But, as I've noted in my previous columns, there's a real need for organizations to also think about their information supply chain and their underlying data, particularly when they start to implement SOA.

Information supply chains encompass the information and data flowing into a company a wide variety of sources, the storage and management of the information as it flows into internal systems (such as PIM, ERP, inventory, order management and other systems) and use of that data and information across a wide variety of channels (such as online commerce systems and POS systems), both within and external to a company.

For illustrative purposes, let's look a little closer at one specific segment of the information supply chain-the part that deals with product data. SOA and current business requirements are making it even more important to mange product data and information supply chains more effectively and efficiently. However, one of the problems is that product data isn't necessarily easy to manage.

In fact, for many organizations, if product data needs to be assimilated and shared across systems, they have a problem, since product data can come in so many different formats, categories, and forms. As a result, in most cases, product data integration isn't easy. But it doesn't mean that companies aren't trying to do it. Let's look at some existing solutions:

One approach to managing product data is through manual effort. While this approach can work, it's expensive, not very scalable and should typically be considered a one-time fix, especially when organizations need to manage tens of thousands or hundreds of thousands of products. Unfortunately, it's very easy to get out of date or out of sync with manual product data management approaches, since it takes so much effort.

Custom code is another alternative. If the product data is very simple or very consistent, this can be a reasonable solution. However, like the manual approach, it can be expensive to develop and hard to maintain. It is also likely to deliver a high percentage of incorrect information as most pattern-based systems are very poor at correctly detecting and interpreting context.