Note: this feature is currently available to Early Access applicants. If you wish to trial it, please email firstname.lastname@example.org.
Some of the most important data fields in any document are vendor name, customer name, VAT IDs, addresses, and payment information. Such fields usually do not change on the documents from the same vendor, unlike dates, total amounts, etc. Moreover, to process the document correctly, you need to get company details exactly right. However, in reality those values aren’t always easy to capture with an OCR reader – or to figure out with human intelligence, for that matter.
The problem with capturing company details
Rossum’s AI captures document data with great accuracy from day one. Unfortunately, not every document is OCR-friendly and some characters might be read incorrectly or not at all. Ask your data entry team, they’ve seen it all:
- invoices might come in terrible scan quality or have strange fonts
- your vendor might put their logo but not their name on the purchase order
- a vendor might use different names on different documents, or omit their name from the documents completely
Unfortunately, in a manual data entry process, a human operator is forced to spend extra time looking up more information to validate documents. In an AI-powered data capture process, you want to automate this step.
How Rossum can help
Rossum’s data matching extension will intake a master data file containing information about your vendors or customers. When capturing data from a new document, Rossum will look up the data from the processed document in the master data file to find the best match and identify the company precisely. This powerful tool helps speed up and even automate document validation without taking up any development resources.
Let’s set this up in Rossum
To access the setup page, go to https://data-matching.elis.rossum.ai and log in with your Rossum email address and password. The portal will take you through the following steps.
Step 1: Upload your file.
- The file format is important, so please download one of the sample files from the prompt and use the same column titles (or attributes in JSON or XML) and the same order of the columns to structure your own file. That is how Rossum will know what data it should use for matching.
- After choosing the master data file, select the encoding of the file, and the name of the Rossum workspace where the company matching should apply. Click Upload.
Step 2: Create matching logic.
- In the Create matching logic tab, select the queue where the logic should apply. The settings pictured below will appear.
- Select the fields you see in Rossum that will be used for matching against the master data. Currently the add-on supports up to 8 fields for matching (defined by their schema ID in parenthesis): Vendor Name (sender_name), Vendor Address (sender_address), Vendor VAT ID (sender_vat_id), Customer Name (recipient_name), Customer Address (recipient_address), Customer VAT ID (recipient_vat_id), IBAN (iban), and Bank Account (account_num). If you see fewer fields, like in the screenshot above, check your schema and make sure the fields are ON with the blue checkmark.
- Choose if you want Rossum to look for an exact match or for a fuzzy (close enough) match.
- Click on the Plus icon to add another rule. The matching process will apply the rules following the order you set up. With multiple rules, even if Rossum can’t find the name or the VAT number on the document, it could still successfully extract and match the IBAN, and identify the company from your file.
Step 3: Choose UX behavior.
- Choose what happens if the match is not found. Rossum could show an error and block the Confirm button until the data is corrected – or it could do nothing and allow the user to act at their discretion.
Step 4: See data matching in action.
- Once you open Rossum’s validation screen, you’ll notice a new field on the left called “Company match”. This field is a dropdown which will pull the value from the master data file.
- To pull in the company match value, the matching extension will compare company details captured from the document to the master data, following the rules you’ve just set. If the captured data matches the data on the master file, the dropdown will show the matched company name from master data.
- If some of the company details fields weren’t captured correctly and, as a result, the match wasn’t found, you can select the field and place the bounding box over the appropriate value on the document. Filling in a proper value will immediately trigger the extension to try and look up the matching company name again.
- Voila! Company identity is now verified, you can confirm the document and the correct company ID will be passed to the downstream system.
If you want to sign up for Early Access and try this out, email us at email@example.com.