In late 1966, five Chicago banks entered the credit-card business at the same time. Unsolicited cards arriving through the mail offered a quick way to reach potential users. The strategy brought the new product to customers before they had applied for it. Federal Reserve System, 1968, p. 73.
The commercial logic was understandable. A card needed places to use it, and merchants needed a reason to accept it. Recruiting one side without enough participation from the other could leave an expensive program with little activity. Sending the product directly helped an issuer get past the problem of waiting for people to request something they had barely encountered.
Bank of America’s earlier Fresno experiment illustrates the scale of that ambition. At Visa’s 2017 Investor Day, executive Jim McCarthy described 60,000 mailings in Fresno in 1958 to test a portable, general-purpose line of credit. Visa, 2017, p. 16.
BankAmericard’s reach subsequently expanded through licensing to other banks beginning in 1966. Federal Reserve History also describes the operational work that broader acceptance required: early interbank settlement was paper-based and bilateral, while merchants often had to telephone for authorization. Putting plastic in customers’ hands was one part of building the service. The rest had to support what happened when someone tried to pay. Federal Reserve History.
The mailing carried the risk
The Chicago report described intercepted cards. Banks called cardholders after first use to confirm who had used the card. The report also identified no known case of an individual being held responsible for unauthorized use. Federal Reserve System, 1968, p. 75.
The mailing was doing two jobs: recruiting a customer and delivering the means to spend. If it reached the wrong person, a bank could face fraud even when its assessment of the intended customer’s creditworthiness was reasonable.
Two rates, two different questions
Marine Midland’s 1966 charge-card program shows why the results need careful reading. It received 221 applications from 33,357 promotional mailings. In a separate group of 731 people sent cards without written requests, 139 were active within 60 to 90 days, about 19 percent. Federal Reserve System, 1968, p. 76.
The application response works out to approximately 0.66 percent. Putting that beside 19 percent makes direct distribution look dramatically more successful. But the two percentages measure different outcomes. One counts applications following promotional mailings. The other counts use among a group that had already received cards. They cannot tell us how the same customers would have responded to each approach.
Skipping the application helped the bank put more cards into use. It also changed the starting point of the measurement. Applications from a mailing and activity among people already sent a card answer different questions about demand.
There is also a difference between observing use and establishing that the recipient wanted the product issued in the first place. A card that was not returned might sit in a drawer, be discarded, or eventually be used. The absence of a returned envelope cannot settle the question of informed customer choice.
The rules caught up
The report’s authors opposed a legislative ban on unsolicited cards, partly over concerns about barriers to new entrants. Federal Reserve System, 1968, p. 4.
In 1970, Congress required a request or application before a credit card could be issued, with an exception for renewal or substitution of an accepted card. The provision took effect on October 26, 1970. It governed card issuance; banks could still mail advertising and application forms. 15 U.S.C. §1642 and historical notes.
The banks had a difficult business to build. They needed enough cardholders and merchants for the service to be useful. That commercial problem did not settle whether sending someone credit they had never requested was an acceptable way to acquire a customer.
Suppose a bank automatically makes a new payment feature available to every eligible customer. That total measures access. Separate counts of customers who knowingly enable it, make a first payment, and return to use it show how the relationship develops. A growth report that combines those stages can make an institution’s distribution decision look like customer demand.
Distribution also deserves its own fraud review. Checking a transaction after it happens cannot establish that the card reached the right person in the first place. The Chicago experience shows how an acquisition decision can create an operating risk before a customer ever chooses to use the product.
A mailing count told an issuer how far its distribution reached. It could not tell the issuer how many people wanted a card, how many would keep using it, or whether the card had arrived safely. Those were separate questions in 1966. A bank assessing adoption today still needs separate answers.
Sources
- Federal Reserve System Task Group — Bank Credit-Card and Check-Credit Plans
July 1968; printed pp 4 and 73–76. Contemporary report drawing on bank-supplied information. Its Marine Midland comparison is not treated as a randomized experiment.
- Visa — 2017 Investor Day transcript
June 22 2017, Jim McCarthy presentation, printed p 16. Retrospective company account of 60,000 Fresno mailings in 1958, rather than an independently verified contemporary count.
- Federal Reserve History — Electronic Point of Sale Payments
Published September 25 2024. BankAmericard licensing chronology, paper settlement and telephone authorization. Historical synthesis rather than a contemporary operating record.
- Office of the Law Revision Counsel — 15 USC 1642
Issuance of credit cards, statutory source and effective-date notes. Public Law 91-508, Title V, section 502(a), 84 Stat 1126; effective-date section 503(1), 84 Stat 1127. Used to establish the 1970 rule and timing.
Historical facts and reported figures are linked to sources. The 0.66 percent response is calculated as 221 divided by 33,357, rounded; it is not directly comparable with the reported active-use percentage. Modern examples and measurement implications are the author’s analysis. Reviewed November 12, 2026.