Best practices for invoice formatting.
This guide has been created to help customers understand how to optimise the extraction automation of invoice documents. In many cases providing guidelines for suppliers on their invoice formats can provide improvements for our customers.
Ensuring that invoices you receive contain all the necessary information in an easy-to-read format is one of the best ways to improve document processing efficiency.
Firstly, ensure what you are uploading is not excluded from the Zudello platform.
For definitions of particular terms, please refer to our glossary of terms.
How does Zudello read invoices?
Zudello extracts all the information from your documents using advanced cognitive recognition. Zudello’s unique extraction technology takes the OCR result from multiple engines and overlays it with cognitive vision.
The engine automatically identifies areas of text and tables on the document, mapping them to keys and labels. There are documents that can’t be processed which are covered in the platform exclusions document referenced earlier. Importantly, the system will not be able to extract invoices that are confusing or unclear for humans to read. Examples of such invoices are discussed further down.
Once the platform has extracted and mapped the information on the document, the system will perform a validation check to streamline documents that progress and to act as a safeguard against excluded document types that may have been accidentally uploaded.
What should an invoice include?
Invoices can contain as much relevant information as you and your supplier would like, provided they also contain the information necessary for Zudello to extract and map the document properly. This information is essential for validating the document. We recommend that all uploaded invoices include at a minimum the following information:
At least one stock code or description
Here is an example of a valid and high-standard invoice:
On the left is the original invoice and on the right is the cognitive vision overlay demonstrating how the platform has successfully “read” the document.
If we break down the example invoice, we can see that all necessary information has been included and formatted in a way that the system has easily read.
Examples of problematic invoices:
The two main issues Zudello will have with mapping your invoices are formatting and missing information.
Example 1 – Unclear formatting:
Looking at the original document on the left, it is not obvious that the number floating on the right is related to the line item. As you can see from the image on the right, Zudello has not mapped the floating number as the line total and thus will view this document as invalid, requiring input from the customer using the mapping interface.
Example 2 – No line item totals:
This invoice does not include the totals for the line items listed, meaning the platform will view this document as invalid, requiring input from the customer using the mapping interface.
Example 3 – Hard to understand table:
At the top of the table, the supplier has included labels (description, stock code, quantity shipped, UOM, unit price and total price), however, information has been misplaced.
The site address has been placed in its own row under the column stock code.
The description column contains many rows of information which were all read as separate rows (see the cognitive vision overlay on the right).
Much of this information shouldn’t have been included in the description, such as the reported fault descriptions and a breakdown of costs.
It is unclear whether information within the description such as materials (340.13 in the first instance and 95.00 in the second instance) and labour (157.50 in the first instance and 105.00 in the second instance) were intended to be separate items or not.
The placement of $802.62 on the second last row is unclear as to where it belongs to.
The final row does include a stock code, quantity shipped, unit price and total price, however as this is the first row of the table to include this information and as it is so far away from the headings, they have not been recognised as part of a line item.
This has left the platform unable to recognise what was meant to be included in the table and thus information will not be mapped correctly meaning the platform will view this document as invalid, requiring input from the customer using the mapping interface.
Example 4 – Hard to understand table:
In this example, you can see that the price of the unit cost, the GST and the line total cost of the item are detailed below the rest of the table. With this formatting, it is hard to recognise those values as belonging to the line item. This means it is read as if the line item does not have a total, leading to the platform viewing this document as invalid, requiring input from the customer using the mapping interface.
Example 5 – Text too close together:
This invoice is an example of what happens when text is too close together. To make it easier to see, the problematic text has been outlined in red. As you the text preceding the ABN is connected to the ABN itself and when the system mapped the document, it read the whole line as a single field. The reverse issue occurs when related text is too far apart and the system reads them as separate fields.
Recommended Basic Invoice Format
Our recommended basic invoice format is provided below to avoid invalid documents. Please feel free to share this with your suppliers if you find their invoices frequently have issues or add this to your new supplier guidelines.
Filled in example: