What does the disastrous launch of Cahoot, the standalone Internet bank from Abbey National, say about the trade-off between speed to market and systems reliability?
Following its launch on Monday last week (12 June), the bank's Web site crashed on its first day, was unable to accept applications for accounts for most of its first two days and went down again on the fourth day.
Similar problems at Egg, the standalone Internet bank launched by Prudential in October 1998, suggest such crashes are an inevitable risk in an age when the company that moves swiftly into new markets reaps benefits that far outweigh the disadvantages of technical failure.
But the successful launch of the Co-operative's standalone Internet bank Smile in October 1999 showed that it is possible to come to market fast without crashing.
So what was Abbey National's big problem? The answer, at the end of Cahoot's first week, was that it was impossible to say because the bank still did not know exactly what had gone wrong.
This much Abbey National was admitting to. The core banking systems that customers used once they had been accepted were working fine. But the registration application, developed separately, was not.
That was not to say there was an identifiable problem in the registration software, merely that IT staff were monitoring it to locate the bug in the network. Customers were being allowed to register in numbers that did not crash the system.
One of Cahoot's technology partners, IBM, said it was not possible to test for every eventuality before launching an e-banking Web site. But testing firms have pooh-poohed the claim on the grounds that the registration application and its integration with the network should have been one of the top priorities for testing.
Take a look at what happened at Smile. Its testing company, Keane, applied the same disciplines that were being used in systems development years ago to the more complex multi-tiered systems typical of Internet banking systems.
It devised a testing harness that replaced the client side of Smile's architecture with a machine that mimicked the chaotic behaviour of thousands of users, based on a study of human behaviour. The company measured performance variables at all stages of the network - in the Web servers, in the application servers and in the mainframe - looking for cascade effects.
What were they looking for? One example might be incorrectly configured software that worked perfectly with a small volume of users but created serious bottlenecks when there were thousands of concurrent users.
Keane developed a bespoke testing tool for the bank, that Smile can use with new testing scripts, because the current range of testing tools could not cope with the two separate 128-bit encryption systems used in communication between the layers of software at Smile.
Not that using standard testing tools would have been cheap, with some suppliers charging hundreds of pounds per concurrent user simulated. To ensure that speed to market was not impaired, testing took place throughout the project - a practice that also aids application development.
One of the factors Keane stresses as most important to the successful launch of Smile was the involvement of the Co-operative Bank's strategy department in developing and specifying its e-banking systems. This strategy department had substantial previous experience of e-banking for the main Co-operative Bank brand going back 18 months prior to the launch of Smile.
Abbey National too had several week's successful experience with e-banking for its mainstream banking brand but chose to use standalone systems for the Cahoot brand which was designed to attract a younger, Internet-savvy customer base.
Even before Cahoot's launch, stockbroking analysts at Morgan Stanley Dean Witter were questioning the commitment of Abbey National's senior management to the Internet bank, saying that they had come away from an Internet strategy briefing with the impression that the bank's heart was not in it. This could be seen in modest targets for take-up and a decision to run the Cahoot business with just 56 staff.
However, an Abbey National spokesman said other analysts were more impressed by the bank's commitment to providing its services over the Internet. "It was clear at the Internet strategy event, which covered retail, business-to-business and Cahoot, that most of the analysts I spoke to were very impressed."
As it was, Cahoot's marketing department could not have dreamed up an incentive scheme for new customers that was better designed to produce a one-off surge of Web site hits than the one it chose. The incentive was 0% interest on credit cards and loans for one year for the first 25,000 people to apply for accounts.
One customer, who described his experiences using the online name Madwolf at the consumer finance Web site motleyfool.com, said he planned to use the overdraft for a year and then close it as soon as the special offer ran out.
Commenting on Abbey National's claim that it had experienced "unprecedented demand" for Cahoot's services the user said, "Since they had just opened it was bound to be unprecedented but given that they were offering 0% APR on credit cards and overdrafts it shouldn't have been unforeseen."
Following the disastrous launch of Cahoot, Halifax has announced plans to stagger the launch of its standalone telephone and Internet bank Intelligent Finance, by launching its telephone banking service first followed by its Web site 10 days later.
Abbey National may still not know what has gone wrong with its systems but other banks are already learning the lessons of its experience.
Cahoot seeks royalties on software sales
Abbey National's Internet bank Cahoot is planning to collect royalties on the sale of its e-banking software to other financial institutions. That is when it has worked out what has gone wrong with its systems.
Cahoot plans to collect the money from Fiserv on sales of Straight Through Processing. The software enables customers to perform banking transactions, such as applying for an overdraft, without human intervention from Cahoot.
While the source of Cahoot's problems are not known, Fiserv has pointed out that its core banking and contact software delivers industry-leading uptimes for other financial institutions such as HFC, ABN AMRO, Banco Comercial Portugues, CIBC and Birmingham Midshires. Fiserv did not provide the registration application that is under suspicion and Cahoot carried out its own systems integration.
A Fiserv spokesperson said, "It would be entirely inappropriate to criticise any of the partners involved in the project at this stage. Criticising Fiserv in particular would be like blaming Tarmac for the congestion on the M25."