Scanned batches and forwarded email attachments often arrive as a single file that actually holds several business documents back to back: an invoice, then a purchase order, then a goods received note, then another invoice. Document Splitting finds where each of those documents begins and ends, cuts the file into separate documents, and processes every one of them as a document in its own right.
Splitting is switched on per organization and runs automatically. Nobody has to mark the page breaks, and nobody has to split the file before uploading it. If a file turns out to hold only one document after all, Docupath leaves it alone and carries on.
How It Works
When splitting runs
Splitting is controlled by the Split documents from this organization setting on the organization or sub-organization a document arrives under. When that setting is off, documents are never split.
When it is on, Docupath checks every uploaded document for internal boundaries after it has read and processed the file, without any action from you. Two cases are skipped:
Documents that are already parts of a split. Once a file has been split, its parts are not checked again.
XML documents. These are never split.
How Docupath finds the boundaries
Docupath reads the text of every page and decides, page by page, whether that page starts a new document or continues the previous one.
A page is treated as the start of a new document when it shows the kind of signals a first page shows:
A document-type heading near the top, such as Invoice, Tax Invoice, Proforma Invoice, Purchase Order, Goods Received Note, Delivery Order or Delivery Note, Credit Note, Debit Note, Quotation, Sales Order, or Statement
A fresh header block carrying party and date fields, such as Invoice No, PO No, Order No, GRN No, Bill To, Ship To, Supplier, Vendor, Customer, Date, or Due Date
A document number that is different from the one on the previous page
A previous page that clearly closed a document off, with a Total, Grand Total, Balance Due or Amount Payable line, a signature or approval block, or a closing thank-you line, followed by a page that starts fresh with headers
A page is treated as a continuation when it clearly belongs to the document before it:
It begins in the middle of a line-item table, with columns such as Item, Description, Qty, Unit Price, and Amount, and no new header
Its item numbers carry on from the previous page, or it repeats the same document number
It holds only the remaining line items, terms and conditions, notes, or a signature block for the same document
A document-type word that simply repeats in a running header or footer on every page does not, on its own, make a page the start of a new document. This is what keeps a long single invoice from being cut into pieces.
Two rules always hold, whatever the pages look like:
The first page of the file always starts the first document.
Every page ends up in exactly one document. A page that is blank or that could not be read is kept with the document that precedes it rather than being dropped, so the parts together always account for the whole file.
Each document that Docupath identifies this way becomes its own PDF, containing that document's range of pages from the original file.
What happens to the file and its parts
While the file is being cut up, the document shows the Splitting status in your document list. After that:
The original combined file is retained, and the first detected document takes the place of the original entry in your list.
Every part becomes its own document, named after the original with its position appended, for example
Batch scan-1 of 3,Batch scan-2 of 3, andBatch scan-3 of 3.Any attachments that arrived with the combined file stay associated with the set.
Each part is then processed from the beginning as a document of its own. Document type detection, extraction, instruction builds, transformations, rejection rules, duplicate detection, flags, and Auto Review all run against the individual part, not against the combined file. A batch holding an invoice and a purchase order therefore ends up correctly classified as one Invoice and one Purchase Order, each with its own extracted data and its own review outcome.
If Docupath finds that the file holds only one document spanning all of its pages, no split happens. The document simply continues through normal processing.
What you see in the product
In the document list, each part is labeled with its position in front of the document number, shown as
1 of 3 -,2 of 3 -, and so on.In the review screen, a Linked to selector lists every part of the set alongside its document number, so you can move between the documents that came out of the same file while reviewing them.
While a document is in the Splitting status, its Flags column shows N/A, in the same way it does for documents in Processing, Retrying, or Failed. See Document Flags.
Where splitting fits in the document lifecycle
Splitting sits between ingestion and review. It applies to documents from every ingestion path that produces a splittable file, and it changes how many documents reach your review queue: one uploaded file can become several Pending documents. From the review stage onward, the parts behave exactly like any other document and each one is approved, validated, or rejected on its own.
Supported Configurations and Options
Splitting has a single control, set per organization.
Option | Where you set it | What it does | Default |
Split documents from this organization | Settings > Manage Organizations > add or edit an organization or sub-organization | Turns automatic splitting on for documents that arrive under that organization | Off |
Sub-organizations follow the parent by default:
A sub-organization created while the parent has splitting on inherits the setting.
Changing the setting on a parent organization updates every sub-organization that is still following the parent.
A sub-organization whose setting was changed on its own keeps its own value, and is not affected when the parent's setting changes later.
Other Technical Specifications
Parameter | Detail |
Applies to | PDFs, and file types Docupath converts to PDF (Office documents, HTML, and images) |
Never split | XML documents, and documents that are already parts of a split |
Page limit for splitting | 200 pages. A document above this limit fails with "More than 200 pages were detected in the document, which exceeds the allowed limit of 200." |
Parts produced | One per detected document |
Naming of parts |
|
Label in the document list |
|
Status while splitting | Splitting |
Navigation between parts | Linked to selector on the review screen |
Notes
Splitting only decides where documents begin and end within one uploaded file. It does not combine separate uploads into one document.
Every part is a full document in its own right: it appears in the document list, goes through review, and is approved, validated, or rejected independently of the other parts of the same file.
Because each part is classified on its own, a file holding a mix of document types produces correctly typed documents rather than one document forced into a single type.
Known Limitations and Edge Cases
Documents over 200 pages are not split. They fail with the page-limit message above and need to be split outside Docupath before upload.
XML documents are never split, regardless of the organization setting.
Blank or unreadable pages are kept with the document that comes before them. A blank separator sheet between two documents therefore lands at the end of the earlier document rather than starting a new one.
A file that holds only one document is left as it is. No parts are created and the document keeps its original name, with no
n of Nlabel.If the split cannot be completed, for example when one of the parts cannot be retrieved, Docupath retries the document automatically. If the retries are exhausted, the document moves to Failed with the reason recorded, and no partial set of parts is left behind.
Boundaries are detected, not configured. There is no way to adjust the detected page breaks for an individual document after upload.