To understand the various operational and organizational problems, we must first understand what it was like to initiate, authorize, clear, and settle transactions in the BankAmericard system of 1968. This is best done by walking through the process of a typical domestic purchase. Note that in this example, we will discuss only the details that will help us understand the specific operational and organizational problems faced by the BankAmericard system; other interesting but less relevant details will be examined in later chapters. This example is also the ideal case; the unfortunate realities of the process will be noted in the following section.2 Imagine yourself in 1968, holding a shiny new BankAmericard. As discussed in the previous chapter, all cards had the same blue, white, and gold bands across the face of the card so that merchants could easily identify your card as acceptable, regardless of which bank actually issued it. Merchants also hung signs with the same marks in their windows, so that you could easily identify those that accepted the card, regardless of which bank represented the merchant in the system.3 You spy a merchant that you need to visit, select your items and present your BankAmericard for payment. If your purchase amount is below the merchant’s floor limit the merchant can complete the transaction immediately without authorization.4 The floor limit varied by merchant type and by card type: some cards had a star on the front, while others did not.5 The floor limit for a general merchant was typically $50 for a non-starred and $100 for a starred card, but airlines, hotels and other services were often granted higher floor limits.

If your purchase is above the merchant’s floor limit, the merchant is required to call for authorization. The merchant dials the acquirer’s authorization center and verbally conveys the transaction details to the authorization operator. The autho- rizer first determines if the card was issued by the same bank or another by looking at the first four digits of your account number. If it is the same bank, the transac- tion is known as local or on-us, otherwise it is known as an interchange transac- tion. If this is an on-us transaction, the authorizer then consults a series of printed re- ports to determine if the transaction should be authorized. At this time, there were no interactive computer systems with CRT terminals installed at the BankAmeri- card authorization centers. When your bank became a BankAmericard licensee, it did receive some “computer software” from the Bank of America, but this was just a simple punch card-based accounting system. This system produced two reports to help the authorizers: a list of known hot cards, which were either stolen or on hold for some other reason; and a summary of each cardholder’s account, listing their current balance, credit limit, purchase and payment history. The authorizer first searches through the hot card list to ensure your account number does not appear there. Then the authorizer manually wades through the massive binder of account sheets to find yours, reviews your details, and consults the hand-written list of au- thorizations already given since the report was last printed. If all is in order, the authorizer gives the merchant an authorization code, consisting of a few letters and digits, and the merchant writes that on to the sales draft. If this is an interchange case, however, the merchant’s authorizer does not have access to your records and is thus required to call or telex your bank’s authorization center. The authorizer puts the merchant on hold, dials your bank’s center and relays the transaction details. Your bank’s authorizer then consults the same type of reports already discussed, and supplies an authorization code. The original authorizer then relays this code to the merchant. At this time, interchange was rare on the average, but there were localized exceptions to this. For example, the National Bank of Commerce in Seattle and Puget Sound National Bank in Tacoma experienced a high level of interchange due to the large amount of business that takes place be- tween those two cities, which are roughly 30 miles apart. The authorization centers at these two banks simply called each other in the morning and kept a line open, allowing them to authorize interchange transactions quickly over a speakerphone After authorization, the merchant then completes the sales draft. The draft is a multi-layer document: the top two layers are like tissue-paper, one for you and one for the merchant. The bottom layer is an IBM 80-column punch card, com- plete with the corner notch. The merchant puts your card and the sales draft into an imprinter, informally known as a “zip-zap machine,” which squeezes the em- bossed characters on your card against the sales draft, thereby transferring your card number, expiration date and name onto each layer via carbon paper. The imprinter also holds another embossed plate containing the merchant’s details. The merchant manually adds the transaction date and purchase amount to the draft, and you sign it to complete the purchase. The merchant is then required to check the signature on the card against your signature on the draft to ensure that you are the proper cardholder, but few do. The merchant tears off the cus- tomer copy and hands it to you, putting the other two layers in the cash regis- ter. On a regular basis, the merchant deposits the punch card layer just like a check. Unlike a check, however, the merchant receives an instant credit, less the discount, the amount of which is negotiated when the merchant signs the contract with the bank (merchant discounts at this time ranged anywhere from 0 to 8 percent, averag- ing 3.5 percent).7 From the merchant’s perspective, the transaction is now complete, but the clearing and settlement process has in fact only just begun. Although the drafts are computer punch cards, they are not yet machine-readable. Banks with very low volume may just manually sort and total the drafts, but others send them to the proofing and data-entry departments to be manually key punched and proofed.Proofing involves verifying that the drafts total to the same amount claimed by the depositor. This was often done by encoding the human-readable elements of a draft into machine-readable form, so that the drafts can be machine-totaled. The drafts are then sorted by card number. On-us transactions are fed into the computer to update the cardholder accounts, and are then added by collation to the drafts already processed for each cardholder since the last billing cycle. At this time, most banks are still performing country-club billing, where the physical drafts are included with each statement. All interchange drafts are then grouped and totaled by issuing bank. The mer- chant’s bank completes a special clearing draft against the issuing bank for the total of all the sales drafts. The clearing draft looks very much like a cashier’s check, complete with the magnetic ink routing characters, and can be submitted through the normal checkclearing system for payment. The physical sales drafts on the other hand are mailed directly to the issuing bank through the US postal system. The clearing draft is often processed before the individual sales drafts arrive at the is- suer, so the issuer is forced to transfer funds, but must wait until the sales drafts arrive to reconcile and add the charges to the relevant cardholders’ accounts. Once they arrive, the issuer reconciles the sales drafts against the settlement payments, and then performs the same actions the original bank did for the on-us case.

Operational Problems Within this simple transaction scenario, we can begin to see a number of operational problems that were greatly exacerbated by the system’s increasing scope and sales volume. Authorization, Floor Limits, and Fraud The first notable operational problem was the interaction of authorization, floor lim- its, and fraud. Payment card transactions differ from those in other payment systems in one important way: they are guaranteed. If the merchant follows the rules of the program, the merchant is guaranteed payment, even if the transaction was fraud- ulent. In the case of a personal check, the issuer simply returns the bad checkand the merchant must absorb the loss; in a payment card transaction, the issuer must absorb the loss. This introduces a certain amount of risk to the issuing bank, and in an ideal world, the issuing bank would like to eliminate that risk by authoriz- ing every transaction. This was not a realistic option in 1968, however, as the la- bor and telecommunication costs would easily outweigh the revenue gained from a low-value transaction. Additionally, authorizing every transaction would delay an already slow process, risking the use of cash or a checkinstead of the card. The floor limit concept is essentially a cost/risk tradeoff made by the banks. Not all transactions are equally risky, and the easiest way to distinguish the higher-risk ones is by the combination of purchase amount and merchant type: a high-value pur- chase from a jewelry store is more risky than a low-value purchase from a shoe store. What most banks did not anticipate, however, was that criminals would quickly dis- cover the various floor limits and make numerous under-limit charges, resulting in significant losses. A new card stolen from a mailbox could be used for a week or more before the issuing bank even saw the first sales draft, and over a month be- fore the cardholder received the first statement for a card the customer did not even know was issued.10 Once detected, banks would notify other authorization centers and mail a postcard to merchants that might likely see the card.11 But relying on the merchants to catch the cards was problematic. The main in- centive for merchants to use the authorization system is the guarantee of payment, not the reward for catching a stolen card. The authorization process is more than just a technical function—it also formally transfers the responsibility for fraud from the merchant to the issuer. A merchant was (and still is) allowed to take a transaction above the floor limit without authorization, but the merchant then assumes the risk of fraud. If an issuer can prove that the merchant did not authorize the transaction, or that the bank warned the merchant about the card number prior to the transaction, the issuer can submit a chargeback into the system, which will eventually debit the merchant’s account. Proving a chargeback required a manual audit, however, and most bankcard processing centers were already struggling to keep up with the sharply-increased sales volume. 10Often the cards were actually stolen by the postal sorters and carriers. The practice of mailing unsolicited cards to consumers was eventually banned by the US Congress in 1970, and most other countries have since passed similar laws. 11Jutilla (1973), pp. 221–223. Eventually the Visa system produced a weekly booklet of hot card numbers, but this was ultimately replaced by online authorization via inexpensive point of sale dial terminals (see Chap. 7). 34 2 Associating: Dee Hock and the Creation of the Organization Merchants were also not inclined to call for authorizations due to the delay it would cause at the point of sale—sources from the time estimated that the aver- age authorization took anywhere from five to twenty minutes, depending on how quickly the merchant could get through to the authorization center, and how quickly the merchant’s bank could call or telex the issuing bank in an interchange case.12 Stallwitz found that nearly all merchants in his study complained about the speed of authorization, and some admitted that they encouraged the use of cash or a check- when the purchase was above their floor limit.13 Others would rely on their own assessment of the customer (often based on appearance) and take the card without authorization, or simply reuse an authorization code from a prior transaction as it was unlikely that the issuing bank would detect this under the manual system of the time.14 Stallwitz also found that suburban merchants in particular would avoid consulting the hot card lists and calling for authorization as it might offend their customers and risk the loss of the sale. Lastly, some merchants were themselves creating or participating in fraudulent transactions. Restaurant cashiers would make additional sales drafts with a customer’s card, or less reputable merchants would submit under-limit drafts using a stolen card and split the proceeds with the thief.15 The actual amount of fraud occurring at this time is difficult to estimate as banks were not required to disclose such information, nor were they particularly eager to do so. Those that did were either inconsistent in the way they calculated and reported losses, or as Spencer Nilson claims “doctored the records so that it would come out to a ratio acceptable to their peers.”16 Nevertheless, Nilson and others attempted to estimate how much the banks were losing on their card programs. Unfortunately, the estimates are difficult to compare as they are for different time periods, different sets of card programs (e.g., bankcards only, bank and T&E and retail, etc.), and different loss categories (total losses as opposed to losses specifically attributable to fraud). Nilson estimated that fraud-specific losses on bankcards increased from a mere $140,000 in 1967 to $2.2 million by 1969.17 Various Federal Reserve studies reported that total losses for bankcards rose from $12 million in 1967 to $115.5 million in 1970.18 Nocera claimed that throughout the late 1960s, the Chicago banks alone lost over $25 million, and the New York banks over $250 million.19 12“It took about 15 to 20 minutes to make a $35 purchase, which didn’t make you very popular at the point of sale” (Russell interview). See also Stallwitz (1968), pp. 44–45. 13Stallwitz (1968), p. 45. 14Reusing authorization codes became much easier to detect after NBI computerized both autho- rization and clearing and settlement, the story of which will be told in the next two chapters. 15Jutilla (1973), pp. 219–229, Nocera (1994), p. 30, Galanoy (1980), p. 149. 16Nilson (11 April 1977), Report No 161. The general accuracy of the Nilson Report was contested by many of my interview sources, so some of his claims and statistics should be approached with caution. 17Nilson (11 April 1977), Report No 161. Dollar amounts are in USD. 181967 data from (Federal Reserve System July 1968); 1970 data reported in Brooke (18 May 1971). 19Nocera (1994), p. 61. Problems in the Licensing Program 35 The growing amount of fraud was clearly a concern for those banks participating in the BankAmericard licensing program. Beyond the actual monetary losses, the shocking headlines were creating a perception that fraud was rampant and bankers were doing nothing to protect their cardholders.20 This perception could not only erode the confidence of cardholders and merchants, but also attract the unwanted attention of lawmakers and regulators. Indeed the US Congress held hearings on the practice of mailing unsolicited cards in 1967 and was drafting legislation to not only prohibit it, but also protect consumers from the cost of fraudulent charges.21 As is typical, these hearings became a thinly-veiled public trial of the entire bank credit card industry, accusing the banks of fueling inflation and tempting innocent consumers to abandon the traditional values of thrift in favor of reckless debt spend- ing.22 Clearing and Settlement of Interchange Transactions The second major operational problem area was the clearing and settlement of in- terchange transactions. Like a check, a payment card sales draft is a claim on funds that must be cleared and settled with the issuing bank. If a different bank would acquire that transaction, there would have to be a mechanism by which the draft can be routed to the issuer, and payment made to the acquirer. As noted in the previous chapter, most banks at this time cleared and settled their checks through the national clearinghouse operated by the Federal Reserve. It would seem that using this same system to clear and settle credit card sales drafts, which were small in number compared to checks at this time, would be a sensible thing to do. The bankcard associations approached the Fed about processing credit card drafts, but the Fed refused to handle them.23 Technically, it would have required 20For example, see Galanoy (1980). Formerly the Director of Communications for NBI, Galanoy accused bankers of being blinded by their desire to build an all-encompassing electronic funds transfer system, ignoring the costs of fraud to consumers. For an example of this concern voiced in the popular press, see O’Neil (1970). 21These laws were passed in 1970 as an amendment to the Truth in Lending Act (Brandel and Terraciano 1980; Fisher et al. 1980, p. 257). The 1967 hearings are documented in 19th Congress, First Session (8 and 9 November 1967). 22Dee Hock provided perhaps the best rebuttal to this in a 1979 interview: “Sure, consumer debt is high, but if you want the consumer to stay out of debt, business and government have to set the example. If we expect consumers to reduce debt and increase savings, then we must create an environment without inflation and with tax laws that favor saving and not debt. After all, interest paid on debt is tax deductible and interest earned on savings is taxed. How can that encourage thrift? We should not criticize the consumer who is learning to play the government invented game of buying now through debt and paying later with inflated dollars.” Streeter (1979), p. 75. 23Russell interview. Hock also commented on this in a 1974 speech: “Had the Federal Reserve agreed when asked (and they were) to clear bank card activity, would the service have evolved as it subsequently has? . . . It is clear there would be no BASE II and no INAS today had the Federal Reserve said yes, and clear that present bank card service would be radically different” (Hock 1974, p. 21). 36 2 Associating: Dee Hock and the Creation of the Organization some modifications to the automated systems: the sales drafts were 80 column IBM punch cards, larger in size than most checks of the day; and they encoded infor- mation as punched holes instead of magnetic characters printed along the bottom edge. But the technical reasons were secondary to the more ideological belief that debt instruments, especially those involving a discount, simply did not belong in the Federal Reserve’s clearing system.24 Recall that one of the Fed’s goals was to eliminate discounts on cleared checks, so it is not surprising that they would refuse to process BankAmericard transactions. With the Fed’s refusal to handle credit card drafts, the BankAmericard Service Corporation (BASC) was faced with a problem: how should the licensee banks clear and settle their interchange transactions? One logical option would have been for the BASC to create their own centralized clearinghouse for BankAmericard trans- actions, but the BASC chose not to do this, partly because the amount of interchange was still very low in the late 1960s.25 Instead, the BASC stipulated that acquiring banks must mail interchange drafts directly to the issuing bank, similar to the way they handled out-of-town checks in the nineteenth century. The issuing bank would then reimburse the acquiring bank, less a discount fee, called the interchange reim- bursement fee. 26 This solved only the clearing half of the problem—the licensee banks still needed a way to settle those transactions (i.e., transfer “good and final funds” from the issuer to the acquirer). Recall that the Federal Reserve System eliminated the need to transfer physical currency between banks when settling payment transactions, and the BASC decided to leverage this system by creating a special clearing draft, which looked like a bit like a cashier’s check. To receive payment for a set of interchange drafts, acquirers completed one of these clearing drafts against the issuer for the total amount of the sales drafts, less interchange fees, and submitted it along with their other inter-bank funds transfer requests. This separation of the clearing draft from the sales drafts allowed banks to use their existing funds transfer mechanisms, but it also created a timing problem that jeopardized the functioning of the entire system. When the issuing bank received payment notice of the clearing draft, it would enter that amount into a suspense ledger and wait for the individual sales drafts to arrive in order to reconcile and bill the cardholder. Unfortunately, this often took quite a long time. This is how Visa’s founder described it: Meanwhile, the merchant bank, having already been paid and under immense pressure to handle its own cardholder transactions, had no incentive to process [interchange] transac- tions and get them to the issuing bank for billing to the cardholder. Since each bank was 24Russell interview. 25Sources estimated that it was between one and five percent of transactions at the most. There were of course localized exceptions to this. In regions where banks were not allowed to operate branches across an entire metropolitan area, the interchange level would naturally be higher. 26Note that the laws governing checkclearing discussed in the previous chapter did not apply to credit card sales drafts. Any similarity in their clearing method was coincidental and not required by law. The legal basis for credit card sales drafts came from the contracts signed by licensee banks, cardholders, and merchants (Katz interview). Problems in the Licensing Program 37 both a merchant-signing bank and a card-issuing bank, they began to play tit-for-tat, while back rooms filled with unprocessed transactions, customers went unbilled, and suspense ledgers swelled like a hammered thumb. It became an accounting nightmare.27 This immense backlog in the system also compounded the fraud problems dis- cussed earlier. Issuing banks would have no way of knowing if sub-floor-limit fraud was occurring on a card until the actual sales drafts arrived and were processed. By the time they arrived, thousands of dollars worth of fraud could have taken place. Even when the sales drafts did arrive, it was often the case that their total did not match the clearing draft amount. Many smaller merchant banks would simply run an adding machine tape over the drafts instead of key punching them, and would inevitably make mistakes. Chuck Russell, who succeeded Hock as CEO, recalled that “Banks couldn’t balance from day to day because they couldn’t get their drafts drawn on other banks settled. It was a disaster.”28 To provide a sense of the scale of the problem, he relayed this story: I was shown a room that was warehoused-sized, full of IBM 80-column tab cards (which were the drafts) that they couldn’t settle. We’re talking millions and millions of dollars … they had never got the debit or the credit side of the transaction through clearing because they couldn’t find them!29 Finally, it should be noted that not all banks experienced problems to the degree described here. But the lack of a centralized clearinghouse, compounded with the timing problems introduced by the clearing drafts, created operational problems that were most definitely threatening the overall system’s stability and impeding its future growth

Organizational Problems Although the operational problems just described may have had potential solutions within a health franchising organization, the organization had problems of its own that further compounded the operational difficulties. It was these organizational problems, even more than the operational ones, that convinced the licensees that a new organizational structure was necessary. The BankAmericard licensing system, like any cooperative payment system, faced a central organizational tension—balancing competition and cooperation.31 27Hock (2005), p. 77. 28Russell interview. 29Russell interview. The “stacks of unprocessed drafts” story was also relayed by others in various forms. 30Jutilla indicated that his bank was typically able to reconcile, but the delays in receiving the interchange drafts were especially dangerous due to fraud. He concluded that the system could not have survived the way it was as the transaction volume increased. 31Evans and Schmalensee (2005). 38 2 Associating: Dee Hock and the Creation of the Organization The licensing system created a new meta-organization comprised of competing fi- nancial institutions that needed to cooperate, at least to some degree, in order to pro- vide a universal payment system that none could have realistically provided alone. Competing organizations in a marketplace normally seek their own self-interests in an assumed zero-sum game for market share. A cooperative organization, on the other hand, offers a different possibility—if all members cooperate, they can pro- vide a larger, universal system that allows them all to benefit even more than if they chose not to cooperate. In other words, each participant’s slice of the cooperatively baked pie would likely be larger than any pie the participant could have baked alone. To accomplish this, however, they need mechanisms that would create trust within the organization, mechanisms that balance out their power and interests and dic- tate how inter-organizational work will be accomplished. In other words, they need something akin to a constitution, as well as operating regulations, to which all mem- ber organizations agree. As we shall see, the licensing program’s key organizational problems lay precisely in these balancing mechanisms and operating regulations. Under the BankAmericard licensing system, BofA retained not only the owner- ship of the BankAmericard name and marks, but also all the power, and this led to a fundamental distrust between BofA and the licensees. The licensees knew that BofA would have opened branches in their territories if the banking regulations had allowed it, and if those regulations ever changed, BofA could easily revoke their license and become the sole BankAmericard issuer.32 The licensees also doubted if BofA had the desire and even the ability to solve the operational problems dis- cussed earlier.33 The licensees believed that any solutions developed by BofA would naturally be in BofA’s best interest and not those of the licensee banks. Although BofA retained nearly all the power in the system, their power to en- force and modify the operating regulations was neutered by two critical flaws in the license contracts. First, the contracts lacked mechanisms for financially punishing banks that skirted or bent the operating regulations, nor did they contain a method for resolving grievances between the licensee banks. The only recourse BofA had was to revoke a bank’s license, but since most of these banks held large correspon- dent deposits with the BofA, and were dominant in their geographic area, this was not likely to happen. Second, the contracts also lacked a clause allowing BofA to change the operating regulations in response to new developments. If BofA needed to modify or add a rule, they had to re-negotiate a new contract. Again, BofA had no recourse if banks simply refused to sign the new license, which they often did if the rules were not in their best interests.34 The fundamental distrust and the flaws in the contracts created a number of orga- nizational instabilities. The most significant and pernicious was the tension over the 32Hock (2005), p. 85. Of course, these regulations were abolished in the 1980s, but by then it was too late, as the Visa system had already been established. 33Russell interview. The BofA paid very low salaries at the time, and the most talented operational people tended to go to their main local competitor, Wells Fargo, which was a member of the Interbank system. 34Katz interview. See also Hock (2005), pp. 83–87. Dee Hock 39 interchange reimbursement fee. As noted earlier, this fee was paid by the acquirer to the issuer during the settlement of an interchange transaction.35 At this time, the intent of the fee was to compensate the issuer for the cost and risk of extending the cardholder credit for the transaction. The rule established under the licensing sys- tem for interchange fees was essentially unenforceable. This is how Bennett Katz, Visa’s long-time general counsel, described it: When I came on board, the rule was. . . if a customer of your bank goes into a merchant belonging to another bank, outside of that territory, then the bank that signed the merchant has a choice as to what it sends to the issuer. It could send the amount of the discount that it received from the merchant less a processing fee (for processing the transaction), or if it didn’t want to calculate each and every one. . . it could send the average discount it was getting from all of its merchants less a processing fee. Well they would say ‘my average is two percent.’ How are you going to audit that? And if the merchant put up a big deposit, their merchant discount might be close to zero, and the issuer would get almost nothing! So the issuer has all the costs because he’s extending the credit and eating defaults, but he was getting almost nothing when the customer traveled. The losses were horrendous. It was literally chaos in the BankAmericard system.

In October of 1968, the BASC called a special meeting of the licensees to discuss the operational and organizational problems facing the BankAmericard system. Card program managers from each of the licensee banks descended on Columbus, Ohio, but the BASC neglected to send their most senior officers. The licensees were in- censed that the BASC apparently did not recognize the seriousness of the situation, and began to make accusations that the BASC was either unwilling or incapable of solving the system’s problems. By the middle of the second day, the meeting had devolved into “acrimonious argument.”37 Unsure of how to rescue the situation, the BASC representatives attempted to create a committee of licensees that would look into the most critical problems. One of those selected to be on the committee, how- ever, had a different idea of what it would take to solve the system’s problems, and after lunch the rest of the licensees were greeted by the card-center manager from the Seattle National Bank of Commerce: Dee Ward Hock.

Dee Hock had been slowly coming to the realization that “money” had become nothing more than “guaranteed alphanumeric data” and that a bank is nothing more than an “institution for the custody, loan, and exchange” of this data. Furthermore, that data was increasingly being stored and manipulated by computers, and would eventually “move around the world at the speed of light at minuscule cost by infinitely diverse paths.”59 He then came to one of his most important conclusions: Any institution that could move, manipulate, and guarantee alphanumeric data in the form of arranged energy in a manner that individuals customarily used and relied upon as a measure of equivalent value and medium of exchange was a bank. It went even beyond that. Inherent in all this might be the genesis of a new form of global currency.

Lastly, Hock realized that he, and most of his fellow bankcard managers, had misunderstood what business they were in: It seems ordinary and obvious now. It was a revelation then. We were not in the credit card business. “Credit card” was a misnomer based on banking jargon. The card was no more than a device bearing symbols for the exchange of monetary value. That it took the form of a piece of plastic was nothing but an accident of time and circumstance. We were really in the business of the exchange of monetary value.62

See creation_of_national_bankamericard_inc.md