Our first reaction to SOA was considerable excitement. Finally, we thought, here was something that would prevent us having to re-implement applications continually, and would help us to link things that go across business processes more easily.

Our first priority was to find out what our main vendors, principally SAP, were doing in this space. We have a large SAP estate and, given this, are dependent on what the company does and how it exposes its underlying applications as services to allow us to take full advantage of any flexibility SOA could give us.

The general rule at Unilever is not to develop applications in-house unless absolutely necessary. In recent years, we've replaced a lot of homegrown software, and now have a highly packaged software landscape. So, if we're going to provide services on top of this, we must have access to underlying data and objects under SAP. But to do this efficiently requires us to upgrade to the latest version of the software.

As a consequence, at Unilever, we're not heavily into SOA for its own sake. We've done very little because we have made the decision to go at the pace of our main packaged software provider and have not yet seen the big paybacks elsewhere that would persuade us to change this view. But we have done some very basic work on it in a couple of places.

I still believe in the original vision of SOA. For us, using a packaged supplier means they deliver the tools that make the underlying complexity, such as granularity and re-usability, less of an issue. When you look at some of the areas this plays out - pulling customer information out of sales and order processes and cross-selling via a CRM environment - it means the information you need at your fingertips when customers call is already there.

SOA is not going to be mainstream for us in the next three years. It will be in the period after that. If, like Unilever, you are using a packaged supplier, it's as much about the speed and cost of the upgrade cycle as anything else. While it is our decision to follow the vendor pace, they often don't appreciate that the cost and effort to upgrade is colossal. We are not going to implement an upgrade across 140 countries in a few months - it's a much longer journey than that.

If you talk to some banks, it's a different story, especially where they write their own applications and can expose legacy applications (which they won't want to change in their entirety as their real-time business runs on them) as services in their architecture.

But if you talk about services as a critical part of business architecture - service-orientation if you like - it's a different story to pure SOA. In that sense, Unilever is all about service orientation, whether it's integrating a full scope BPO or buying a new SaaS offering to support a specific process. But for IT, if we can get it right, SOA is ultimately a bigger prize.

For more on the SOA debate, read the viewpoint of Anne Thomas Manes, vice president and research director, Burton Group.