Field extraction rates
AP Essentials (formerly ReadSoft Online) Capture Service commits to an average field extraction rate not less than 80%. The average field extraction rate is calculated by dividing all unchanged fields with all extracted fields on all manually verified documents. If a field is unchanged by user, it is considered as correctly extracted. Only fields actively used by customers are taken into consideration.
Depending on the quality and layout of the documents, as well as the verifier’s work, the extraction rate can reach higher than the committed 80%.
Learning process in AP Essentials (formerly ReadSoft Online)
The method used for supplier identification and field extraction is called Knowledge Processing (KP), which is a customer specific database and consists of data from previously verified documents. Every time a document is processed through AP Essentials (formerly ReadSoft Online), learning information is collected and stored in the customer’s Knowledge Processing database. Every supplier record has its customer and buyer unique learning information connected to it. Once stored, the learned information is applied the following time the supplier document is uploaded to AP Essentials (formerly ReadSoft Online).
The learning in AP Essentials (formerly ReadSoft Online) is a continuously ongoing process, constantly refined by every processed document.
A properly recognized supplier is the basis for learning to function, as all learning information is connected to the supplier Master Data record.
First time recognition
When a supplier enters AP Essentials (formerly ReadSoft Online) for the first time, the information in Supplier Master Data is used in an attempt to recognize it. The best match from the Supplier Master Data is suggested. If there is too little information to find a best match, the supplier must be manually selected during verification.
Since first time is only a suggestion, the supplier name field will be flagged as uncertain and must be confirmed by the Office user.
AP Essentials (formerly ReadSoft Online) learning
As soon as the first document for a new supplier is successfully processed, learning data based on the OCR content of the invoice is stored in the customer’s Knowledge Processing database and connected to the appropriate Master Data record. Next time a document is received from the supplier the stored learning data is used to identify the correct supplier name.
When multiple buyers are used, suppliers will be recognized and learned in two possible ways, based on how the master data service is set up:
Master Data on buyer level
When using this setting, AP Essentials (formerly ReadSoft Online) creates a unique learning record for every buyer. Even if a supplier may occur for different buyers, every supplier/buyer combination will have a unique learning record and need to pass through the first time recognition before the learning is applied.
Master Data on customer level
This setting creates one learning record per supplier and is used across all buyers. Once a supplier is recognized for the first time, learning information will be collected and applied on the next uploaded supplier document regardless of which buyer is used.
AP Essentials (formerly ReadSoft Online) uses the learning information from manual verification; position and value given by the Office user, to learn where to find specific fields. The learning information is combined with a snapshot of the field’s surroundings, such as close-by words, to maximize the ability to find the field next time.
Initially, field formats are based on default settings. Once a field is confirmed with another format, the new format will be learned and used on the next document from the supplier. For multiple buyers, every supplier/Buyer combination needs to be learned separately.
Line item extraction
Initially the line item area and fields are defined by using the Draw Wizard in Office. The extraction is improved with learning information after every processed document.
Considerations for learning in AP Essentials (formerly ReadSoft Online)
For a previously unknown supplier at least one batch must be processed before learning can come to play.
The more information you have in Master Data, the better first-time matching. When identified once, the stored learning data is utilized, and the content of the Master Data is less significant.
In order for the system to be confident enough that the selected supplier is the correct one, and set supplier field status to green, at least one of the following master data fields needs to be found on the invoice.
- Supplier name
- Organization number
- Tax registration number
- Telephone number
- Bank account(s)
When Master data on buyer level is selected in the Master Data service, every supplier/buyer combination has its unique learning information and need to pass first time recognition to start collect the learning.
When Master data on customer level is selected, the supplier learning information is on customer level and used across all buyers. First time recognition is only used once per supplier.
The learning information is connected to the supplier Master Data record. If the record is deleted, the learning information is deleted as well and thus the supplier will need to be re-learned again if added.
Once saved, the learning information only affects invoices uploaded after the processing is done. Invoices already imported into AP Essentials (formerly ReadSoft Online) will not be affected.
AP Essentials (formerly ReadSoft Online) is constantly re-learning, always using the data from the last processed invoice.