This October saw the 30th Anniversary of Big Bang. Press coverage has concluded the event has secured London as a leading trading hub and provider of financial services. Successful application of technology played a key enabling role and continues to do so, and the rise of a healthy FinTech industry in London must also be a welcome outcome.
I have been asked to provide a few words on my experience at the London Stock Exchange as a technologist in the years preceding Big Bang, and I hope the following is illuminating.
I joined the Exchange in November 1972, just as it was moving into its new home at 125 Old Broad Street. At 26 floors it was then the tallest building in London, and built in the brutalist pre-cast concrete style of the 60’s. It was a design that appeared to signal the exchange was at the centre of things and there to stay.
Alongside the tower was the cathedral-like market floor. Here stood the high priests of the Exchange - the Market Makers or Jobbers. These tall men dwarfed the brokers who would scuttle onto the floor from boxes located around the periphery to buy and sell stocks and shares handled and priced exclusively by the Jobbers. Floor etiquette was governed by arcane rules and protocols. This was a male only bastion with member only access, and a strict dress code.
My first weekend at the Exchange was spent on hands and knees in the catacomb-like structures under the market floor. Here an army of telephone engineers had been contracted in to transfer thousands of telephone wires to a newly installed private telephone exchange big enough to serve a small city.
I was familiar with this kind of work having enjoyed an apprenticeship in advanced telecoms development at British Telecoms Research, a Bletchley Park type of establishment set in Thames-side grounds next to Clivedon.
Immediately before joining the Stock Exchange I had worked on the team designing the next generation of computer controlled digital exchanges called System X. As I grappled with a wrap wiring gun and colour coded wiring schedules I couldn’t help thinking that the pristine telephone exchange we were commissioning, which was of an arcane electro-mechanical design, would be short lived. Indeed that was to be the case and within a few years it was replaced with a version of System X.
What I could not have predicted is the systems I would go on to instigate would eventually render the market floor redundant.
However, as I started to understand the machinations taking place at the heart of the Exchange, it was not long before I came to realise that it was feasible for the processes to be automated. I was also to learn that there was no appetite for this amongst the membership.
At the time I joined the Exchange there was little awareness of the pending digital revolution and the attendant opportunities for business development. Certainly there was no vision or blueprint for automation. Substantial investments had been made in computer systems but they lay idle in a nearby basement. A large ICL mainframe had been procured as the basis of an inter-firm accounting system that had not seen the light of day, and the consultants were in.
Also hidden away in a corner were two Ferranti computers of the type used by the navy for weapons control. They were painted in battleship grey and came with a host of spares and manuals. Huge umbilical cords spewed forth from these machines, in a somewhat intimidating fashion. On close inspection many appeared not to be connected to anything.
I learned that the system was intended to support collection of market prices by a team of market reporters operating on the market floor. The information was to be stored digitally, formatted as price pages and broadcast via a citywide closed circuit TV cable system with one page per channel. Hundreds of TV sets with push button tuners had been set up in broker’s offices around the city in anticipation of a launch, and there were many more in storage. The community system was intended to replace several systems that had been set up by individual broking firms for their customers. White boards located in brokers boxes were marked up with current market prices collected from the market floor. As a simple expedient TV cameras were pointed at the white boards and the signal transmitted to TV sets located in offices around the city via several private cable systems.
My manager, who headed the Telecommunications Office, said that getting the system up and running could usefully be the immediate focus of my attention. I had at my disposal a fully equipped laboratory and a cable TV technician, my first team member. I was also provided listings of the application program written in machine language, and hand drawn schematics of the general arrangements authored by property management staff. He also explained that not all of the Exchange members were supporting the introduction of the communal system. The bigger brokers with boxes nearer the market floor, which commanded a premium, could refresh their prices quicker than their smaller competitors. A communal system would level the playing field to the big firms’ disadvantage.
I uncovered numerous design and implementation faults. Surprisingly given their critical role in military applications I also found design problems with the Ferranti machines themselves that required factory modifications. The MPDS (Market Price Display System) was launched in a low-key fashion when all was ready, and the service was rapidly adopted first by the smaller firms and then the bigger firms.
By this time the mainframe consultants had set up a new department called, somewhat grandly, the Directorate for Information Systems and Settlement, and I was transferred there.
A large and complex inter-firm settlement system called TALISMAN (Transfer accounting and lodgement for investors, stock management for Jobbers) was being planned based on IBM mainframes. Previous efforts involving the ICL mainframe were to be scrapped. A meticulous plan had been worked up and a large team of analysts had been drafted in to do systems analysis in a classical waterfall project approach.
What was needed, as a first phase, was a transactional system for capturing post trade deals and matching them as a precursor to overnight batch settlement processes. The existing manual process involved a team of clerks rerunning the day’s business in a checking hall to build a ledger of matched transactions. They had to scrutinise and match hastily made out trading tickets which often had not been spiked but rather swept up from the market floor. The whole process looked like the sort of vote counting process you see at general elections but more chaotic.
It had been determined that the transactional system would be best implemented using Digital Equipment Corporation (DEC) minicomputers and I was asked to lead a small team of specialist contractors to develop the solution.
I decided on an incremental rapid development approach I had learnt from my R&D experience. This involved development of a bare bones prototype constructed in close concert with broker’s operations staff. This was subsequently elaborated and refined. This is the approach I was to adopt for future developments using small teams of highly motivated experts. The style included a flat organisation, an open office to aid communications, and an enabling management style to ensure team members had the best tools and equipment to accomplish their tasks. Today this would be recognised as an agile methodology.
The trade matching solution involved the development of algorithms to correlate counterparty and security identification, price, size, time and a host of other bargain conditions some of which were mandatory and others optional. The development went well and we soon had a working prototype that we progressively refined. The mainframe team could not believe the rapid progress we were making whilst they were mired in deeper and deeper systems analysis that was throwing up huge amounts of impenetrable documentation. Some of things we learnt during our prototyping proved useful in helping them clarify things such as data exchange definitions.
As we neared successful implementation and launch I started to think about next steps. It occurred to me we had the basis of a completely automated trading system. All that was needed was additional functions to capture price quotations from Jobbers and the capability for Brokers to enter orders via screens that could hit or take the quotes using a suitable matching algorithm. This would obviate the need for post trade data entry. I decided to test the water.
To achieve buy-in of its activities by the powerful Jobbing community the Directorate had appointed a chairman seconded from the largest Jobbing firm, Wedd Durlacher. I decided to put the idea to him. He graciously took me out for a City lunch and he invited me to regale him with my best thoughts on the changes computer technologies could be expected to reap on the city over several jugs of port.
The conversation centred on Ariel (Automated Real-time Investment Exchange Limited) an automated trading system that had been launched, somewhat controversially, to match institutional trades via, as it happened, Ferranti computers. The system was being promoted by a group of City Accepting Houses and mirrored a similar initiative called Instinet in the US that used DEC computers. His advice was that powerful forces were at play and Ariel would not receive the support it needed to be viable. He also thought that it was far too early to be proposing such a development at the Exchange. His prediction about Ariel was soon to prove correct, and I quietly shelved thoughts about automated trading systems.
I was steered back towards information system developments. There was a growing need for clean machine-readable data delivered in real-time by some brokers who were beginning to focus on quantitative and technical analysis. Quote Vendors were also keen prospective buyers.
I devised a system based on DEC Computers to capture quotes continuously from the market floor and to disseminate them in real-time as computer readable feeds. The information was also accumulated in a price history database. The system was called EPIC (Exchange Price Information Computer) and was readily taken up, generating significant income from real-time and historic data sales.
By this time the Market Price Display system was proving to be an information bottleneck as only 22 channels (each equivalent to a displayed page) were available on the closed circuit TV distribution system. At the time the Post Office and BBC were experimenting with a system for encoding and distributing text and basic graphics to colour TVs equipped with as suitable decoding chip. Mass production of the necessary Teletext TVs was in prospect.
I roughed out a design that would provide fast interactive access to an unlimited amount of pages from any number of standard Teletext equipped TV sets. The sticking point was finding computers powerful enough to achieve the sub-second page request response times that I anticipated users would demand. I found the answer at NASA.
They were using fast computers to process the numerous telemetering signals and real-time responses inherent in rocket launch control. The machines manufactured by a company called Modcomp in Florida were a perfect fit for what I had in mind.
There were many other innovations required to achieve success, but in short order we had a system ready for launch in the early 80’s. My team comprised the brightest computer science graduates from universities such as Imperial College, Strathclyde and Manchester as well as some seasoned contractors.
In addition to price and company news pages generated from Exchange operations, we also provided facilities for member firms to create and post their own pages containing research and recommendations to nominated closed user groups.
The take up of the service was phenomenal and soon revenue from information services was growing by leaps and bounds to become the Exchange’s largest income segment. This was fortunate as the TALISMAN settlement system that was finally introduced in 1979 had considerably overrun budget and timescales and had caused the exchange to mortgage the Tower to balance the books.
The large TALISMAN development team was to move on to attempt a paperless settlement system called TAURUS (Transfer and Automated Registration of Uncertified Stock).
In the meantime EPIC and TOPIC were combined to launch a new international market based on displayed quotes and telephone trading. This was to prove highly successful in attracting significant trading in equities listed in Paris, Frankfurt and other European centres to London. This was called SEAQ International (Stock Exchange Automated Quotations).
The EPIC and TOPIC systems were also use to underpin a nascent traded options market.
I then was asked to join a team at working on a system for settling high value Gilt-edged securities trades to be operated by the Bank England.
I devised a platform and architecture that featured settlement by Book Entry Transfer using a distributed ledger. This was replicated and synchronised over multiple computer nodes implemented on a Tandem Non-stop system. User access was on a permissioned basis via a VPN connection using IBM PC-based thin clients. RSA Public Key Encryption was used for authentication. PCs had hardware-accelerated cards for encryption and X25 protocol support.
Gilts BET was launched successfully in the early 80’s.
It was at this time that the Exchange was embroiled with Cecil Parkinson, Trade and Industry Secretary about ending the restrictive practice of fixed commissions. This was to be the first brick out of the wall in restructuring, and opening up the Exchange, as has been well documented elsewhere.
Now there was an external and urgent imperative to review trading structures and come up with a plan for a fresh approach. Given the urgency it was decided to create a SEAQ Domestic System modelled on SEAQ International, using expanded versions of EPIC and TOPIC and this was to be pressed into service in October 1986. In the planning I voiced reservations to the implementation team, as my calculations indicated that not enough hardware capacity had been factored in to deal with the surge in traffic that could be expected on market opening. Unfortunately this was to prove to be the case and the system suffered from overloading and crashed early on the launch day. This was subsequently rectified, and the system was to serve it purpose reliably for many years thereafter.
As a postscript, in 1987 found myself on the Stock Exchange Managing Board with responsibility for developing a blueprint for future development. This was to be a bleak time as following the crash of 1987 the Exchange started to turn inward under the leadership of a new Chairman and Chief Executive. There was to be drastic cost cutting, divestment of its information technologies and outsourcing of the IT functions to management consultants. The only development that was ring fenced was TAURUS, and this was consuming ever-growing resources. The Chief Executive, an accountant, took on personal responsibility for delivery of the project.
I decided to move on to create my own vision of an automated Exchange. This led to the successful launch of the Tradepoint Investment Exchange in 1995 that was to be a game changer.
Unfortunately the Stock Exchange’s TAURUS project was to fail in the early 90’s at considerable cost to the exchange and the City. The Chairman of Stock Exchange was forced to go to the Bank of England cap in hand requesting they take on the mantel of development of a paperless settlement system. The Exchange Chief Executive lost his job.
The Bank was to extend the Gilts BET system to create CREST (Certificateless Registry for Electronic Share Transfer). This was successfully launched three years later.
To conclude I would like to take this opportunity to thank all those bright and able people who worked with me at the Exchange, and those who loyally followed me to Tradepoint.
Peter Bennett will be at the Mondo Visione Exchange Forum which takes place on 10th November in London.
