Documentation/ Uploading documents/ Formats and limits
Formats and upload limits
Dijit.app takes PDFs, images and ZIP files. The limit is 15 MB a document and 100 MB a ZIP, holding at most 100 files.
On this page
What it accepts
| Format | Where it fits |
|---|---|
| The one to aim for. PDFs produced by the sender's own system carry the text inside them, and read most accurately. | |
| Images | Photographs and scans in the usual formats. Common when capturing on a phone. |
| ZIP | A compressed file holding documents. It is unpacked in the browser before anything is uploaded. |
Spreadsheets, word processor files and email files are not accepted. A document in one of those has to be turned into a PDF first.
Size limits
| What | Limit |
|---|---|
| One document, PDF or image | 15 MB |
| A ZIP file | 100 MB |
| Documents inside a ZIP | 100 |
| Each document taken out of a ZIP | 15 MB |
| One batch on an accountancy practice account | 250 pages or documents |
An invoice PDF made by the sender usually runs to a few hundred kilobytes. A colour scan of several pages rarely gets past five megabytes. Going over 15 MB nearly always means the scanner is set far higher than it needs to be.
What happens when a limit is exceeded
One oversized file does not stop the batch:
- A document over 15 MB is dropped and you are told.
- Inside a ZIP, files over 15 MB are dropped one by one; the rest go in.
- If a ZIP holds more than 100 files, the first 100 in alphabetical order are taken and you are told the rest were left out.
To make a scan smaller, turn the scanner's resolution down. Between 200 and 300 dots per inch in black and white is plenty for reading a commercial document.
File names
No file is ever refused because of how long its name is. Long names are trimmed automatically as the file goes in:
| What | How much is kept |
|---|---|
| The document's base name | 80 characters |
| The original file name | 150 characters, keeping the extension |
The trimming touches the name the file is stored under, not what is on it or what was read off it. If the name is how you find your documents, put the part that matters at the front, the supplier, the number or the date, because that is the part that survives.
What counts as one unit
Your plan is spent on what comes out of processing, not on how big the file was:
| Account type | Unit |
|---|---|
| Business | A page processed |
| Accountancy practice | An invoice processed |
On a business account, throwing away the pages with nothing useful on them before processing, cover sheets, terms and conditions, annexes, cuts what you spend. You do it on the upload screen.
Quality of the original, and accuracy
The 99 % figure is for an original PDF. On photographed documents it follows the quality of the image. Most reliable first:
- A PDF made by the sender's system. The text is inside the file.
- A scan. Whole page, straight, evenly lit.
- A photograph. The whole document in frame, no shadows, nothing covering it, and the sheet as flat as you can get it.
- Worn-out copies. Photocopies of photocopies and faded till receipts cut the accuracy and mean more correcting by hand.
A supplier whose documents always read badly is rarely an image quality problem. It is usually how their data is laid out, and it is fixed with a supplier correction rule, set once and applied to everything that comes afterwards.