Dext decides two Cost items are duplicates by comparing a fixed set of extracted fields. It doesn't compare the images, and it doesn't judge whether the documents look alike. Every field in the set has to match.
Which fields are in that set depends on the document type, so a receipt and an invoice are matched on different rules.
What Dext checks on every Cost item
Before Dext looks at anything type-specific, 4 conditions apply to every comparison.
If any one of them fails, the items aren't matched.
Both items have the same type. A receipt is never matched against an invoice, even for the same purchase.
Both items have the same supplier record.
Both items have the same total amount, and that total isn't 0.
The existing item has finished processing and sits in the same Dext account.
Note: Dext compares the supplier record, not the supplier name as printed. If extraction reads the name differently on the two documents, they'll sit under separate supplier records and won't be compared at all.
How Dext matches receipts
For receipts, Dext compares the supplier, total amount, date, and document owner. Two receipts are duplicates when all 4 match.
Dext also uses the time printed on the receipt, where it can read one on both documents. In that case, the two times have to be less than 10 minutes apart for the receipts to count as duplicates. If either receipt has no readable time, Dext skips the time comparison and matches on the other 4 fields alone.
Two receipts with no owner set are treated as having the same document owner, so a blank owner still counts as a match.
Dext doesn't compare receipt numbers. Two receipts from the same supplier, on the same date, for the same amount, to the same owner will be matched as duplicates even when the printed receipt numbers are different. This is the single most common cause of a genuine purchase being flagged, such as three identical fares bought minutes apart, or repeat charges of the same value on one day.
How Dext matches invoices, credit notes and refunds
For invoices, credit notes and refunds, Dext compares the supplier, total amount, and Document reference, the invoice number printed on the document. Date is used only as a fallback.
Two items are duplicates when the supplier and total match, and either:
both items carry the same Document reference, or
neither item has a Document reference, and the dates match.
When both items have a Document reference, Dext ignores the date entirely. The same invoice dated differently on two copies will still be matched. Dext also ignores spaces inside the reference, so "INV 1043" and "INV1043" count as the same.
Note: Document owner isn't compared for these document types, so the item Dext matched against may belong to a different user.
How Dext matches statements and remittance advice
For statements and remittance advice, Dext compares the supplier, total amount, date, and document owner, and all 4 have to match.
This is the same rule set as receipts, without the time comparison.
Summary of the matching rules
This table shows which fields Dext compares for each document type it checks.
Document type | Supplier | Total | Date | Document reference | Document owner |
Receipt | Yes | Yes | Yes | No | Yes |
Invoice | Yes | Yes | Only if no reference | Yes | No |
Credit note or refund | Yes | Yes | Only if no reference | Yes | No |
Statement/remittance advice | Yes | Yes | Yes | No | Yes |
Receipts are also compared on the printed time where both documents show one.
Document types Dext never checks for duplicates
Duplicate detection doesn't run on every Cost document type. For these types, Dext creates the item and processes it normally, whatever your duplicate detection mode is set to:
ATM withdrawal
Delivery note
Dental statement
Expense statement
Mileage
Other
Rental statement
Supplier statement
Note: Supplier statement and Statement/remittance advice are different document types, and only Statement/remittance advice is checked. If you change an item's type to one on this list, it stops being compared.
Why a missing field stops a match
Every field in the rule set has to match, so a field that was never extracted can't match anything. This is why Dext might sometimes let an obvious duplicate through.
A blurred total, a supplier name read two different ways, or a missing invoice number on one of the two copies can break the comparison. For what to check when this happens, go to Why didn't Dext detect a duplicate Cost item?