The tea shops that ruled the IT world

A new book tells the story of Leo, the world's first business computer, devised by British catering company J Lyons in 1951. Ross...

A new book tells the story of Leo, the world's first business computer, devised by British catering company J Lyons in 1951. Ross Bentley talks to two members of the original team.

This is the story of how the humble British cuppa changed the face of global business for ever. In October 1947, the directors of J Lyons, a British catering company famous for its teashops but with a strong interest in new office management techniques, decided to take an active role in promoting the commercial development of computers. Four years later, the Leo I computer was operational and ran the world's first regular routine office computer job.

Based on the electronic delay storage automatic calculator (Edsac) machine developed at Cambridge University, Leo I occupied 2,500 sq ft of floor space and used technologies such as thermionic valves and paper tape readers.

By today's standards it was very primitive but back in post-war Britain it enabled Lyons to drastically streamline its clerical operations.

"Prior to Leo, we had developed a very efficient clerical system. We knew we could go no further with the existing equipment. We realised we needed something extra," says David Caminer who was systems manager at Lyons at the time.

Caminer went to work for Lyons at its Cadby Hall headquarters in Hammersmith, London, during 1936 with no idea that his later work would be intrinsically linked with the development of British business computing.

According to Caminer, one reason why Lyons was so progressive in terms of its commitment to business computing was because it had a champion of the Leo programme on the board. "John Simmons, controller of the company at the time was our advocate," he says. "Everything we did was fed back to the top managers so that they fully understood what we were working on. We made a point of simplifying things like the flow diagrams so that they could be understood by everyone."

Caminer says the post-war period was a time of intense technological innovation when the rate of development was incredible. "Solid state electronics was coming in, multi-layered circuitry had been developed but in those days there was no structure for R&D. We were so close to it in those days we could not see where it was all going. But we knew we were working with the most advanced computer in the world at the time."

There is no doubt the Leo team were galvanised by the idea that they were working at the forefront of computing innovation but being a trailblazer also had its drawbacks. The first job Caminer's team worked on concerned the automation of payroll but it never went live. Although the details of 10,000 bakery staff were placed on the system and the team went through hours of rigorous testing, the decision was made to pull the project because of a lack of business continuity planning. After all the Leo computer was a unique machine with no failover. If things had gone wrong "it would have been difficult to capture the damage", he says.

One of the first applications to be used in anger was to facilitate the distribution of food items to Lyons' outlets. "The company produced 250 different food items to be distributed to 250 teashops across the country," he says. "The original application we wrote relied on clerical staff spending a great deal of time inputting items. A new list was required from each tea shop every time fresh items were ordered. So to simplify things, we came up with a standard list and asked manageresses to send in alterations. You could say this was the first example of systems re-engineering. This meant that orders could be changed at the last minute to cater for specific demands. For example, if there was a heatwave or a cold snap we would have an upsurge in demand for beef and dumplings, say, or salads.

"Manageresses could also devote themselves to overseeing customer service rather than keeping on top of the clerical work as they had done before."

But despite the advances being made by the Leo team, no one expected or received special treatment. "People in the company were aware of the work we were doing but we were still looked upon as members of the clerical department," Caminer says. "The advantage of working for a catering company was that we got treated to the odd slap-up meal at the Trocadero or Northumberland Hotel."

The closeness of the Leo team, is something that comes across when talking to veterans of the time. This feeling of belonging is demonstrated by the healthy turn-out at regular reunions of the Leo Society which take place every two years or so.

"In the early days we were a small body of people peaking at about 12 members. If someone couldn't solve a problem we spoke about it at coffee break and inevitably we would find a solution. I'm mystified by the number of people it takes to build an application today. It's a wonder they don't trip over each other."

The team spirit is fondly remembered by Frank Land who was recruited into the Leo team as a trainee programmer in 1952 and at 75 years old still lectures in information systems at the London School of economics.

"Someone was always doing something new. At the time we were aware that we were pioneers and that created a buzz about the place. We would discuss problems at lunchtime and there would always be someone on hand to listen."

His first assignment was to build a program to keep track of the stock in the reserve stores. "At the time rationing was in place. Sugar was hard to come by so we had sugar substitutes and sweetened fats in our store," he says. "The program was very sophisticated, it integrated transport, forecasting and stock control."

But when rationing came to an end in 1953, the market conditions changed making the application obsolete. Again this program never made it live. "It was an interesting lesson I learnt early on: the best-laid plans can change if the situation changes," he says.

Land says there were many companies eager to emulate the successes that Lyons had achieved through the use of computers. "They were looking at computers but many firms just wanted to automate processes not business process re-engineer. Our attitude was that computing was not about bean-counting and the mere automation of processes - it was about extracting value for the business.

"But, in those days organisational methods were very strict and many people were constrained by their roles. The situation was also hindered by the staunch class divisions that existed. Managers did not communicate with people on the ground and as a result they really did not know their business."

Leo was spun off as a separate company in 1954 as Lyons decided to focus on its core catering operation. "It's incredible to think that, for a time, this little offshoot took on IBM, one of the biggest corporations in the world," says Caminer. "Our business came out of nothing whereas IBM had a rich history in heavy-duty office machinery."

Land agrees, "Leo was thoroughly well-engineered and superior to the IBM 360 [mainframe] in many ways."

There was a lot of demand, says Caminer, "we were mopping up jobs". The company devised PAYE tax tables for the government, payroll for Ford Motor Company and Kodak, as well as writing an application to calculate the distances between all British Rail stations in order to work out pricing structures.

The company released several versions of Leo over the years. In all 85 computers were sold. Leo II computers were installed in many British offices, including British Oxygen Company and the "clerical factory" of the Ministry of Pensions at Newcastle upon Tyne. Leo III computers ran at Customs & Excise, Inland Revenue and the Post Office and in Australia, South Africa and Czechoslovakia.

The company Leo Computers merged with the computer interests of English Electric in 1963 to form English Electric Leo and later, English Electric Leo Marconi. Subsequent mergers eventually found Leo incorporated into ICL in 1968.

Both Caminer and Lane agree that, besides the Leo computer, itself, the greatest legacy from this remarkable era of innovation at Lyons is the impact people from the Leo team have had on the wider world of IT. "John Aris went on to become director of the National Computing Centre. Peter Hermon went to British Airways and set up the world's first flight reservation system," says Caminer. Land adds, "Then there was John Gosden who became a US presidential adviser on IT issues."

How Leo came about  - an extract from A Computer called Leo, by Georgina Ferry                 

The total time taken to build a modern PC probably amounts to a matter of minutes. Circuits are mass-produced, etched in silicon or printed on boards. Even when several such boards are assembled in one box, together with hard and CD-Rom drives, there are just a few contacts to screw down. And the completed product, in the case of a laptop, is not much bigger than a hardback book. 

When John Simmons, controller of the company, began to plan for the building of the Lyons computer, soon after Pinkerton (the head computer engineer) arrived in 1948, he set aside 2,500sq ft of floor space on the second storey of WX Block at Cadby Hall. But essentially it consisted of exactly the same elements as a modern PC: an input device, a store or memory, a processing unit (combining control and arithmetic units) and an output device. The Edsac team (the computer, on which Leo was based) in Cambridge made all these elements by adapting pre-existing technology such as thermionic valves and paper tape readers: they did not have to invent anything. 

There was a problem with Edsac's 3,000 valves, they were not very reliable. When it broke down, the engineers would simply replace valves one by one until they found the dud, a tedious and time-consuming process. 

Pinkerton decided that for ease of maintenance he would make the computer out of interchangeable units. Each unit could hold up to 28 valves, with their associated circuitry. The units slid easily in and out of metal racks holding up to 12 units each. All the wiring, carrying the power supply and the electrical connections from one unit to another, was threaded through the racks. 

Given the calibre and dedication of Pinkerton and his team, the task of producing what was essentially a copy of Edsac was not intrinsically difficult, and they completed it within two years. The real challenge for the engineers was to build the additional parts of the system that Edsac did well enough without. 

What were needed were high-speed methods of delivering data and programs to the machine, and of producing the results. There would be little point in building, at great expense, a machine that could carry out a large number of simple calculations at electronic speeds if the data which the machine was to manipulate could be fed to it only much more slowly. Similarly, it would be wasteful to keep the computer waiting for its next job while the results of the previous calculation were slowly printed out. 

Thompson and Standingford (Simmons' assistants) had thought the answer might lie in recording the input and output data in binary form on to spools of magnetised wire, a technology that has been used for some time to record sound for broadcasting and was actively being researched both at Harvard and the Moore School during their visit. The spools could be wound back and forth at a speed much greater than traditional punched tape. 

The programmers were desperate to run something other than the test programs beloved of the engineers, to demonstrate the worth of the machine to the company. Systems manager David Caminer chose a job known as Bakery Valuations, because it made relatively low demands in terms of the volume of data to be introduced to the machine, yet it would still produce information that was useful to managers.  

It involved calculating the value of the week's output of bread, cakes and pies from the Cadby Hall bakeries, taking into account materials, labour and indirect costs such as the power to run the ovens; calculating the value of bakery products leaving the site for distribution to grocers shops, teashops, restaurants or private customers in terms of factory costs, retail price and profit margin; and calculating the value of the products held in stock. This process was conventionally done as three separate operations, occupying 50 hours of clerks' time per week. 

Caminer specified the requirements of the job and drew up a flowchart to show how the different parts of it related to one another. "We decided that everything should be charted meticulously, and laid down absolute standards for the charting," he says. 

"I was at particular pains to make sure the charts were understandable by ordinary people. First of all for the person who was going to write the code; secondly for the person who was going to alter the code; and thirdly to be explained to the users on the ground."  

The job of writing the Bakery Valuations program went to John Grover, with assistance from Tony Barnes, "sometimes leading, sometimes restlessly in step" in Caminer's words. It ran successfully for the first time on 5 September 1951, but only as an experiment. Less than three months later rehearsals were over, and it was time to launch the new program as an integral part of Lyons' management system. On 29 November 1951, Leo took over Bakery Valuations from the clerks who had previously done the work, and became the first computer in the world to run a routine office job. 

For the rest of the life of the machine, Leo printed out weekly reports to be incorporated in the trading analyses delivered to Geoffrey Salmon, the director in charge of the bakeries. Using the slow-speed inputs and outputs, all that Lyons had available for the first year or two, it did the job in less than five hours, plus eight hours to punch the data on to paper tape.  

This performance, as Leo's designers and programmers were well aware, was far from state of the art. But the successful Bakery Valuations job demonstrated that a computer could be relied on to work week in, week out, not exactly without fail but with failures kept to a manageable level. And they knew what needed to be done to build up to the speeds that would be necessary to make bigger jobs such as the payroll worth doing. It was, at least, the end of the beginning.   

Reader offer

A Computer called Leo  (ISBN 1-84115-185-8) is published by Fourth Estate and retails at £15.99. However, it is available to Computer Weekly readers at the special price of £12.99 including post and packing. To order call the HarperCollins Publishers hotline on 0870-900 2050 and quote code 838E.

Read more on IT risk management