The Unlikely Story of a British Tea Company, the First Business Computer, and Payroll

Ian Zapolsky

min read

Ian Zapolsky is the Head of Partner Engineering at Check, and joined us in 2019 as the company’s first engineer. When he’s not reading up on the history of how people pay each other, he’s probably helping one of our partners launch and scale their payroll business. As a certified payroll nerd, Ian volunteered to share what he’s learned in a blog series about the history of payroll in the US.

The development of the electronic digital computer was hastened dramatically by World War II and the accompanying need to quickly and efficiently do things like calculate the trajectories of bombs and rockets, or decode encrypted messages sent by the enemy. The first computer of this kind, dubbed “Colossus,” was developed at Bletchley Park in the U.K. between 1943 and 1945 to help crack Nazi codes.

But in the immediate aftermath of the war, the utility of Colossus and its successors was not obvious to most business leaders, much less the general public. One reason for this was that the computers built during the war were not programmable, but rather had been developed for the sole purpose of completing one specific kind of task. How would a machine built to crack a Nazi code be of any use in an office setting?

In the years after the war, academic research and experimentation intensified around the question of how to turn the specialized computers of the early 1940s into general computing tools. In particular, Alan Turing, John von Neumann, and Grace Hopper made important contributions to the conception and development of “stored-program” computers, which were able to store instructions in memory, or have them input via switches or magnetic tape. Stored-program computers were a revolution because they allowed the same physical machine to perform a wide variety of tasks, depending on the instructions that were loaded into it.

Several academic institutions raced to develop the first stored-program computers in parallel, with several coming online around the same time in the late 1940s. One of the most notable of these projects was the “ENIAC” (Electronic Numerical Integrator and Computer), developed at the University of Pennsylvania from 1946 - 1948. ENIAC’s design and construction had been financed by the U.S. Army to calculate artillery firing tables, but because of its programmability it was also used to perform computations to aid the development of hydrogen bombs at Los Alamos.

This is where our story takes an unexpected turn. In 1947, J. Lyons and Co., one of the U.K.’s largest catering and food manufacturing companies, known for its chain of 250 teashops and London restaurants, sent two senior executives to the United States to scout for new business methods developed during World War II. During their visit, the executives met with Herman Goldstine, one of the developers of the ENIAC. To their credit, the executives saw the potential of the general programmability of the ENIAC to automate clerical and back office tasks. They were also encouraged when Goldstine told them that a similar machine, the “EDSAC” (Electronic Delay Storage Automatic Calculator), was being developed back in the U.K. at the University of Cambridge.

The executives returned home bursting with ideas and presented to the board about how a general purpose business computer could lower costs and improve quality for certain repetitive office tasks. The man who had sent the executives abroad in the first place, John Simmons, had been hired 20 years earlier out of the Cambridge mathematics department to devise new methods to improve back office operational efficiency. He was the right person to hear this ambitious plan, and after convincing the board of the direction, secured permission to approach the EDSAC research team at Cambridge with a proposal.

J. Lyons and Co. would offer £3,000 of funding to the EDSAC team, in return for permission to copy aspects of the EDSAC design, and the ability to collaborate with the research team and ask for advice while developing a machine of their own.¹ Simmons assembled a team of engineers, and over the next three years they developed a working computer, very similar to the EDSAC, which was dubbed the Lyons Electronic Office, or LEO I. When it was completed the machine occupied 2,500 square feet, had over 5,000 vacuum tubes, and used ultrasonic delay-line technology based on tanks of liquid mercury to implement its 8.75KB of memory. For context, that’s about 1,000,000 times less memory than an iPhone.

John Simmons standing in front of the LEO I

LEO I’s maiden voyage took place on September 5, 1951, when an application called “Bakery Valuations” was run to calculate the cost of ingredients used in bread and cakes.² The program ran from 3:50pm - 5:35pm without fault. Despite the success, one of Simmons’s early goals for the project remained unmet: payroll automation. It took a few more years of ironing out the kinks with the new machine before it was deemed reliable enough to handle the company payroll. But on December 24, 1953 the milestone was finally achieved. The task of calculating an employee’s pay, until then, had taken an experienced clerk 8 minutes. The LEO I could do the same job in 1.5 seconds.

It’s true then that running payroll was one of the first tasks performed by the world’s first business computer, and that the cost and complexity of payroll in part led to the necessity behind the innovation of the LEO I. Reinforcing this point, in 1954 J. Lyons and Co., at Simmons’s urging, stood up an entirely separate business unit called LEO Computers Ltd. to manufacture and sell computers, as well as rent out processing time on the LEO I to other businesses. By 1956 they were running payroll for some of the largest employers in the U.K., including the Ford Motor Company.

The technology was obviously revolutionary, and competitors arrived quickly. The larger, more focused, and better capitalized IBM quickly took over the business computing market, and by the 1960s both LEO Computers Ltd. and J. Lyons and Co. were losing money and market share. After a long string of mergers and acquisitions, LEO Computers Ltd. would eventually find a home in Fujitsu in 1990.

Below, find a short video telling the story of the LEO, from the London Science Museum.³