Middleware Matters

I think the classic metric is that 50% of all IT projects fail to deliver all of their functionality on time and on budget. This was based on a recent Gartner report, reported in Comuputerworld.

So if the failure rate of SOA efforts is only 13%, they've found a magic silver bullet.

Since I've read the mythical man month, I don't believe in magic bullets; SOA is likely to hit the same failure rate (+/- some small percentage) as every other architectural technique invented since 1960...

While I tend to agree with Steve’s points about the failure rates quoted, I think it is dangerous to get too complacent about SOA as necessarily delivering greater success. For instance, AMR has survey results that show 36% of SOA adopters find themselves "unable to reconfigure business processes as needed" and that amongst organizations that have *not* adopted SOA, only 13% have difficulty with reconfiguration of business process.

It is scary of course because SOA is meant to delivery agility above pretty much anything else. It is no good claiming the results are wrong simply because you don’t agree with them – we need to look deeper. My own view (as an advocate of SOA and vendor of SOA-based products) is that this is because there has been an over-emphasis on SOA as abstract architecture to the exclusion of how you actually implement the architecture.

It is not sufficient to embrace SOA as an agile architecture without selecting a product that can also deliver on this agility. If the industry believes that a maturity model (for instance) on its own delivers the benefits of SOA, 2006 will the year of SOA disaster stories. I have just blogged about this very topic... (http://polarlake.typepad.com/ronan/2005/11/the_gap_between.html).

To the points already made, we are not being realistic if we think that many of those looking to leverage SOA technology today at a project level are not falling on their faces, at the very least due to inexperience and perhaps some inflated expectations from vendors. I feel this trend will continue. Again, not as much as a function of the missing pieces in the technology but the natural learning curve of the end user. We’ve seen this pattern many times before.

I do believe the survey. It’s pretty much what I’ve been seeing as well, albeit informal data point. I also think this is to be expected.

By the ways, here is my take on why these projects are failing. http://weblog.infoworld.com/realworldsoa/archives/2005/11/why_soa_fails.html#more.

I don’t want to start a discussion on the reliability of the methods used the research. (They are not. There’s that, I’ve said it anyway) I’d just like to remind that even with reliable research method it always comes down to what do you consider a success and what is your viewpoint. I’ve often seen projects that were highly successful from vendor perspective, have been complete failures when viewed from customer perspective. (This especially applies for product vendors.) Likewise, I know projects that have been (at least partial) failures from the vendor perspective, but have been highly successful from customer perspective. The real analysis on whether or not a particular project was successful can not be made until the project has finished, it’s deliverables are deployed and the real benefits can be analyzed. How many SOA projects are in this stage of their lifecycle?

A senior program/project manager whom I have a lot of respect for, has an excellent track record and is extremely good at his work once told that 100% of his projects have been failures. He said that he has never had a project that would have been 100% in time, 100% in budget, would have fulfilled 100% of the stakeholder requirements and expectations, would have behaved according to ROI and TCO calculations, etc. There are always some areas where you have to compromise during the project, and these compromises can be seen as partial failures by (at least) some of the stakeholders. He did not even expect that a project with 100% success would be possible. (In fact he said that if there would be such a project he would not want to work for it, because it would probably be dullingly simple.)… Long term research seems to indicate that 70-80% of IT projects fail. Why would SOA projects be any different? Especially when SOA projects are probably conceptually much wider and complex than IT projects in general.

Why should one be surprised by the stats? SOA is not a solution to all your problems. I have seen people who want to implement SOA for an application which is ideal candidate for desktop.

IMHO SOA is a better way of writing procedural code. And the reason being standardization provided by XML. In years to come it will travel the same path as other technologies. There is too less inteligence to replace majority of thinking. ;-)