The reduction comes from images, because that is where the bytes are. Photos and scanned pages are resampled and re-encoded; a re-encoded image is kept only if it is at least 8% smaller, so already-efficient files are left alone and compressing twice never degrades a document. Masks, 1-bit images and anything the browser cannot decode are copied byte-for-byte. The text layer is untouched, so a searchable scan stays searchable. If your file is dominated by a long scan and even Email & web stays above 500 KB, split it under a size and send the parts — that tool measures each part by actually writing it.
How it works
- Drop the PDF into Compress
- Read the computed size on each of the three cards
- Download the level that is under 500 KB and still looks right at 100% zoom
Questions people ask
How many scanned pages fit in 500 KB?
Roughly four to six ordinary text scans at the Email & web level, fewer at Balanced, and fewer still if the scans are in color or full of photographs. This varies a lot by scanner settings, which is exactly why Bindery shows the measured result for your file instead of a rule of thumb.
Can I compress once more if the first result is still too big?
You can, but it will not help much: images that were already resampled to the target resolution are not re-encoded again unless it saves real space. Move to a smaller level instead, or reduce the page count with Extract pages or Split under a size.
What does the published benchmark say?
Bindery measures its own compressor on the sample files that ship with the app and publishes the numbers at /compression-benchmark, with the method and files so anyone can repeat the run. Image-heavy decks drop by 90% or more; text-only samples move by a few percent.