The computer that makes sure your salary is paid every month probably has less computing power than your phone. That’s the sad truth of the state of technology inside Brtitish banks. “Legacy systems” is the polite way of describing this millstone around their necks.
One of the things that always surprises me, coming from a tech finance startup, is meeting people from the big High Street banks who say, “We wish we could have the technology you guys have”. It’s surprising because we’re 14 people in a shared office in Shoreditch and they’re a global mega-corporation. We feel good about channelling £20 million to British businesses via invoice finance; for them, twenty million quid is a rounding error.
But, we run our system on modern computers. They have to rely on what one banking IT expert colourfully described to me as “1980s, green-screen pieces of shit”. Most bank systems at some point rely on a variation of the IBM AS/400 – machines introduced in 1988, older than most of the people who are trying to fix them. But the 1980s PCs are not the worst part: almost every banking system has to deal with parts coded in a language called COBOL, which was invented in 1959, and designed for vacuum tube computers like UNIVAC.
This all goes back forty years, to when computer technology was not only expensive, it often had to be custom-made. Due to the fact you couldn’t order off the shelf, you had to get engineers to literally build your mainframe, and then programmers to code it. The first banking systems were built like this: everyone had their own custom made system and it was incredibly expensive.
When new machines and systems arrived, the old systems, then five or six years old, were deemed too expensive to replace. Rather than replacing the old systems, they were simply connected up to the new machines. Since then, there has been a huge market in financial services for “middleware” firms to come in and make different generations of computer talk to one another.
The last major round of this layering of technology happened in the late 1990s – that’s why, if you peek around the screen next time you’re in a bank to see what your “customer relationship manager” can see, it looks like they’re working on something like Word Perfect 5. No one has yet taken the plunge, cleaned the house, and modernised their IT systems top-to-bottom.
This is further complicated by the wave of takeovers that created our current “too big to fail” banking system. Every bank had its own system – when two banks amalgamated, they patched the two totally different custom systems together with overnight fixes and workarounds, many of which are still in place.
These systems are quirky, to say the least. As one IT professional explained to me recently: “One system we worked on, the system command to end a program run [essentially turning off the machine] was a long string of number 9s. A lot of the fixes involved trying to convince it to accept payments with more than 4 nines in them – for example, £9.99 on the 9th of September”.
An odd quirk like this, with an old mainframe talking to newer systems, was at the root of the recent RBS meltdown that left millions of customers unable to use their accounts for as long as a fortnight (you can read a fuller version of what happened here, at The Register). It wasn’t the first major banking technology failure the world has seen; it was simply the first that the mainstream media in the UK noticed.
On New Year’s Eve last year, Lloyds TSB chip and pin machines double charged 200,000 transactions, because the antique programming language that underpinned the system went wrong; in Australia, in 2010, a catastrophic failure at National Australia Bank stopped customers being able to use their debit cards for over a week.
Creaky, old systems are breaking down more and more regularly, as more and more data is being pushed through them. The banks need to overhaul these old systems, before another catastrophic IT failure happens. Next time, it could be a lot longer than a fortnight that customers are stranded without functioning bank accounts.