According to Billentis, up to 85% of all invoices in the UK were issued on paper in 2016. With billions of paper documents entered into accounting records, making data entry more efficient is critical for all businesses. Here’s how OCR and machine learning can help you ditch manual data entry for good.
Manual data entry has been around for ages
Rewriting documents manually dates back to the pre-printing times when it was the only way to produce additional copies of any written material. It’s fascinating that the job of today’s data entry specialists hardly differs from that of monastic scribes from long ago. The switch from quills and parchments to keyboards and monitors is the only change in process which is essentially the same.
Despite the fact that manual data entry is such an outdated business process, it’s still widely in use. Many companies rely primarily on human labour for inputting data from documents rather than applying suitable technologies. But do these technologies even exist?
Manual data entry issues
There’s a large number of issues related to manual data entry that make it one of the business fields most in need of improvement. A solution that boosts the speed and accuracy of how bookkeepers process documents is just the technology that is needed.
Even a single error can have huge consequences for your company. A well-known 1-10-100 rule says that it costs $1 to verify a record, around $10 to fix it, and approximately $100 for subsequent problems arising if you do nothing about that error. As a rule of thumb, the error rate for manual data entry is 1%. Of course, you can reduce the error rate to a negligible level by allocating more resources for additional verification.
- The volume of documents processed
The number of documents processed manually is limited, and it varies. The average typing speed for a typist is around 50-80 words per minute but you can’t expect anyone just to sit and type continuously for the whole day as it’s a mundane and laborious task. To make things worse, average productivity graphs show that your data entry specialist is likely to be fully productive only during specific periods of the day. As a result, you may expect that the pace of document processing will vary depending on many factors.
- Data integrity
Many issues around data entered manually are related to inconsistent naming conventions, misunderstanding of data and misinterpretation of what data means. In other words, manual data entry needs a lot of costly and time-consuming verification to ensure data integrity.
Paper invoices dominant in Europe
While the EU allows businesses to make transactions entirely without any paper documents, it’s rarely the case in most countries.
According to the E-invoicing Report 2017 released by Billentis, the share of paper invoices in both B2B and B2C transactions is as high as 60-85% in most EU countries, including the whole EU5 (France, Germany, Italy, Spain, and the UK).
For central European markets, such as Poland, Czech Republic, or Hungary, paper invoices are even more dominant, with more than 85% share in all transactions.
With billions of paper documents issued every year, some invoice and receipt data entry is inevitable for accountants and bookkeepers in the foreseeable future. But what’s even more important is that moving to e-invoicing won’t solve the data entry problems that companies face today.
When businesses receive e-invoices, data from PDFs, emails or image files still must be entered into accounting systems, such as Quickbooks or Xero. Copying and pasting are probably faster than rewriting but prone to human errors and data integrity issues as well.
OCR and machine learning for data entry
While the advantage of machines regarding speed and cost is undeniable, they also make mistakes. It happens primarily when they work based on human-created input material, such as invoices and receipts.
OCR is a technology for converting paper documents or images into machine-encoded versions that can be edited using computer software. While OCR has been around since before computers, its accuracy is far from ideal, especially for bookkeeping. Read this guide to learn what OCR is and what it needs to read business documents.
Even a single digit error can cause a lot of damage.
The solution is systems that read invoices and then learn based on how documents are verified and booked into an accounting system. This is exactly how arbitrue solves issues around manual data entry.
Why not see for yourself how arbitrue can help you speed up your expense document processing using machine learning and OCR. Just sign up following this link and start using arbitrue at no cost.