages.
Does compressing a PDF twice make it smaller? Not meaningfully. The first pass removes the data above your chosen quality; the second re-encodes what’s left, gaining a percent or two while blurring images further. Change the level, not the repetition count.
Does converting to Word and back shrink a PDF? No — it usually grows the file and damages the layout on the way. Conversion is for editing, not for size; we’ve documented what survives a Word conversion honestly.
Why does text barely compress? Because it is stored compressed already. The PDF format deflates text and structure streams when the file is written, so a compressor finds that work done. This is also why a text-heavy file that feels big usually isn’t — check the per-page number before assuming something is wrong.
Will converting to grayscale reduce the size? On colour scans and photo-heavy files, noticeably — often 30–60% on top of compression, since one channel replaces three. On born-digital text it changes almost nothing. Grayscale PDF does this in one step, and our guide to black-and-white conversion covers when it makes sense.
Is there such a thing as lossless PDF compression? Only structurally: rewriting the file cleanly, subsetting fonts, discarding history. That yields perhaps 5–20% on cluttered files and nothing on clean ones. Every large saving on an image-heavy PDF is lossy by nature — the honest question is which pixels you can afford to lose.
Can a PDF get bigger after compressing? Rarely, yes — when a file was already optimised harder than your chosen level, the rebuild’s re-encoding overhead can outweigh its savings. If that happens, keep the original; it was already at its floor. And if your file grows every time you save it in an editor, that’s appended history — a single rebuild clears it.
