PC networks are taking over in areas where mainframes were once thought unassailable.
Periodically, the IT industry convulses and produces a new technology. The most recent example involves two aspects of the industry. First, downsizing, where minicomputers replace mainframes and PC networks replace mini-computers. Second, the introduction of more powerful languages, capable of achieving, in a single line of program code, what older languages need dozens of lines to do. The combination of the two is likely to transform the industry.
Casualties are already mounting simply because of downsizing. IBM, in particular, is suffering as customers switch from its products to smaller and cheaper offerings from other vendors. After announcing 1992's $5-billion loss - the largest in corporate history - IBM's chief executive John Akers was forced out earlier this year for failing to formulate a viable counter strategy.
Until now, mainframe manufacturers have taken comfort from the belief that their core market - the giant corporations who were the first into computers 30 years ago - was unassailable. It was inconceivable, for instance, that a bank or building society would throw away its mainframe and rely entirely on PCs. This was because many in the IT world were convinced that only mainframes had the capacity and required robustness for what is termed 'on-line transaction processing' (OLTP).
OLTP operations differ from normal commercial functions, such as invoice processing, in that most of them have to be carried out rapidly. A bank, for instance, may have thousands of receipts, withdrawals, statement requests or balance enquiries to process at any given time. Customers must not be kept waiting. Furthermore, it would be disastrous if, due to a breakdown, these operations were brought to a standstill. Only mainframes, ran the argument, offered OLTP-using organisations the required processing power and reliability. The banking sector has invested heavily in IT on both sides of the Atlantic. Staggering sums have been spent in pursuit of productivity gains or services that would differentiate them from their competitors - witness the Bank of Scotland's home-banking experiment, or the Midland's First Direct 24-hour dial-in service. According to Ernst and Young, the top 35 US banks spent over $10 billion on IT in 1991. The benefits have been slow in coming - home banking has yet to take off, and 90% of bank payments still involve paperwork rather than pure electronics.
Flexibility is also an issue: bankers bemoan both the cost of their automated systems and the constraints that result. Stories abound of branch managers being unable to fulfil customers' perfectly reasonable requests because 'the computer won't allow it'. To keep customers, some banks have been obliged to build in expensive manual sub-systems to bypass the computer.
Change, however, is on the way. A number of banks and financial institutions worldwide are now seeing costs slashed and flexibility enhanced by a full-function OLTP banking system running on PCs. Unfortunately, getting approval for the idea in conservative British financial circles has proved an uphill struggle. Ray Moy, director of financial software specialists Apak Systems in Bristol, says, 'It's still hard to get people to accept that an ordinary 486 PC can support hundreds of bank terminals.'
But Moy's proof of the system's worth is incontrovertible - even if you have to travel to Canada to see it. Richmond Savings Credit Union in Vancouver is a bank-cum-building society whose asset base ranks it among the top 10% of North American financial institutions. The company's entire computer processing is carried out on a PC network - the largest such OLTP network in the world. The dozen or so industry-standard PCs at its heart handle 100,000 transactions a day for some 50,000 customers. It also has on-line links to credit card operators, ATMs (cashpoint machines), touch-tone banking and credit-check bureaux. The story behind the system is one of luck and foresight. In the early '80s, four colleagues working for a Canadian computer company became excited by the growing potential of the IBM PC - and particularly the Intel 80286 chip used in the 'AT' version. Realising the scope for downsizing that the new technology offered, the four tried to persuade their bosses to seize the opportunity. But their ideas were ignored. They left in l984 to set up Prologic Computer Corporation in Vancouver. They decided they needed a new high-level language specially designed to exploit the potential of networked PCs. And they set about writing one which they intended to market as a development tool. However, while the new language - called Probe - was being created, they were approached by one of their former employer's customers - Richmond Savings. The firm, hard hit by economic downturn and the victim of inflexible systems running on an outdated mainframe, was desperate enough to gamble. Could Prologic create a banking system for them with Probe? Such a request would normally have been treated as a joke. If, for example, a new banking system was written in Cobol, a standard mainframe language, it might take decades and mean writing over two million lines of program code. The risks involved in pioneering such an application were enormous, and time was of the essence.
Allen Lacroix, Richmond's vice president in charge of technology, helped plan how the bank would manage the risk.The basic strategy was to develop the system in three stages. The first step - desperately needed - was to get a better understanding of the bank's customer base. Prologic were given the task of creating links to the data on the mainframe and building a decision support tool for the bank's marketing and sales people to use. Which products were selling - and to whom? Which other customers might be interested? Should such-and-such a customer be given the money for a mortgage? 'It paid dividends straight away,' says Lacroix. 'We at last started to understand our customers.' Encouraged, they moved on to phase two: replacing all the bank's redundant terminals with PCs hooked into the mainframe; and running wordprocessing or spreadsheet programs on others. Again, says Lacroix, the objective was to minimise risk: 'If it all went horribly wrong, at least we'd be able to keep the PCs.'
Phase three was the creation of the actual banking transaction system to bridge phases one and two. Transactions on branch PCs would instantly up-date customer information obtainable from those same the PCs. The emphasis, says Lacroix, was on step-by-step evolution with milestones agreed and payment made when Prologic had delivered the work. Relationships were at times strained. Lacroix is blunt about the extent of Richmond's leverage, and the extent to which he used it. 'Prologic had only one customer - us. We were their future.'
It is undeniable that Prologic secured Richmond's future. The group had given it flexibility and responsiveness and, therefore, says Lacroix, an edge on their competition. In the six years since the system was installed, market share has steadily grown. With twice as many customers, operating income has increased 850%, assets have grown from $400 million to over $1 billion, the number of branch locations has doubled and (thanks to better-informed lending decisions) the level of bad debt has dropped sharply. IT costs, not surprisingly, have also plummeted. 'We didn't go into it to save money - but we have anyway,' says Lacroix.
Richmond's enthusiasm has proved infectious. Over 100 other banks and financial institutions have adopted the system, now marketed by Prologic in its own right. After visiting several of the sites, Apak Systems' Moy decided last year to market the system in the UK. Already the fledgling Bank of Edinburgh has signed up. Ironically, the mainframe company DEC referred the bank to Apak. Moy is bullish about the system's prospects, but realises that the concept may take time to flourish. Only then will they be able to reap the full rewards.
If the analyses now emerging from research institutes are anything to go by, they will not have too long to wait. Industry watchers, the Gartner Group, forecast that the OLTP total market will continue to grow at around 7% a year through to 1996, reaching annual sales of around $130 billion. The estimated proportion of this destined for mainframes will shrink sharply, from almost 50% now, to around 35% in 1996. Meanwhile, the proportion going to PC networks should virtually double - bringing it level with the mainframe sector's market share. After that, the future belongs to the PC.