I'm stating the obvious here but the underlying key element to services is the data. Agility, reuse, composibility and all that other jazz is difficult to achieve when you do not really understand the data. Chris Madrid has posted an article on Master Data Management over at SOA Magazine. I think Chris's points on the issues facing the enterprise are outlined very well. Multiple sources of the same or variations of the same data are the heart of the issue. Keeping them in sync and knowing which one to draw from can be difficult at best. The solution as outlined by Chris, is basically a central repository that serves as the master data. A very noble cause within the enterprise.
I think there are a few issues with this that some enterprises will run into. Does the enterprise actually have data architects? I going to throw out there that surprisingly few do. That can be a problem. Developers, DBA's and even Enterprise Architects don't always make for the best data architects. And even when they do understand it very well, it's usually not their primary focus. Moving to the Master Data Management concept requires resources that are focused on just that. This leads to the next issue which is buy in. Can you get the enterprise to agree on the concept and if so can you then get them to agree on the data model?
I don't think the issues facing Master Data Management are all that different from the concepts of a Common Information Model. It can be a large black hole for an enterprise. Attempting something like that a can turn into a long effort that ultimately doesn't keep pace with the changing business and is outdated before it's even put into place. I do agree with the concept though and I think the closer you can get to this state the more successful your SOA can be.