The Purchasing Insight blog has taken a timely look at the issue of scanning accuracy percentages, and how claims of 99% scanning accuracy through OCR, are misleading.
As the blog post points out, all this means is that 99% of the characters are recognised: it does not mean that they accurately interpreted. It’s what happens after scanning that is the critical step in automation – for example, the ability to recognise and interpret a purchase order number, or a tax reference.
It is only this intelligent capability which is truly meaningful in the context of automation, enabling functions such as line matching. It’s useful to have a reminder that we shouldn’t be blinded by data!
read more at http://purchasinginsight.com/lies-damn-lies-and-statistics-interpreting-accuracy-claims-for-scanning/