Back in 1998, I was doing a Masters degree at Leeds University. An IT Module on my course posed a question something like “What is the case for transactional websites in UK retail banking?”
Do you remember 1998? It was the year Netscape announced they were going open source. Home broadband did not exist in the UK. The vast majority of people did not have a home internet connection of any type and probably almost no-one over the age of 25 regularly used the internet or had a personal email ID. There was not one single UK bank offering meaningful transactional capabilities on their website. The internet (or should I say Web 0.1) was only just taking off.
Polls at that time were concluding that only about a third of the British public expected to have online retail banking within 5 years with some really tech savvy people (13% of the population) wanting it within 2 years! Wow!
About 3,000 words (on paper) later I concluded that there was no current short term ROI case for fully transactional internet banking but it was blatantly obvious that it would take off and any bank that didn’t take the leap of faith and invest, would get left behind both in customer service and cost per transaction.
This unremarkable conclusion somehow managed to get me a good mark for that paper – it seems even more unremarkable today.
So is SOA in the same boat then today, being blatantly obvious but needing a leap of faith? I don’t think so.
Whilst I agree that the justification of any IT initiative should be an enterprise challenge (not an IT challenge), I don’t see the point in SOA unless it delivers ROI. If that ROI is delivered via creation of an “agile enterprise” this is great news but why can we not anticipate the advantages that brings and quantify them in measurable business benefits?
In my opinion a much better route is to take the bottom up approach, solving problems one by one at a local level. Each one contributes ROI in its own right (or it doesn’t get done), but adds to the central pool of “services” or perhaps “business processes” is a better description.
This apparently anarchic approach sits very uncomfortably with those who must have a fixed big picture to aim for. However, business can no longer predict more than a few weeks ahead so what is the point of a big picture – by the time you deliver the big picture, your picture is not one the business can use any more.
If SOA is supposed to deliver the agile enterprise why cannot it not be implemented in an agile, and incremental (and therefore cost effective) way?
In this debate I find myself siding with McKendrick’s colleague Dana Gardner, see Get real: ROI is a necessary passage for SOA — SOA can show ROI from the get-go on a project basis.