I just came across this story at Application Development Trends referring to a report from ROI-specialists Nucleus Research which:

found an underwhelming correlation so far, based on a survey response. Of 106 enterprises surveyed, just 37 percent indicated a positive ROI from SOA.

The article and, I presume, the research focuses on the software development and integration aspects of SOA, rather than the broader service lifecycle and it's the latter where the real business value of SOA lies.

However, it's not this big vs small thinking that's the real problem: it's the search for mythical SOA ROI grail. You can think about SOA a bit like the plumbing in your office or investment in methodologies. It's easy to explain their benefits and their costs but there is only an indirect link between cause and effect. We have to move away from the comparative safety of spreadsheet ROI models and instead get out there and really sell it to the business, as we discussed in our latest signposts newsletter:

The problem is that when it comes to business managers in particular, SOA is just too involved a set of ideas to try and get across. Regardless of how relevant the underlying concepts are to senior managers, it seems that SOA as a term is something that’s seen as very technical. It’s tempting to think, of course, that this is not the way things should be, and that what we all need to focus on is education.

This just isn't practical though. What's more productive is to separate the name from the concept. The solution lies in focusing on two concepts that are pretty well naturally understood in the business leadership community, and that are are absolutely related in practice to SOA - "service" and "process".

Most senior business managers have experience of having services provided by outside parties (facilities management; payroll processing; etc) and in some instances also have experience of providing or consuming "shared services". And we also find that many senior managers are also pretty sharp when it comes to understanding business processes and their management and optimisation. In getting buy-in for SOA initiatives, we find that quite often it's best to avoid "SOA" completely, and focus instead on helping business managers understand how the concepts of "service" and "process" in a business strategy context map down into more flexible and granular IT capabilities that can be developed, configured and reconfigured quickly as business service and process priorities change.

I am certainly not saying that the investment side of the equation can be ignored. It can't but costs need to be dealt with them in an incremental fashion, aligned with particular business benefits as an SOA initiative evolves. Trying to do that up front just doesn't make sense.