What's in the numbers?

Every day tens of thousands of pieces of paper pass through the bank, bearing information on what its customers earn and owe, and who its competitors have as clients. Checks. Banks process reams of them, but until lately have let their potentially lucrative marketing data slip through processors' fingers.

Now banks increasingly are realizing that the imaging software they use to facilitate check storage and retrieval can also be turned to marketing advantage. We're talking about treating the images you capture while processing incoming checks as data stores on which marketing strategy is premised. For instance, the bank with a customer's checking account can see who's getting their mortgage payment while the same bank doing check processing for a major store in town can see where the store's customers have their checking accounts.

"I could pretty much tell if you're keeping a mistress on the other side of town," jests Tommy Green of the data mining capabilities inherent in his state-of-the-art check imaging system.

But the enthusiasm of Green--president of Greenway Corp., a Carollton, Ga., start-up--is not matched by many of his counterparts in banking. Market observers spoke of the "moral barrier" felt by some bankers.

Pat Koster, national marketing co-ordinator with Document Solutions Inc. (DSI), the recognized lead provider of check imaging software says, "there's a fear among banks that they're not supposed to be doing this." Others won't talk because they don't want the competition to spy back on them, he says.

"We never did data mining in a primitive way, therefore we'd never do it with imaging"

STUMBLING ONTO A MARKETING APPLICATION
"Credit card companies have been using information like this for years," Koster says, adding, "now that banks have to compete with nonbanks they have to stop acting like banks and start acting like businesses." Actually many community banks review on-others checks manually to glean insights on bank and nonbank competitors. The application of image technology to this process is new.

Koster is one of many who say the practice--of what we might loosely call the data mining of check images--has gone on only within the past year on an automated basis.

In fact, Greenway says it was unaware of its system's sideline benefit when it was released in August, 1995. Tommy Green says his customers unearthed the data searches PrimeImage was capable of, thanks to system features that were unique (at least at the time of release).

The prerequisites for an imaging system to be used for data queries--that it operates off a relational database and a multi-tasking operating system--are features starting to become standard among check imaging vendors.

Tommy Houston, president of Cabbage Mountain Consultants, imaging specialists in Roswell, Ga., explains why, though it may seem retrograde, relational databases (RDBS) are only now coming to check imaging systems. Check imaging, a five-year-old technology, developed in relative isolation from document imaging, a 15-year old technology where RDBS is common. The significance of a relational database is that it allows users to pull data according to numerous criteria, whereas earlier databases pulled files in strict order. "The old flat file databases were sequential, so to get to record number 27 you had to read the first 26," Houston says.

The other requirement, multi-tasking, is being met as system providers move onto Windows NT. Multi-tasking is required, Houston says, because data querying is an interactive process in which a response often begs a further query to clarify the outcome. Equally, queries cannot be allowed to disrupt the fundamental business of item processing. Hence, the system needs to function on different levels simultaneously.

A popular application of checking systems' newfound capabilities is counter-acquisition tactics by community banks. When a competing community bank is being taken over by a large institution, the other community banks in its market do a data search on the routing number of the takeover candidate. Then they write to its customers, urging them to switch to a community bank.

Greenway, DSI, and Wausau Financial Systems, a provider in Mosinee, WI, all said they had customers doing this. Among those willing to go on record, was Exchange Bank, a Wasau client. Gary Searby, vice-president of operations with the $670-million assets bank in Santa Rosa, Calif., said his bank reacted this way when Bank of America bought Security Pacific Bank, and, again, when Wells Fargo bought First Interstate. In contrast to the huge acquirers, Exchange operates only within Sonoma County.

"We didn't formally track the results, but we did increase our customers after the mailings, which we presumed was as a consequence of them," Searby says.

Exchange has been doing check imaging since May, 1995, and mining the images since mid-1996, he says.

HOW MINING HELPS
With imaging, the same information as before is available, but it has been made more accessible, commentators concur. Referring to microfilm/fiche, the predecessor to imaging, Gary Searby remarks: "What I would do in an hour, before, I can do in seconds with the current technology."

Other banks gain from being able to easily identify social welfare recipients' checks. Having searched the system for a distinguishing code, the banks encourage recipients to take direct deposit of their benefits. A mailer by First Commerce Bank, a $110-million assets institution in Commerce, Ga., persuaded 57 of 700 recipients off its teller line. "This," says Randy Whitehead, head of operations, "saves us a few hours a week of teller time plus the processing of physical checks."

First National Bank of Griffin hopes this year to move half of its welfare customers to direct deposit. Bill Homes, senior-vice president, says the Griffin, Ga., bank processes between 3,000 and 5,000 welfare checks a month.

The $165-million assets bank hasn't yet recouped its $350,000 investment in Greenway's system, but it expects a 25-month payback, Holmes says. Griffin has been on PrimeImage since October, 1995.

Another Greenway client reportedly did a search on big-ticket withdrawals to determine what customers it was losing to a local brokerage.

Others reportedly search to see where their competitors are doing good business in case they too should open branches in those areas.

"A check search is the only option other than picking up the phone book and calling people to ask, Ôwho do you bank with?'," says Tommy Wyatt, vice-president of operations with ABC Bancorp, Moultrie, Ga. ABC will mine its check images to help determine its relative market share in the nine markets where it has branches, Wyatt says.

Local banks know who is desirable but often they don't know which of the desirables they have as customers for what accounts, Wyatt adds. Before ABC began using Greenway software in January 1996, it would have to look at each check on microfilm to weed out its competitors' checking accounts. Now, using data query tools, "maybe it'll take one hour of processor's time," he says. (Off-the-shelf software for relational databases has given nonprogrammers access to the database.)

THE LIMITS OF IMAGING
What imaging software adds to the check review process is a quicker way of looking at checks--on screen rather than by hand or on microfilm--plus a way of whittling down which checks to be looked at. Unlike an undifferentiated pile of a competing bank's checks, the software can isolate those above a certain amount that were also drawn within a certain period.

However, one can query only the data contained in the Magnetic Ink Character Recognition line, not the entire check. As checks go through the check reader, MICR data gets stored as a text file while the check image is stored as an image file. In other words, the check gets stored in two unrelated formats, with only MICR data going into the RDBS from which it can be queried.

A query of the MICR data isolates a set of checks, which the imaging system is instructed to pull up. Then a human being must look at those images to review non- MICR data.

A key omission is the name and address of anyone not already a checking account customer. (Existing customers will already be in a database from which the MICR data could pull their records.)

Handwritten elements, such as the type of payment, also are omitted. Whether the amount of the payment is omitted depends on whether the bank uses courtesy amount recognition software.

Moreover, some banks image only checks drawn on themselves so they obviously cannot mine for prospective customers.

Banks that find pros-pects in their check images have to employ someone to key information from the imaging system into a marketing database because of the incompatibilities outlined above. However, most don't see this as a problem since it costs little.

Imaging providers are working to overcome these limitations. For instance, Wasau has devised a system to make the payment-type information comprehensible to a computer. Instead of writing this information on the check, consumers mark one of 20 icons that represent types of payments. The system can read these as data fields.

Meanwhile, Advanced Financial Solutions, a small provider in Oklahoma City, recently developed a common method of accessing imaged data, regardless of its type. Users of ImageEnterprise also can cut and paste information from the check image into a word processing document.

NO, THANK YOU
Sterling Bank, an AFS customer, is indifferent to whatever mining capabilities AFS might offer because the Houston-based bank refuses to "snoop" on its customers. Glen Rust, executive vice-president, says Sterling is the trusted intermediary for the 80,000-odd checks that pass through its premises every day. Rust, who foresees a consumer backlash to intrusive marketing practices, says, "we never did data mining in a primitive way, therefore we'd never do it with imaging."

The scope of a bank's operations may partially determine whether it values check data mining. First National Bank of Donna, Los Cruces, N.M., says it doesn't need an automated means of tracking competitors. "It's no different from thumbing through a pile of checks," says Bob Johnson, chief financial officer with the $375 million bank. If he wants to review competitors' checks Johnson just walks over to where they are filed, he says. Every bank in town sets apart checks drawn on each of the others and the banks hand back each others' checks at the end of business, he explains.

American Bank and Trust, Coushatta, La., seems equally unlikely to data-mine checks. The bank, an AFS user, (through a service bureau), is an ideal candidate for check mining. As George Drew, senior vice-president says, "the utility company deposits with us so we know where everyone in town holds their bank account." However, the bank doesn't image foreign checks and is relatively satisfied with manually reviewing 20,000-odd checks a day. Drew notes, "We've just created a college fund so we've alerted tellers to watch for checks made payable to universities."

"It's no different from thumbing through a pile of checks"

Likewise, Alpine Bank of Basalt, Colo., does a manual review of checks paid to a local grocery store. This is the only store for which Alpine processes checks and the annual search takes just a few hours of processor time, Glen Jammaron, president says. The 18 western Colorado banks with which Alpine is affiliated are just getting into imaging, but so far they have no check mining plans. At $125 million in assets, Alpine sits on the threshold at which Tommy Green says a bank should invest in imaging.

However, Green goes so far as to question why banks would do third-party check processing in its usual form. "Check processing for big entities is low margin," he says. "It only becomes worthwhile if you can use the data."