Dollars Ex Machina
Before the advent of automated teller machines (ATMs), most people’s experience with devices that dispensed money involved pulling a handle and hoping three cherries would line up. For this reason, perhaps, many people did not trust ATMs at first. In the early days it was common for users to count their cash each time. After all, how could you rely on a machine to dispense the correct number of $20 bills when your photocopier regularly skipped pages and kicked out blank sheets?
Nowadays many customers never count their cash and rarely set foot inside a bank. Whether in San Francisco, London, or Tokyo, all one has to do is insert a card, type in the appropriate code, and out comes the requested amount of dollars, pounds, or yen.
Machines that performed the function of bank tellers began to appear in the late 1960s, though deciding which of them deserves to be called the first ATM is largely a matter of definition. The antecedents of today’s ATMs would either take deposits or dispense cash, but not both. They certainly couldn’t make fund transfers, print out your account balance, sell tickets, or do any of the myriad other things ATMs do today.
One of the first automated deposit acceptors was Luther Simjian’s Bankograph, which New York’s First National City Bank (now Citibank) installed in several branch lobbies in 1960. The idea was to allow customers to deposit cash or checks or pay utility bills and get a receipt without a teller. Bankographs were not burglarproof, however, so their use was confined to normal banking hours. The Bankograph contained an internal camera that took pictures of each deposited bill or coin (checks were recorded and traced by normal means) on 16mm film as it was fed into the machine, also recording the date and time. After about 30 seconds the machine then spit out a strip of photographic prints as the receipt.
After the Bankograph came the cash-dispensing machine. One of the first, perhaps the very first, was introduced in England in June 1967 by Barclays Bank at a branch in Enfield, near London. The dispenser, which was made by De La Rue Instruments of London, initially used a system of paper vouchers, which had to be purchased in advance from a Barclays teller (usually by deducting funds from one’s account). Each voucher was worth £10. They were redeemable at Barclays cash-dispensing machines, which were conveniently located outside Barclays branches. The advantage of this system seems to have been that since the vouchers could not be spent, they were easier to save for emergency use.
It quickly became obvious, however, that paper vouchers were not the way to go. They were easy to counterfeit, and when one got lost or stolen, anyone could use it. So in 1968 Barclays and a few other British banks introduced a second-generation machine that encoded cash on plastic cards, which still had to be bought from a live teller. Punched holes on the card identified the account and encoded a monetary value—£10, 20, 100, or however much the customer asked for. There was no password or other security procedure.
The problem was that each time the customer used his plastic card, the cash machine ate it. The card, which had the purchaser’s name and address printed on it, was sent to the bank’s central processing area, where a clerk deducted the withdrawn amount and issued a new card with the remaining value. The clerk then mailed the card to the customer, a process that could take several days.
This may sound cumbersome, but it proved quite popular. Ian Buxton, an executive with National Cash Register in the United Kingdom, says, “This was truly a folk-driven device. The service came to be known by word of mouth rather than by marketing.” Cash-dispensing machines proliferated rapidly, and three British companies began manufacturing them. They all used punched cards that were read with a matrix of pins. Some had a dot of radioactive paint to mark them as genuine, and some introduced what would be a familiar feature of later ATM cards: the personal identification number, or PESI. By 1969 cash-dispensing machines had spread to Switzerland, France, Sweden, Australia, Brazil, and even Thailand. In the early 1970s similar machines began to appear in the United States and Japan. All were still cash-only machines, not true ATMs.
The next step came from a small Irving, Texas, company called Docutel that had previously specialized in automated baggage-handling equipment for airlines. Its first great contribution was the use of a magnetically encoded plastic card—of the type that is now universal—instead of one punched with holes. The stripe identified the user, the user’s checking and savings accounts, the date of the last transaction, and the bank’s transit and routing numbers. Docutel installed its first such machine, known as the Docuteller, at the Rockville Centre branch of New York’s Chemical Bank in September 1969. Other companies were developing similar machines at the time, but since Docutel made the earliest patent application, the Smithsonian Institution’s National Museum of American History considers Docutel the inventor of the ATM.
To receive cash, the customer had to punch in his or her PIN (which was also encoded on the card) on a keypad. Withdrawals were limited to specific amounts per day. Because the Docuteller was an off-line machine—that is, it had no electronic links to determine how much money was in someone’s account—the withdrawal allowances were based on the bank’s opinion of the customer’s trustworthiness. The best customers were allowed three withdrawals of $50 each per day; others were allowed two or one. Some customers were not offered cards at all. Still, it was a step up from previous cash dispensers, with which you had to pay a teller for the cash before you got it from the machine.
The person responsible for the Docutel cash machine, and ultimately for that company’s ATM, was Donald C. Wetzel. Wetzel led the development effort, and he had a true ATM in mind all along; the 1969 cash dispenser was merely a first step. Wetzel had grown up in New Orleans and begun his career in the late 1940s playing professional baseball in the New York Giants chain. When he realized he wasn’t bound for the majors, he enrolled at Loyola University in New Orleans. After earning a degree in foreign trade, he took a job with IBM selling punch-card computer systems to banks. As he advanced up the IBM sales ladder, he became quite familiar with bankers and how they thought.
In 1968 he was living in Texas and was offered a job at Big Blue headquarters, in Armonk, New York. As he recalled in a 1995 interview with Dr. David K. Allison, curator of the National Museum, “I’d moved five times in the last 12 years, and I decided that I just wasn’t going to do this one.” Instead, through a colleague, he found a job as vice president of product planning for Docutel in Dallas. He was given a mandate to expand the company’s business into financial equipment.
Docutel had been established in May 1967 as a division of Recognition Equipment, Inc. (REI), which sold it in 1970 to a Houston firm called Information Processing Company. As Docutel’s chief planner, Wetzel set up weekly meetings with company executives to kick around ideas for new products. At first nothing seemed right. Then, after Wetzel had stood in line at a bank one day, a light clicked on.
“The idea came into my mind very quickly,” Wetzel recalled in 1995. “I went back to our people at Docutel and told them I thought we could build a true automated teller machine that would perform at least 90 percent—perhaps more than 90 percent—of all the transactions processed by a bank teller.” The idea sounded good to others at Docutel, but it needed more study. “We weren’t close to designing anything or building anything; we were talking functions. Then we had to determine what the economics were. How many of these machines could we sell if the idea turned out to be any good at all? … There was one other early issue that we had to resolve: If we built this machine, and if banks were interested in buying it, would their customers use it?”
The company recruited some master’s students at the University of Dallas to perform a feasibility study. According to Wetzel, “The students came back with a report that basically said that most people would use it, some wouldn’t, but that younger people were more likely to use it and older people probably would not.” Since banks, like most businesses, wanted to attract young customers, the demographics suited them.
When Docutel executives presented the idea to REI’s board of directors, only one board member was against it. That lone dissenter was the only banker on the board. His reaction was not surprising; in the 1960s the officers of most financial institutions were resisting the spread of technology into their businesses. In keeping with the American tradition of small, local banks, a banker typically thought of himself in much the same way as a friendly corner grocer.
One reason bankers disliked computers was that they required numbering customers’ accounts. Numbered accounts were perceived as unfriendly, and they allowed competitors to advertise, “At our bank you’re not a number, you’re a person!” In fact, numbered accounts were inevitable, and the larger and more progressive banks went over early to punch-card computers. Still, the practice made bankers nervous, and with ATMs it would be mandatory.
For similar reasons bankers believed strongly in having tellers deal face to face with customers. Not only did this provide the prized personal touch, but friendly tellers who cashed checks and chatted with customers could (in theory) sell them the same sorts of services that the ATM was supposed to sell to younger customers: savings accounts, loans, credit cards, insurance, and so forth. But as Wetzel explains, “Tellers didn’t usually cross-sell to anybody. The teller’s creed was, ‘You have a check, I’m going to cash it, and then I hope you go away. That way, if you move fast, I’ll get to the next customer, and everybody will be happy.’ ” The few bank tellers with genuine sales ability did not stay bank tellers long.
Toward the end of 1968, after only six months with Docutel, Wetzel persuaded REFs board to fund development of his machine. The board budgeted $4 million toward the effort; this later grew to $5 million. Most of the work was done by a fiveman research team, and Docutel came up with its first commercial model in less than a year. But it was only an initial step.
The engineers quickly figured out that a full ATM would have to be built around four interdependent subsystems: a method to identify the user and retrieve information about his or her accounts; the cash-dispensing mechanism; an electronic logic board to manage the mechanical functions; and a recorder and printer that would keep tabs on all the transactions and issue receipts.
Because the bank computers of the day were fairly rudimentary, the first full-service Docutel ATM, like the cash dispenser, was an off-line machine. It still had no electronic link to a customer’s account. Wetzel and his R&D team knew that such links were coming, though, and Docutel would be ready when they arrived. On-line capabilities were being developed even then by IBM and Lloyds Bank in Britain. But for the time being, each off-line machine would have to stand alone.
The first challenge was to come up with a secure system for identifying the customer, one that didn’t use vouchers or cards that had to be mailed back. The answer was magnetic encoding. Wetzel found a company in Owings Mills, Maryland, named Malco Plastics, that had just started making plastic cards with codable stripes. They contained four tracks, one of which was tailored specifically for use by airlines. The other three were open. Docutel decided to use those three tracks, and a Docutel electrical engineer named George Chastain quickly developed a typewriterlike device that would allow nearly anybody to encode numbers onto them.
At first Malco had problems mass-producing usable cards. The heat required to laminate the thin plastic sandwich tended to warp the magnetic stripe. Eventually, after some adjustments to the production machinery, Malco greatly reduced this problem, but some dud cards continued to be produced.
It occurred to Chastain that anyone with a knowledge of magnetic encoding might be able to duplicate a Malco/Docutel card. He therefore assigned an engineer to make counterfeiting as difficult as possible. The engineer came up with a system that rearranged the bits of magnetic information on the stripe each time the card was used. The code reader was programmed with several different algorithms, each given a letter designation: A , B , C , D , and so on. One spot on the reshuffled stripe told the machine which algorithm to use for that transaction and then changed it to another algorithm for the next transaction.
The second system, the one that actually delivered the cash, did not come so easily. Tom Barnes, a mechanical engineer, tried a number of fruitless approaches to dispensing individual bills and finally took his inspiration from the printing industry. He ended up using tiny suction cups, powered by two small vacuum pumps, to peel one bill at a time out of a cassette that resembled a restaurant napkin holder. Then, as now, only crisp new or nearly new bills would do.
After the vacuum pickups had peeled a bill off the stack, they laid it on a conveyor belt, which moved it past a light sensor to gauge its opacity. If too little light was transmitted, the machine knew that it had picked up more than one bill, so it diverted them to a separate bin and started over. If too much light was transmitted, the machine knew it had picked up no bills. But if the transmission of light was within the acceptable range, the bill would continue on its path and be dropped into a receptacle at the front of the machine.
Today’s ATM dispensing mechanisms work essentially the same way, although rubber rollers or flat rubber belts often replace the vacuum pickups. Some ATMS also have multiple cassettes for dispensing different denominations of bills, traveler’s checks, coupons, tickets, and so forth. Modern cassettes can hold as many as 2,600 bills.
While the off-line Docutel cash dispenser, and later the ATM, had no links to bank computers, each one did contain a small, simple, stand-alone logic circuit. This “black box” read the magnetic stripe on the plastic card, correlated the input PIN with that on the card, read the amount of cash to be dispensed (which the customer typed in on a keypad), told the suction cups how many bills to peel off, and activated the printer so that both the customer and the bank would have a record of each transaction.
Wetzel says that the trickiest and most bothersome part of those early machines was the printer. “We had to print the bank receipt in MICR [magnetic inscription character reading] so it could be read by the automated equipment in the back office. … You know those funny-looking numbers at the bottoms of your checks? Those are in MICR. …
“We had some problems with those early printers. They made mistakes—just simple mistakes like not printing one digit. But that was enough so the bank couldn’t tell who withdrew the money, or when, or from which branch. In those early days, I dreaded Mondays. When the phone rang in the morning, I just knew what was coming. It was almost always a printer problem.”
After Docutel’s four basic subsystems were up and running, there was one more hurdle: the physical security of the machine itself. Wetzel realized that bankers wouldn’t want to risk thousands of dollars in a wall-mounted cash machine without vaultlike security in and around the dispenser. To give bankers confidence, the earliest Docutels were overengineered.
“We built a mechanical box,” he says, “that would have taken somebody eight hours to cut through with an acetylene torch. We had 5/8 inch of stainless steel as the housing. Looking back, we really didn’t need that much, but we thought it would be important to the banker. Bankers understood steel vault doors and safe-deposit boxes.” Docutel also built alarms into each machine.
By the spring of 1969 Docutel had rigged up a prototype, and Wetzel invited bankers to come to Dallas to see it in action. Unfortunately, it didn’t always work, so the demonstrations were often fraught with suspense. One morning five executives from a bank in Atlanta flew in to check out the Docutel prototype. This was on a day when the machine decided not to function.
“I told George Chastain that I would have the bankers come into the conference room, and I’d give them a little history of Docutel, talk about the idea, the concept, automatic tellers, and so on. I left the door open a little bit, and I said, ‘George, periodically you walk up and down the hallway, and if the machine is down, give me a thumbs down. If you give me the thumbs up, I’ll know the machine is working, and we’ll go in for a demonstration.
“Well, I started out. I went through our company history, and George came by twice and gave me the old thumbsdown signal. So I went around again, did this for about an hour and a half. Finally one of the executive vice presidents says, ‘I know more about Docutel now than you do. The machine isn’t working, is it?’ And I said, That’s right.’ He said, ‘That’s O.K. Let’s go look at it anyway.’
“So we went into the lab, and George and I showed them the machine. They understood what we were doing, and they could see we were serious that we were going to build this thing. We showed them how it was supposed to work, and we showed them the guts. The prototype never did work that day. Two days later they called and ordered five machines.”
Even so, Docutel found considerable resistance in the beginning. Bankers had two basic reservations. One was the cost: Why pay $18,000 plus substantial installation, maintenance, security, and promotion costs when tellers were already doing the same job at $10,000 per year or less? Wetzel showed that the machines would more than pay for themselves in decreased staff costs.
The bankers’ second reservation was their fear that customers would be afraid to let a machine handle their money. To answer this reservation, Docutel installed mockups in bank lobbies to show employees and customers what their machines could do and how to use them. Most customers said that if a real machine had been standing there that day, they would have tried it.
The Chemical Bank in Rockville Centre promoted its introduction of the Docutel machine with a newspaper advertising campaign: “On September 3, 1969, our branch will open its doors at 9:00 A.M. , and we’ll never close again!” The machine stood on an outside wall and became an immediate hit with customers. After that, sales to other banks were a snap because they didn’t want to be left behind.
Meanwhile the company kept working on what it called the Total Teller—a true ATM. Docutel sold its first Total Teller sometime in 1971. At about the same time, the Diebold Company (which had shown a prototype ATM at a 1965 banking conference) installed its first ATM at a branch of Citizens & Southern National Bank in Valdosta, Georgia, and Fujitsu installed its first at a Bank of Yokohama branch in Japan. Exact dates are hard to come by, but these three companies were running neck and neck in what might best be declared a three-way tie.
Diebold’s early ATM, designated Model TABS-500, was similar to Docutel’s except that it didn’t use cassettes to store and dispense money. Instead, bills were loaded directly into a hopper—a simple open box—where a suction device peeled them out one at a time. By 1973 an estimated 2,000 automated teller machines were operating throughout the United States, most of them manufactured by Docutel and Diebold. They sold for around $30,000 apiece. Diebold added video security cameras that year and in 1974 launched the next important step: on-line ATMs.
By 1974 Diebold was marketing three ATMs: the off-line TABS-500, an on-line TABS-600, and a TABS-550 that had both on-line and off-line capabilities. Mechanically, the three models were much the same. Where they differed was in electronics.
Pat Green, Diebold’s vice president and general manager of self-service systems, including ATMs, began working on the machines as soon as he joined the company permanently in 1974. He recalls: “On the magnetic stripes on the card, there was a big debate early on about standardization for interchange. Docutel, for example, had a four-track proprietary scheme for recording data on the card. Then the American Bankers Association proposed a different standard called Track 2. We used Track 1, same as the airline industry, because Track 1 was alphanumeric and we could put names and nonnumber codes on it.
“Then the thrift industries came out with a specification for a Track 3 card. You ended up with a lot of debates and arguments about future standards, because at that time you could use Diebold cards only in Diebold machines and Docutel cards only in Docutel machines. There was no standardization.
“So we opted to use the Track 1-2-3 concept, which uses all three tracks, rather than a proprietary scheme like Docutel’s. Turns out we ended up going in the right direction, because the industry did come around to standardize for interchange capabilities, which allowed on-line networking to grow.” Since the mid-1970s most ATMs have used Track 2.
In addition to being linked to a bank’s mainframe computer, on-line ATMs in most countries are interconnected by huge global electronic networks and smaller local ones. These began to be formed in the early 1980s, first as regional networks (like Yankee 24 in New England and NYCE in the New York City area) and then on a nationwide basis. Over time they have proliferated, withered, and merged, though it means little to the consumer, since by the mid-1990s almost all of them could be accessed from almost any ATM. On an international scale the largest networks are Cirrus, Plus, and Star. These networks, which use telephone lines and bounce digital information off satellites, allow you to use your ATM card in distant cities and foreign countries. Your request for cash is routed from the remote ATM to your hometown bank, which debits your account by the proper amount.
Today the world’s largest manufacturer of automated teller machines is NCR, formerly National Cash Register, in Dayton, Ohio. NCR supplies 25 percent of the worldwide ATM market, followed by Diebold at 18 percent and Fujitsu at 10 percent. Olivetti, which bought Docutel in 1982, is among a group of smaller ATM manufacturers that also includes Triton, Tidel, IBM, Phillips, Siemens-Nixdorf, and Interbold, a Consortium of IBM and Diebold.
According to Diebold, there are some 190,000 ATMs in the United States, or roughly one for every 1,300 people. An average of 30 million ATM transactions take place every day. Withdrawals account for 66 percent of all ATM transactions; the average amount withdrawn is $70. Deposits account for another 16 percent, 10 percent are balance inquiries, 4 percent are fund transfers, and the remaining 4 percent are combinations of these categories.
Thirty years on, the bankers’ initial trepidation seems misplaced. Modern customers would as soon spend their lunch hour standing on an old-fashioned bank line to get cash as they would give up indoor plumbing. Yet in the context of the 1960s, the bankers’ reservations made sense, because that era was suffused with fear that machines were going to take over people’s lives. Today we understand much better the ways that machines, by reducing mind-numbing routine for workers and consumers alike, can actually increase the human element in our daily life. It all results from being comfortable with technology. By enlisting technology to satisfy perhaps the most basic human need—money—ATMs have made an important contribution to raising that comfort level.