Put More Automation Into Data Capture

by Adam Throne

Whether paying insurance claims or handling account applications, plenty of mission-critical processes remain stubbornly rooted in paper. That doesn't mean you can't reinvent the way you're handling such forms.

State-of-the-art data capture systems can easily automate almost any business form, and you can skip some of the cumbersome steps companies used to go through - like presorting documents into separate batches by document type before scanning. German Bank SEB, for example, routinely scans as many as 20 different types of forms together; the software does the work of sorting images into the proper queue.

Grayscale and thresholding technologies are minimizing costly rescanning at Complementary Health Care, Portland, OR, by maximizing image quality and cutting data-entry costs with Web-based distributed processing. At the very cutting edge, new systems are for the first time tackling forms that are similar by type but variable in layout and exact nomenclature.

The bottom line in data capture is that it pays to keep up with the Joneses. The latest technologies can make the difference between customer satisfaction and customer defection. If each paper-based transaction represents money in the bank, efficient data capture can make the difference between profitability and disappointing earnings.

Tracking Company Automates Variable Insurance Forms

Insurance is one of those businesses for which paper-intensive processes are still mission critical. In the case of Balboa Life and Casualty, however, the documents in question are not claims or applications. This Irvine, CA-based company provides a tracking service that verifies that insurance is properly maintained on collateral tied to loans. The service is aimed at mortgage lenders, banks and other companies that secure loans against property or automobiles.

To keep track of coverage, Balboa receives proof-of-insurance forms from insurance companies and agencies nationwide. While most forms processing systems are designed to process large volumes of forms of the same design, Balboa has no control over the forms that come in, and many of them keep changing.

"The fact that there are thousands of different form types presented an issue," says David Grundy, Balboa's vice president of application development.

To automate the 30,000 variable forms received each day, Balboa is using AnyForm from Microsystems Technology Inc. (MTI), Tampa, FL. An optional add-on module to MTI's conventional OCR for Forms system, AnyForm is designed to handle forms of the same type that might vary in layout, type style or even terminology. (For example, one form might refer to the "policy holder" while another might say "insured.")

Instead of using a traditional template-based approach that maps out the position of specific data, AnyForm uses a rules-based approach to recognize and extract data from different locations on a page image. MTI recommends using templates and AnyForm in combination to deliver the most efficient solution. Grundy says that in Balboa's case, "20 percent of the form types Balboa receives represent 60 percent of the market. We have templates for the large carriers, but the remaining forms may not be cost effective to create templates for and refine."

For the variable forms Balboa has chosen to automate, Grundy says that AnyForm is fairly effective because the data tends to be consistent (even if the location of that data varies). "About 80 percent to 90 percent of the data in these unstructured forms can be lifted once the rules have been set up," he says. The result is faster, automatic data entry with less manual effort.

Balboa uses MTI's standard OCR for Forms software when dealing with consistent forms. Grundy says he's planning to use the software's latest browser-based capabilities to support distributed validation. This feature would allow the company to use at-home workers and respond to workload fluctuations more flexibly.

German Bank Captures Customer Records

While leading data capture vendors view variable forms as the next big thing in data capture, many customers have yet to make the most of today's proven technology. Automatic document identification, for example, has been around for a few years, yet many companies still spend time and money presorting documents. Automatic ID lets you scan documents in intermixed batches; the software automatically spots forms by type and directs the images into the appropriate work queues.

Late last year, SEB, a European bank based in Frankfurt, Germany, implemented a data capture system with document ID to read account information from more than 10 million customer records dating as far back as 50 years. The bank selected an Integra forms processing system form Top Image Systems (TIS), Tel Aviv, Israel, and Carlsbad, CA.

To complete its massive back-file project by this fall, SEB is scanning more than 80,000 forms per day, many in duplex mode, yielding 160,000 images per day. Once the documents are scanned, TIS's FormID module automatically distinguishes among 20 different types of forms. Data extraction is then automated with Integra's template-driven OCR, ICR, mark sense and barcode recognition technologies. The system also employs database lookups to confirm data and add supplemental information.

By September, SEB plans to augment its system to process incoming mail for branch offices at the bank's centralized facility in Frankfurt. FormsID will help automate 80 different customer forms, including those used to open accounts, file tax statements and handle other day-to-day business.

Wolfgang Weyand, SEB's project manager for document management, says the system allows new form types to be defined during the capture process without disrupting production, and he adds that workflows are flexible in combining automated and manual (key-from-image) operation. "Information that we can't read from documents [with recognition technology] is keyed in manually," he says. "Depending on the complexity of problems, we can find additional information by defining new processing steps."

Adding yet more functionality, SEB plans to add TIS's ePortal module next year, which supports Web-based forms. "Today, a customer has to go to a branch to open an account," says Weyand. "We are going to change the system so they can do so through the Internet."

New York Bank Speeds Pension Payouts

The Bank of New York is another financial institution improving automation and eliminating document sorting with automatic document ID. The New York-based bank processes some 20,000 paper-based securities transactions per day. The bank also handles as many as 5,000 daily requests from customers to withdraw funds from pension plans. These requests (many of which include attachments such as hand-written letters) have to be processed quickly and accurately to ensure timely payment.

The Bank of New York was already using InputAccel software from Action Point, San Jose, CA, to capture the securities transactions with key-from-image data entry. Without upgrades, adding the pension disbursements would have complicated processing, so the bank added automated document ID and data capture modules supplied by Mitek Systems, San Diego, CA.

"Now we can take anything through the scanner, which cuts down on manual preparation," says Joe Gerbino, vice president of technology.

Once documents are scanned and cleaned (despeckle, deskew, and so forth), the images are identified and routed to the appropriate processing queues. From there, Mitek's recognition and quality assurance modules extract and validate the data. If data can't be recognized with OCR or ICR engines, operators revert to conventional key-from-image processing.

Currently, the Bank of New York is automating about 500 disbursement requests per day, and Gerbino says the technology has cut turnaround time for these forms in half. The system still needs tweaking, he says, but he expects to ramp up to at least 2,000 check requests per day.

Distributed Processing Lowers Claims Costs

"Automation" is sometimes a bit of a misnomer for data capture technology because it can't completely eliminate manual data entry and validation. This was the case at Future Vision, a Salt Lake City service bureau that was having trouble with staffing. The company is using distributed processing functionality to make use of lower-cost at-home and overseas data entry workers.

Future Vision specializes in processing HCFA 1500 medical claims for insurance companies. Devin Cox, Future Vision's vice president of marketing, says that HCFA claims are the most expensive forms to process in the insurance business because accuracy demands manual data entry and validation.

With quick turnaround of HCFA claims mandated by the government, Future Vision has automated - to the degree possible, that is - with MediClaim from Cardiff Software, Vista, CA. In February, the company began experimenting with distributed processing capabilities with the hope of cutting labor costs. Remote workers can now log onto Future Vision's systems from anywhere using the Internet, which has opened up new staffing options.

"Everything is done in real time on the servers in our Salt Lake City office," says Cox. "All information is stored on our servers, and nothing is sent out."

In contrast, Cox says many of his competitors send images to data entry operations overseas. "There are now questions about whether federal regulations will allow this," Cox asserts. "We had questions on our part about control - namely, what could we do if overseas workers didn't finish on time?"

With everything running from Future Vision's own servers, Cox says he doesn't have to worry about these issues.

One issue he does face is bandwidth, as Cardiff's distribution approach requires a Citrix server to operate the data capture applications from remote workstations. With a DSL-speed connection, Cox says remote workers can view snippets of images and validate data at production speeds without downloading information locally on remote workstations. Cox says that the Citrix server cost the operation about $10,000 in addition to the $10,000 spent on additional Cardiff software.

"We spent about $25,000 altogether, but it has been worth it to us for the flexibility," Cox says.

Depending on workloads, Future Vision now employs as many as 45 data entry operators, with 25 working from home and 10 available through a subcontractor in China. Cox says this has cut costs by at least 20 percent.

"This is exciting because it allows an organization to tap into a new workforce," he says, adding that there is great interest in the region from mothers who want to be able to work from home.

Better Images Drive Better Recognition

Many think of document imaging as a mature technology, but advances are still being made in fundamentals such as image quality. And the better the image, the more effectively it can be automated and transformed into actionable data.

Complementary HealthCare Plans (CHP), a supplemental insurance provider in Portland, OR, wanted to move from manual claims processing toward full automation.

"CHP was [manually] processing 7,500 claims per month, and that number was doubling every year," says John Amond of Answers for Better Computing, a Lake Oswego, OR, consulting firm that helped CHP implement data capture software.

Amond says CHP selected software from Datacap, Tarrytown, NY, because of the vendor's experience with HCFA forms. The company needed to implement the system quickly because of deadline pressures to handle new claims stemming from new contracts.

The technology quickly slashed manual data entry costs, according to Amond, automating 6,000 of the 8,000 claims now processed each month. As for those still entered manually, Amond says, "Most of that gap involves working with providers to ensure that lines on the forms aren't too close or don't exceed the maximum number of insurance codes. We hope to get up to 95 percent [of forms] automated, though we'll never quite get to the point where we are scanning everything."

CHP was among the first to use Datacap's new grayscale technology. Grayscale scanning captures more information and greater detail in the initial scan, which is then thresholded to bitonal for OCR data extraction. Amond says that poor-quality documents with light text can be scanned once with clear results.

Datacap president Scott Blau says the grayscale scanning system was developed in response to customer demand. "All of our customers complained that if claims were printed on dot matrix printers, many were too light," he says. While scan settings could be changed to address this problem, Blau says this approach often left normal images looking too dark. Light documents could also be sorted out, but it added time and cost.

Grayscale scanning and image perfection has been popularized by Kofax and its Virtual ReScan (VRS) technology, but Blau says Datacap's technology is optimized for OCR data capture.

"In our conversion, we look at only the typed data, while VRS looks at everything on the page," he says, asserting that Datacap's approach yields better recognition rates.

Technology Offers A Better Way

Data capture vendors are working on advances such as variable forms automation, distributed processing and color scanning techniques. Yet there are plenty of more proven features and capabilities that aren't being used - even by companies that have deployed data capture systems.

According to Gartner, U.S. businesses spend an estimated $12 billion each year entering and validating data. In the worst cases, companies are still keying in data from paper by the thousands of forms per day and then storing documents in file cabinet after file cabinet. In other cases, companies are now scanning and storing documents electronically, but they're still keying in data from images rather than relying on recognition technologies.

OCR, ICR and dynamic document ID are now proven technologies that knock time and cost out of your operations. The greater the number of documents you handle each day, the more you stand to gain.