Skip to content
ProCutOut

Case studies

What happened when we ran real files through the tools

Before a guide goes up on ProCutOut, the tool it describes is run on real files and the results are written down: the size before, the size after, what broke and what did not. This page collects those tests in one place, with the numbers that surprised us.

To be clear about what these are: they are our own tests, not customer stories. Each one used a handful of files, so treat the numbers as a worked example and not a benchmark. Your files will give different results.

1Image optimization

How far can a photo be compressed before it shows?

What we tested: Two photos and a screenshot, re-encoded at 13 quality levels, several sizes and three formats with the same in-browser encoder the Image Compressor uses.

What we measured

  • A 687 KB photo became 119 KB at 80% quality and 102 KB at 75%, with little visible change at normal size.
  • Re-saving that same JPG at 100% made it bigger: 897 KB.
  • For a 20 KB limit, resizing to 400 px at 72% quality scored 36.0 dB. Forcing the 1024 px photo down to 9% quality reached the same size but scored 30.7 dB.
  • Shifting the picture by one pixel between saves, as a crop does, dropped the quality score from 38 dB to under 29 dB after 25 saves.

What to do with it: Start at 80%, and reduce the pixel dimensions before you reduce the quality.

Read the full testTry the Image CompressorChecked

2File formats

What converting between JPG, PNG and WebP does to file size

What we tested: Photos, a UI screenshot and a transparent icon, converted between the three formats with our converters.

What we measured

  • A 687 KB JPG became a 2,006 KB PNG: 2.9 times larger, with identical pixels.
  • At the same quality score WebP was 29% smaller than JPG for a portrait, 45% for a product photo and 62% for a UI screenshot.
  • A transparent icon saved as JPG lost its transparency and grew from 9.1 KB to 11.3 KB.
  • Lossless WebP came out bigger than PNG for the screenshot.

What to do with it: JPG for photos other people must open, PNG when you need transparency, WebP when you control where the image is shown.

Read the full testTry the Image ConverterChecked

3PDF compression

Which PDFs shrink, and which get bigger

What we tested: A text report, a screenshot PDF and a photo PDF, run through the presets of the PDF compressor.

What we measured

  • A 1.4 MB screenshot PDF became 154 KB for a 200 KB limit, and a 571 KB photo PDF became 152 KB.
  • A 2.1 KB text report grew to between 56 KB and 199 KB depending on the preset, because the compressor turns each page into a picture.
  • The Balanced preset (150 DPI, 72% quality) cut the image PDFs by about 61 to 62%.
  • Best quality (200 DPI) saved only 28% on the screenshots and made the photo PDF 4% larger.

What to do with it: Compress scans and photos. Re-export a text document from the program that made it instead.

Read the full testTry the Compress PDFChecked

4PDF organization

What merging does to size, duplicates and form fields

What we tested: Image-based PDFs, small text PDFs and a 10-field registration form, combined with the Merge PDF tool.

What we measured

  • For image-based PDFs the merged file was within 0.1% of the combined size of the inputs: 1,982 KB from a 1,412 KB and a 571 KB file.
  • Merging the same 571 KB photo PDF three times gave 1,712 KB, with three copies of every picture.
  • The registration form had 10 form fields before the merge and 0 after.
  • Selectable text stayed selectable, because pages are copied and not redrawn.

What to do with it: Merging does not save space, and a form should be filled in before it is merged.

Read the full testTry the Merge PDFChecked

5PDF editing

Why a one-page extract can be as big as the whole PDF

What we tested: Two PDFs, one with three different photo pages and one with eight pages sharing a letterhead image, put through delete, extract, rotate, reorder and split.

What we measured

  • Deleting one of the three photo pages cut a 571 KB file to 416 KB (27% smaller).
  • Deleting a page from the 8-page letterhead PDF changed its size by only 0.3%.
  • Extracting page 1 of the letterhead PDF gave 256 KB from a 261 KB original.
  • Splitting it into 8 files totalled 2,049 KB, 7.9 times the original.
  • Rotating every page 90° and reversing the page order left both files exactly the same size.

What to do with it: Pages that share an image or font each carry a full copy of it when they are separated.

Read the full testChecked

6OCR and extraction

How scan quality changes OCR accuracy

What we tested: One known page of clean printed English, scanned at 14 settings and read by the in-browser OCR.

What we measured

  • 150, 200 and 300 DPI all gave zero wrong words for 11-point text. At 50 DPI more than a third of the words were wrong.
  • A 2° tilt was fine, 5° made 17% of the words wrong and 10° made 54% wrong.
  • Noise, faint ink and heavy JPEG compression caused no wrong words.
  • The tool's own confidence score followed the errors: 95% went with 0% to 2.3% wrong words, 85% with 17% and 75% with 54%.

What to do with it: Straighten the page before you worry about resolution.

Read the full testChecked

7Document conversion

What survives a PDF to Word conversion

What we tested: Six PDFs, among them an invoice with a table, a scanned page and a photo PDF.

What we measured

  • Every word of the four text PDFs came through, headings were detected by size and page breaks were kept.
  • The invoice's table became 13 paragraphs, one per row, with no Word table.
  • The scanned page and the photo PDF produced no file, only a message.
  • Tables, images, columns, fonts, colours and form fields were not carried over.

What to do with it: It keeps the words and the order. For a table use a spreadsheet converter, and for a scan run OCR first.

Read the full testTry the PDF to Word ConverterChecked

8Digital signatures

How much a signature adds to a PDF

What we tested: Test PDFs signed with a drawn signature, a typed name and two uploaded signature images.

What we measured

  • A drawn signature added about 3 KB and a typed name about 9 KB.
  • A clean 3000 × 1200 PNG added about 55 KB.
  • A phone-photo style JPG added about 223 KB, which turned a 2.5 KB invoice into 225 KB.
  • All 119 text items in the invoice were still there after signing.

What to do with it: Draw or type the signature, or shrink an uploaded one to about 600 pixels wide first.

Read the full testChecked

How these tests are run

Each study is a script that pushes test files through the same code the tool uses in your browser, then saves every result. If a figure appears on this page, a script produced it.

Because the tests are scripts, they can be run again when a tool changes. When a number moves, the guide and this page are corrected, and the date on the study changes with it.

The tests are run and written up by Hardik Sindhav, who builds the tools. How drafts are checked, and how AI is used, is set out in the editorial policy.

Quoting these numbers

You are welcome to quote any figure here in an article, a lesson or a video. Please link to the study it came from so readers can see how it was measured. If you think a number is wrong, tell us and we will re-run the test.

Used ProCutOut for real work?

An exam form with a 20 KB photo limit, a shop listing, a poster for print: if a tool here got a real job done, or failed at one, we would like to hear what happened. With your permission the story can be written up on this page.

Tell us about it