An accounting firm was re-keying a large volume of invoices by hand and looked for a way to process up to 3× more of them in the same time. We designed OCR data capture that reads the details from each document and prepares them for the information system.
OCR reading of scanned and electronic documents
Details are read from every invoice, whether it arrives as a scan or in electronic form.
Configurable templates
The template defines which fields are extracted from an invoice, so the output always has the same structure.
Learning technology
Recognition becomes more accurate over time based on documents already processed, rather than staying at its default settings.
The accounting firm sought to make the manual processing of a large volume of invoices more efficient.
Its concerns were accountant workload, processing time, manual-entry errors, inconsistent data from its customers’ systems and connection to a central system.
After analysing the processes, we selected a solution using OCR, configurable templates and learning technology to speed up input-data processing.
The system can be configured to validate data through a database and export documents in a format supported by the central system; client training and ongoing support were included.
Instead of re-keying details, accountants can focus on checking and working with the data.
Accountants re-keyed details from a large volume of invoices by hand, which took up a substantial part of their working time. Manual entry produced errors and data from individual customers arrived in inconsistent forms, so transferring it to the central system required further work.
Scanned and electronic invoices pass through OCR, which reads the required details according to a template. The data can be validated against a database before export and reaches the central system in a supported format and a consistent structure, reducing manual entry.
The figure of up to 3× more invoices in the same time is the target the solution was designed for.
Configurable templates and learning-assisted OCR turn scanned and electronic invoices into structured data. The extracted data can then be validated and prepared for the target information system.
The solution works with both scanned and electronic documents. Paper invoices are scanned first, while electronic ones enter processing directly. In both cases OCR reads the details according to a configured template, so the output is structured data with the same fields regardless of the document’s original form. That data can then be validated and prepared for the target information system.
Templates define in advance which fields the system should read from an invoice, and they are set up to suit the accounting firm’s needs. Learning technology then improves recognition based on documents the system has already processed. Together they speed up input-data processing and help keep the output consistent, so it can be passed to the information system without further re-keying.
The system can be configured to check extracted details against a database before export. Discrepancies are caught before the data reaches the central system, reducing the errors that used to arise from manual re-keying. Whether and how the check is used is decided during configuration, based on the client’s processes, which we analysed before selecting the solution.
After extraction and any validation, the system exports the document in a format supported by the central information system. The details are stored there without manual entry and in the same structure, regardless of which of the accounting firm’s customers the invoice relates to. This also addresses the original problem of inconsistent data from individual customers’ systems.