|
On the retail side of financial services, delivery times are usually much faster because the competitive nature of the market demands us to forever be one step ahead of the competition. The down-side to this is that solutions are usually not really at the 95% confidence limit. The 95% confidence limit is a statistical expression of the readiness of the solution to be deployed. In the case of the Internet for instance it may be the level to which the security protection at the website is strong enough to keep systems secure. At the application level, it may be the proportion of the functionality that has passed testing. This latter may seem strange but most applications today have to be released with a number of both known and unknown bugs. Technicians continue to work on the code to eradicate known bugs and wait for users to find the rest.
So managing technology in this environment has more to do with mechanisms which minimise the hassle for the user - by having the main functionality operative, and a reporting-fix-patch cycle that gives users confidence to keep the applications. The 95% confidence limit is defined in the business specification for the project as a key numeric deliverable. In other words, it is a constraint on the technologists and in particular quality control, not to release the project to its users until it can meet certain expectations. Again, in application delivery this might, for example, be a screen response speed.
For the technologists this would translate into systems and server architectures - how fast are the connections over which the application must run; how fast are the processors as well as how the application itself is structured. If it is a database for instance, the way in which tables are structured can affect performance in indexing and access. Any one of these or a combination can cause a failure in quality assurance and resultant overrun on either budget or time frame of both. Business people can only articulate in resultants; hence common 95% confidence statements such as ‘sub-second response time required'. Of course, any business person is going to come up with end-point suggestions ‘I want instant response'. It's the job of the technologists and in particular the project manager to evaluate and communicate the balance between what the business wants and what the cost is. The result on the retail side is a methodology known as the domino effect. This occurs when a solution does not meet the 95% confidence limit and is released anyway. The solution does not immediately fail; nor does it fail catastrophically, but it does fail just enough to tip the project into a reactive stance.
Proactive stances are commonplace in most technology developments in financial services. Software versioning is a typical example. This allows the technologists to indicate that the product has more developments or improvements to come which will be delivered as different versions. Reactive stances are unfortunately just as common. These show up as ‘un-foreseen' failures that result in technologists having to ‘shoe-horn' new releases into a series of releases as ‘bug fixes' or ‘patches'. In non-application level developments these may evidence themselves as changing suppliers (because connectivity resources failed) or non-standard ‘testing' and ‘roll back' of servers out of the normal sequence. All tip the next domino, putting the standard development methodology onto a knife-edge of always trying to catch up. Resources that were well planned to deal with a strong development are now not enough to deal with the domino effect with everyone running around trying to fix things and get ahead of the curve at the same time.
So, best practice is primarily about communication. Unfortunately the two parties concerned rarely speak the same language and need interpreters, more commonly known as project managers. Best practice, almost by definition, requires benchmarking. As we've seen earlier in this article however, best practice is not always something that easily lends itself to the numeric side of benchmarking.