Template OCR solutions had been on the market for a long time. Many of them are already integrated into ERP/DMS systems. In this article we will take a look at the following topics:
- how template OCR solutions work,
- how users are using them,
- why they don’t fit into an ERP.
Template OCR – the basic idea
The idea behind the template OCR solutions is simple. Invoices are semi-structured documents. Most of them are produced by ERP or similar software. There is a finite number of those software solutions. Therefore, there is a finite number of invoice templates. While OCR is already working, we only need to define areas that contain data of interest. To create a template, the user should mark an area on the invoice and tell the computer – this is an invoice number. Afterward, the computer will save coordinates for each area. The next time user will be able to extract data one the same spots with selected OCR template.
Problems in template OCR land
Our user, let’s call him Ben, has bought template OCR software which is connected with his ERP. He has created a new template for his supplier, let’s call it Super company. It was made on this invoice which he scanned himself:
It was scanned perfectly – no rotation in any direction and it was aligned perfectly. He created an OCR template for Super company as marked on the upper image. Therefore, he was able to extract data from the scanned invoice and import it into his ERP system. Later that day, his secretary started to scan invoices. She was in a hurry when scanning the invoice from Super company, so the scan was rotated.
The template used coordinates to define areas for extraction. Unfortunately, coordinates changed since Ben set them – it was not a perfect scan anymore. So, what to do? She created a new template in OCR template software that covered this rotation. A few weeks later, scanning becomes the job of a new employee. He was sloppy so he scanned it like this:
Again, he has to open template OCR software and create a new template to extract data. Now he was able to extract it and import into an ERP.
Invoice data extraction in ERP
Our user Ben needed two software solutions to extract data from invoices. He also has an additional step in his process when adding a new supplier. He has to create templates for its invoices. Not to mention repeated corrections on OCR templates.
The first improvement would be to integrate a templating system into an ERP. You would need to implement all functions that were in OCR templating software by yourself. At some point, it would be easier to write the whole templating system form scratch.
However, our user would still have an additional step in his business process. Is templating really a part of his business? Of course not! Is the data extraction part of his business process? No, but it could automate part of it.
So what is a better solution than a template OCR? An AI OCR data extraction with continuous learning. And the best part about it? No extra steps in the user’s business process. As seamless as it can get.
As an ERP you want to help your customers control their business processes. You don’t want to load them even more. Your customers will love you if boring and time consuming parts of their job will be suddenly automated. They will gladly pay for it.