In a world where technology lifecycles are measured in months, the automated teller machine seems like Rip van Winkle - it's been asleep for 25 years.

The venerable ATM has changed little, either in form or function, since its first appearance in the 1970s. A customer who used an ATM during the Carter administration would feel quite comfortable withdrawing $40 from a unit some 25 years later. Indeed, the biggest innovation in ATMs over the past two decades has been the dispensing of postage stamps.

Herein lies a puzzle. With an ATM on every downtown street corner, and with the Internet having raised the standards for rapid innovation, these networks should be hotbeds for new capabilities. Unfortunately, almost every aspect of the legacy ATM environment inhibits innovation, from the user interface back through the support systems. It's turned out to be a lot tougher to transform the humble ATM into a 21st-century point of access than most people thought.

Why is this? And can anything be done about it?

For starters, the typical machine is underpowered from both a processor and telecommunications point of view. Most ATM software is based on outdated, proprietary technologies. This means that modifying applications is slow and must be meticulously redone for each hardware manufacturer's platform.

Within the Tower of Babel that is the ATM environment, it's almost impossible to link to a Web site or add a new type of card dispenser.

Look at the systems running behind the cash machine, and the story is just as disheartening. Network protocols are proprietary and inflexible. Given that the networks are disaggregated and owned by banks or bank consortiums, it's no surprise that change moves at a glacial pace.

Security is another problem. Reliance on cards and personal identification numbers, and the lack of human-assisted services, have blocked service introductions. As long as the primary identifier of customers is a card and a PIN, adding products takes the nearly unanimous buy-in of the networks. That slows market acceptance to a crawl. And as long as there is no call center tied to the cash machine for customer service and product underwriting, product options will be limited.

Over the years banks have tried to jazz up their ATMs with everything from video games to advertising. None of these attempts have succeeded. As a result, the ATM industry has focused on driving down support costs with cheaper (i.e., lower-processing capacity) devices and shifting to narrower bandwidth (dial-up) communications, rather than expanding capabilities.

So should banks give up trying to innovate, and figure that the ATM we know today will stick around for the next, oh, 50 years? I believe that innovation is possible through an automated, customer-friendly, cash-dispensing platform. However, the solution is radically different from the legacy network.

The next generation of automation has been installed in nearly 1,000 retail locations under the brand name RPM, for rapid pay machine. They perform check cashing (in essence, cash advances of payroll and other checks) and traditional ATM functions, and offer advertising and targeted content.

The RPM uses an advanced technology platform, combines facial biometrics with traditional PIN-based identification, and is linked to a call center for real-time customer support and product underwriting. By early 2001, several new products will be added to the RPM, including money orders, money transfers, and loans.

Field experience with the RPM suggests that customers overwhelmingly prefer multifunctional ATMs to traditional alternatives. Before installing RPMs at one national grocery chain, the retailer cashed customers' checks for free. After installation (and once the grocer shut down the free service) the average number of checks cashed doubled - even after the machines started charging for the service. Customer surveys indicated that 89% of check cashers felt that the RPM was a better experience than the staffed counter.

Retailers also prefer multifunction ATMs. Whereas a standard teller machine in a convenience store may generate 600 transactions a month, dispensing $60 per transaction, a multifunction check-cashing teller machine generates well over 1,000 check-cashing transactions per month and dispenses $250 per transaction. Cash to the floor (and consequently to the retailers' bottom line) shoots up from around $30,000 to nearly $300,000 per month.

Research from one retailer has shown that RPMs can generate an annual bottom-line impact of $100,000 per store from the combination of increased foot traffic and in-hand cash, revenue share from the devices, and elimination of check-cashing expenses (including losses). Convenience stores typically report a 15% to 30% uplift in sales.

The road to successful ATM innovation requires total overhaul of the legacy environment - from technology and operations to marketing and customer service.

Obviously, a hardware and software upgrade is required. Higher-end processors must be installed and the software platform must shift from proprietary OS/2-based code to flexible operating systems and languages.

Technology isn't the whole story - there are also human factors. Keep in mind that customers do very strange things. RPM users have been observed doing everything from inserting checks into the card-reader to licking the biometric camera. You can't predict what people will do to a multifunction ATM.

The ATM network can be a highly optimized access channel for new financial products and services that provide convenience and value to customers. The big question is whether traditional players can awaken their networks from a decades-long slumber to reap the benefits.

Mr. Rees is executive vice president of business development for Innoventry, a San Fransisco firm owned by Wells Fargo & Co. and Cash America Inernational that sells electronic cash-access systems.

Subscribe Now

Access to authoritative analysis and perspective and our data-driven report series.

14-Day Free Trial

No credit card required. Complete access to articles, breaking news and industry data.