The First Network
History has forgotten what may be the first true electronic information network. In the 1930s, Kaufmann’s Department Store in Downtown Pittsburgh had a fully automated point of sale accounting system installed throughout the store. How well the system worked, and for how long, remains a mystery.
Today, Pittsburgh is known for a lot of science and technology, including a fair number of “firsts.” Pittsburgh is where Jonas Salk developed the first effective polio vaccine in the 1950s at the University of Pittsburgh, while decades later, the work of Thomas Starzl also in Pittsburgh made him the father of modern organ transplantation. But of late, if you were to pick the latest technology coming out of Pittsburgh, many will point to the rapidly evolving field of autonomous vehicles, which is just the latest iteration of the region’s long history in robotics, computer science, and really, much of modern information technology.
There is nothing new about computers in Pittsburgh. The Carnegie Institute of Technology, later to become Carnegie Mellon University (CMU), founded its Computation Center in 1956 and CMU dedicated its Robotics Institute in 1979. But to understand the role of technology in Pittsburgh’s history, you really have to go back much further.
Just after the Civil War, George Westinghouse came to Pittsburgh to start his first successful business, which developed a revolutionary braking system for trains. He went on to shape the modern electric power industry and created the Westinghouse Electric Corporation in 1886. In 1906 Westinghouse established one of the first permanent corporate research centers anywhere in the United States in Pittsburgh, and in 1913 the Mellon Institute was founded in Pittsburgh as one of the first academic research centers focused on supporting commercial research.
Research activity in the region grew through the first half of the 20th century. In addition to Westinghouse, Gulf Labs was a worldwide leader in developing new technologies in energy discovery and production, while Pittsburgh’s Alcoa Corporation set up a lab in New Kensington for research in aluminum and materials science. By the early 1960s, greater Pittsburgh was the home to at least 60 different corporate research centers including recently created labs advancing cutting-edge fields of nuclear power and space technology.
But there is one technology that defines the modern world that Pittsburgh does not get credit for, and maybe it should. Pittsburgh may be where the first operational electronic information network was installed? It’s not a claim you have probably ever heard before and goes back a lot further in history than you might presume. Decades before anyone began conjuring robotic cars and even before any computer research began at either Carnegie Tech or Pitt, there was once a fully functional electronic information system installed in Pittsburgh.
It was not even at a local university or corporate research laboratory where the network was installed, but in Kaufmann’s Department Store in Downtown Pittsburgh. The system connected electric sales terminals throughout the store, collecting individual transaction and credit information, and then was able to transmit the data over electrical wires for centralized processing. All this was in operation by the early 1930s, decades before similar technologies would become standard across retail industries. The system would not survive through the depression and, for the most part, has disappeared from history, but it may have been the precursor to all of the computer networking we rely on today.
Computer networking is not a technology that came from a single eureka moment from which all future developments can be traced. Modern networks are the result of multiple technologies that evolved along many parallel paths that eventually converged to build our hyper-connected world.
You might think it should have worked out the opposite way, but the history of Wide Area Networks (WANs) — which connect computers over long distances — came before computers were connected in local clusters. Early digital computers typically took up entire rooms and were very much stand-alone devices, each a virtual island of technology. Initially, there were so few of these large devices that there was no need, nor any method, to connect them to their distant peers in other cities, but that would change.
Many of the earliest computers were being used by the Department of Defense. At least as early as the late 1950s, advanced air defense systems were sharing electronic data and operating very much as integrated computer networks do today. By the 1960s, general-purpose computers within the Department of Defense were beginning to be connected by a new system called ARPANET. The name derived from the Advanced Research Projects Agency, later DARPA, that sponsored the network.
In the early 1980s, academic institutions created their own WANs, such as BITNET, and in the 1990s NSFNET, a network of federal computers outside of the Department of Defense was created. When these early computer networks and others began to be connected to each other, the Internet as we know it today was born.
Local Area Networks (LANs) were only needed once computers began to shrink in size and the number of computers expanded. But LANs evolved from several independent paths across a number of institutions in the late 1960s and into the 1970s as smaller computers, and later personal computers proliferated.
Where does Pittsburgh fit into the early history of computer networks? Local Pittsburgh corporations were among the first customers of the earliest digital computers such as the UNIVAC, which was made by the Remington Rand Corporation in the 1950s, but like other similar computers of the time, they were very much stand-alone devices. Pittsburgh was home to one of the earliest nodes on ARPANET when a PDP-10 mainframe at Carnegie Tech was connected to the network by at least 1971. Then in the 1980s, CMU began building out its campus-wide computer infrastructure called Andrew.
CMU can also claim a role in creating the exponentially expanding “internet of things” when in the 1980s, students connected a basement vending machine to the school’s network, making it capable of responding over the Internet to queries on its inventory and even how cold cans were.
But there is a much earlier story to tell, a story that has mostly been forgotten in the history of information technology. It was in Downtown Pittsburgh that a fully integrated information network was operating at the beginning of the 1930s. The network was installed in the Kaufmann’s Department Store, the largest in Pittsburgh, and connected dozens of its sales counters, and processed credit transactions at a centralized tabulation system. Similar technology is embedded across the retail industry today, but it was more than revolutionary nine decades ago.
The system was the invention of Ed Rogal, an electrical engineer and a 1918 graduate of the Massachusetts Institute of Technology and MIT professor Louis Woodruff. Beginning in the 1920s, the two had worked on connecting electrical tabulating machines and getting them to communicate with each and work together, something that was unheard of at the time.
Tabulation devices were the precursors of modern computers. Capable of aggregating data stored on paper punch cards, it was a technology that revolutionized data processing. Modern tabulation machines date back to at least the 1870s when Herman Hollerith devised a mechanical tabulation device. Hollerith machines were used to compile data from the 1890 census and found ever-growing uses in the private sector. Electrical tabulation machines supplanted mechanical devices and by the early 20th century were common in major corporations, typically used for key accounting functions.
But tabulation machines, much like early computers, were very much stand-alone devices. The only connection between them was in the form of paper punch cards, which served as a medium for both data storage and transmission. What Rogal and Woodruff had spent much of a decade perfecting was a way to get tabulation devices to automatically ‘talk’ to each other and share data via electrical wires.
In 1926 they formed Central Records Inc. and applied for a patent for an “Automatic Accounting System”, an interconnected system that included front-end user interfaces and data input via punch cards. The true innovation came from electrical connections over dedicated wires or phone lines to a bank of central processors or “totalizers” that would be able to integrate data from hundreds of remote machines.
By 1928 Central Records Inc. was ready to put the system to the test the system and went looking for clients. Anticipating the source of future clients, the system they designed was intended to automate much of the accounting needed within a modern department store.
American Department stores were becoming ever larger enterprises by the 1920s. Flagship stores for iconic department stores like Macy’s or Gimbels had thousands of employees and sales happening simultaneously across literally hundreds of locations within their stores. Department stores were also systemizing the issuing of credit, which generated enormous amounts of new data, data that needed to be tracked to individual accounts and constantly aggregated in order for the stores’ accountants to keep track of sales, profits, and inventory. Long before the term “big data” was coined, data management and the back-end accounting of daily operations were major expenses for large department stores across the nation.
Department stores had not found a practical way to use tabulating machines to ease their accounting burden. Typical store sales were made of a large number of small individual transactions at separate locations. The cost and time required to create individual punch card records of each transaction was impractical and costly. What Rogal and Woodruff realized that if data could be collected automatically at each point of sale and tabulated at a centralized location, department stores could benefit from economies of scale and could greatly reduce their accounting costs.
After showcasing their idea at a May 1928 convention of the National Retail Dry Goods Association, an invitation was accepted to build a test system at Kaufmanns’ Department Store in Pittsburgh. Kaufmann’s was not only the largest department store in Pittsburgh but, in many ways, one of the most innovative. In 1931, Kaufmann’s would become a rare non-industrial member of Pittsburgh’s Mellon Institute. The institute had recently spun out of the University of Pittsburgh but continued as an independent center for industrial research. Kaufmann’s was slated to become a “living laboratory” to study scientific management practices in the retail industry. [1]
By 1930, a test system was in place and expanded to 10% of the sales counters in the store. In 1932 a contract was signed to put in a complete system throughout the store. Eventually, the full plan was for 250 point-of-sale recorders connected by dedicated wires to 15 online “typewriters” and 20 centralized “totalizers.” [2]
In operation, the system was unlike any of the manual procedures using sales slips that were common across the industry. At the point of sale, two punch cards or tokens were used. One token encoded information about the sales agent and another with information on the cashier processing the sale. The cashier inserted both tokens along with a card containing information on the item being purchased into an electrically operated transmitting machine at the sales counter. The information on the cards was then delivered electrically to one of the central records machines, where data was aggregated and stored on punch cards for future accounting.
The system also automated the processing of store credit and facilitated a real-time credit authorization procedure. If the sale was a charge, instead of the cashier’s token being used, the customer provided an individualized shopping token on celluloid, essentially a synthetic plastic punch card with their store credit account number. Before the sale was completed, information on the item and purchaser was transmitted electronically via a “charge authorization unit” to a central location where a store clerk would confirm the customer’s account and electronically approve the transaction. The system presaged much of how modern credit cards work today, with the store providing each authorized customer an individualized punch card encoding their personal information.
While there is evidence that the test system was installed and worked as it was intended, it is unknown just how well and for how long the complete system ever operated. In 1932, installation of the full system was reported to have been almost completed, and a 1937 Pittsburgh Post-Gazette article mentioned that the system would be showcased at a conference of retailers hosted in Pittsburgh that year. [3] But a 1938 article by Professor Woodruff gave a complete description of how the system worked but only referred to a “large experimental installation” and gave no explicit reference to Kaufmann’s even though that had to have been the site described in detail.[4]One illustration in the article even showed a punch card price tag used in the system that identified Kaufmann’s on Fifth Avenue. Local or professional news documenting how the system worked through those years is fleeting. There is little evidence the system survived through the end of the decade, or what eventually caused it to be taken offline.
Whether the system had technical problems or failed to provide the promised savings is unknown. The ongoing depression could have been a significant factor in the system’s quiet demise. Department stores suffered more than most other industries as customers drastically cut back on discretionary spending. And in an era of heightened unemployment and depressed wages, the need to automate processes was less a priority. Whatever the reason, only limited histories of the system built by Central Records Inc. and its implementation at Kaufmann’s have survived. Though the system installed at Kaufmann’s was repeatedly hailed as a prototype for what could expand to major retailers across the nation, there is little evidence any other major installations ever happened.
Central Records, Inc. survived for at least the next decade, and Charles Rogal was present at several early conferences on the development of early computers into the 1950s. But there is no evidence the company ever built a complete system or successfully commercialized its “Automated Accounting System” that first found a home in Pittsburgh. Louis Woodruff continued at MIT into the 1940s before retiring and is credited with an invention that enabled the tracking of household radio listening preferences, a system that presaged the creation of aggregate marketing data such as the Nielsen ratings.
Real-time point of sale electronic tracking in the retail industry would be relatively uncommon even into the 1970s. In the 1960s, Edward Rogal became associated with the American Totalizer company which developed a system called Uni-Tote for retail point-of-sale automation. Uni-Tote had many of the features of the system that was installed at Kauffmann decades earlier. But Uni-Tote and some of its contemporary competitors in the 1960s were still described as experimental systems and were not universally adopted. [5]
There are two interesting sidebars the story of Central Records Inc. and the system it installed at Kaufmann’s. In 1932 the company proposed installing an automated accounting system for the city of Pittsburgh, but there is no evidence the offer was accepted. Then in 1933 the company submitted a more intriguing proposal to Allegheny County to create a system similar to what they were then building for Kaufmann’s, but to automate and integrate the county’s entire voting system. [6]
Mechanical voting machines were becoming common early in the 20th Century. But those mechanical voting machines, much like the general-purpose tabulation machines of the era, were also stand-alone devices Individual machines aggregated votes but then typically produced a paper record that needed to be aggregated across all the other voting machines in a jurisdiction. The company was essentially proposing a district-wide integrated voting system, with electronic data shared between individual voting machines and a centralized tabulation system literally decades before any such system was re-envisioned. But for unknown reasons, the county did not take up the offer — likely because such a novel system never received necessary certification from the Pennsylvania Secretary of State — and pursued more conventional tabulation machines. Like the retail technology that would become routine later in the century, the automated voting machines proposed were many decades ahead of their time.
Eventually of course, the automated point of sale system pioneered by Central Records, Inc. in Pittsburgh would become ubiquitous across the retail industry. But for a period, it was at Kaufmann’s where the information age was evolving long before it arrived most anywhere else.
(If anyone has any more information on the operation of, or what became of, the Central Records Inc.’s system at Kaufmann’s, please let me know)
______________________________________________
[1] Push Research in handling commodities, Pittsburgh Post-Gazette, March 16, 1931. p. 6.
[2] L.F. Woodruff, Remote Control Accounting, Technology Review, July 1934. p. 376.
[3] Retailers Plan Convention Here, Pittsburgh Post-Gazette, May 10, 1937, p. 6.
[4] L.F. Woodruff, A System of Electric Remote-Control Accounting, Electrical Engineering Transactions, Volume: 57, Issue: 2, Feb. 1938 pp. 78–97
[5] James W. Cortada, The Digital Hand: How Computers Changed the Work of American Manufacturing, Transportation, and Retail Industries, Oxford University Press. (2003), p. 295.
[6] Vote Machine Offer is Asked, Pittsburgh Post-Gazette, January 16, 1933. p. 1.