Ending the ERP Paper Chase?

Keying data from documents into ERP systems is a hidden running cost - but a new generation of data capture software can reduce these costs and improve data quality. Colin Dean of ITESOFT UK takes a look at the issues.

Despite boardroom skepticism, the ERP bandwagon continues to roll. Among the many benefits of ERP is that data need only be captured once. This means fewer key-strokes when handling data compared to disparate legacy systems and fewer data errors as a result. Whilst much data is captured electronically there is a significant amount of information in many organisations that needs to be keyed by hand.

Typically this originates outside of the enterprise and is often communicated via paper documents. Examples include sales orders, supplier invoices, credit notes, delivery notes, remittance advice and every imaginable type of form. And as data entry volumes continue to grow, costs continue to rise and opportunities for savings become more apparent.

Manual data entry is labour intensive and represents a significant ongoing cost for many organisations. For example a typical UK invoice costs just under 5 to process but it is not unusual for the cost to be three times this. For a company processing 100,000 invoices per annum this equates to an annual cost of half a million pounds. In many cases, data entry is being performed by expensive knowledge workers such as mortgage approval managers and is a hidden barrier to productivity.

Paper is Here to Stay

There is clearly a gap between the data-centric world of ERP and the document-centric world of many business processes. Information is the life-blood of an organisation yet documents are often peoples preferred way of exchanging it. 

So if people like paper, how do you bridge this document / data divide? e-Forms and e-Document applications are widespread and are intended to replace or at least complement paper forms. By capturing data at source, an organisation loses the costs associated with manual data entry, improves capture cycle times and eliminates errors due to re-keying but in reality businesses continue to create and consume paper.

A major European automotive supplier decided to implement an EDI program with its suppliers. The objective was 100% EDI transactions. The reality, however, was that only 70% of transactions could be treated in this way, 30% of supplier invoices (1.5 million per annum) could not conform to the proposed financial model and therefore still needed to be delivered on paper. If the corporate might of an automotive giant cannot drive 100% EDI conformity then how can smaller, less structured enterprises be expected to?

Bridging the Gap

On one hand, the path to digital printed output is mature and sophisticated. A choice of quality, high volume, robust, colour-printing systems are available from numerous global vendors such as Xerox, Hewlett Packard and Canon. But on the other hand, the capture of data from documents or forms relies extensively on teams of data entry clerks much as it did in the era of the mainframe. Conventional thinking suggests 4 possible approaches:

1. Re-key the data from paper.
2. Re-key the data from image (a scanned copy of the paper)
3. Pay someone else to re-key the data.
4. Force the originator to use an e-form instead of conventional paper

Capture Solutions Capture Savings

Another, increasingly popular answer, however, is to use data capture and document recognition software to replace or complement manual data entry.  This class of solution will read data from paper using OCR/ICR and upload directly to information systems with greater accuracy and at a fraction of the cost of human operators.

Automatic data capture is not a new technology. It has been around in various forms for over 20 years but it has been unable to shake off a reputation for being unreliable, difficult to manage and prone to recognition errors. Steady advances over recent years, however, have resulted in a new generation of solutions that are both cost effective and technologically robust.

Recognition rates for machine print alpha-numeric text, bar codes and tick boxes are close to 100%. Handwritten numerics are misread four or five times per thousand and non-cursive handwritten alpha text has an average success rate of around 97%. Only cursive (joined up) hand text remains a significant challenge and it may be some time before doctors prescriptions can be read with any real confidence.

Optical Character Recognition (OCR) uses shape recognition techniques and is used to identify printed text. Intelligent Character Recognition (ICR) uses Artificial Neural Network techniques to recognise hand written text.  A number of OCR/ICR engines are normally used in parallel within a weighted voting mechanism to deliver high recognition rates.  A successful capture solution will handle scanning, integration with upstream systems, on-line data integrity checking, image storage and document routing.

In Conclusion

Data capture using OCR embedded within workflows has become a mature and effective technology. Solutions are robust and straddle the gap between corporate data systems and document based processes, offering security, audit, reporting and management functions that fit within corporate regulatory environments - delivering a measurable return that is unique amongst most IT solutions.