Performance metrics

ABBYY FlexiCapture (or the System) extracts data from documents arriving in streams, so performance is measured in volumes processed per period of time. To design the System, define the target performance using performance metrics. The required processing time is defined by internal procedures, service-level agreements, and business processes requirements of a client company. Processing volumes are based on previous data and business development trends, or a company’s business plan. Some occasional or seasonal leaps in volumes may occur, for example because of successful advertising campaigns or the fiscal year-end. These parameters shape the system workload:

• Average batch size in pages
• Image color mode: color, grayscale, black-and-white
• Pages per day (24 hours), average/peak
• Pages per hour, average/peak
• Average document size in pages
• Number of scanning operators
• Number of verification operators
• Document storage time

​

Average batch size

A batch stands for a set of related documents processed together. For example, a customer submits a dozen documents for processing, all under the same request, because crosschecks and business logic forbid their independent processing.

​

Image color mode

Document images come in all shapes and sizes as:

• Scanned copies in color, grayscale or black-and-white
• Photos in different resolutions
• Email attachments, such as vector PDF files

Color grade of document images depends on:

• The ability to control and alter input data. For example, if FlexiCapture clients are chosen for scanning, a company can set the same scanning mode (color grade) for all incoming documents.
• Long-term storage requirements. For example, according to corporate regulations, all documents should be stored for 5 years as grayscale images only. In this case, FlexiCapture clients can convert color images to grayscale images at the scanning stage.

Companies are often obliged to store incoming documents in their original formats. Even so, they can usually estimate what formats to expect and provide sample images. The most expensive scenario is all-color document images, which drive up both network transmission and file storage costs.

​

Pages per day and pages per hour

Average and peak performance are the numbers of pages processed in a period the company chooses, such as 1 hour or 24 hours. Count color, grayscale, and black-and-white pages separately.

• Specify precise time intervals. 24 hours is better than 1 day, which can be misinterpreted as one workday, that is 8 to 12 hours only.
• Make them meaningful to you, so you can easily see whether the system performs according to your needs and expectations.

For example, a better checking point for a customer is the 1000 pages in 24 hours estimate, not 0.01 pages per second. Estimate the processing volume in pages rather than documents, because documents vary significantly in size. At the same time, it is typically easy to guess the average size of documents of one type in pages. For example, an invoice may contain 1 page or over 100 pages, but typically it has 3 pages on average. Finally, convert the figures to the bytes and bits per second commonly used to calculate hardware performance. Use these typical A4 page sizes for each color mode: For a more precise estimate, a sample of actual documents is required. Combine the typical page size for each color mode with the average and peak pages per day or hour. From those, estimate the average and peak input flow in bytes per second.

​

Number of users

This is the number of users accessing the system concurrently while document processing is in progress. There are two types of users:

• Scanning operators scan, check, and edit document images, and then feed them to the Application Server.
• Verification operators verify and revise extracted data, downloading images from and sending corrected data to the Application Server.

​

Document storage time

Document storage time has a great impact on System configuration and hardware costs, because longer storage times require a larger FileStorage. The document storage time within the System is an important parameter; it should not be confused with the document storage time inside the organization. The average document storage time within the System is often the average processing time. Sometimes when a number of processing stages with manual operations are involved, this may be weeks. Sometimes it is the average processing time plus the time for storing images and data at the Processed stage. FlexiCapture treats a document as processed once it is exported to the company ERP system. Processing inside the organization may still be in progress at that point. The document may therefore be re-sent to any of the initial processing stages within the System. For this reason, documents with Processed status (that is, document images and captured data) are stored inside FlexiCapture until:

• They have gone through all business processes
• They are placed in the company archives