Product information is one of the few remaining strategic information assets within an enterprise that does not have a dedicated system of record. This is not surprising when one considers the complexities of product information management (PIM). Specifically, a comprehensive product information management application must be able to handle structured and unstructured information, share data, accommodate a variety of constantly changing data models, manage multiple inter-relationships, and enable multiple hierarchies, all without duplication of data. While enterprises and vendors alike have turned to content management and relational database architectures to address the challenges of product information management, these approaches have proven to be very limited and far less than satisfactory.

Recognizing the shortcomings of traditional approaches, several leading companies are now utilizing an emerging type of adaptive XML application to more effectively manage their product information. This new adaptive XML approach allows management of both unstructured and structured data and also provides extensive and rich functionality. Rather than limiting XML's use to inbound/outbound transformations, this approach uses XML as a core data model for the application. Traditionally, methodology for building applications utilizes compile-time data models; this makes the application rigid and tightly bound to the data model. The adaptive application utilizes an XML-based runtime data model and is used in all tiers of the application. In short, an application that is built using this approach easily handles the complexities of product data management and adapts to changing data models, which allows further extension of the capabilities beyond those of traditional applications.

Product Information Management and Its Challenges
Generally speaking, product information management (PIM) refers to a system for managing all types of information related to products. A PIM system needs to provide a system of record for products throughout the enterprise and maintain that information in a centralized location. In addition, it must provide rich functionality to address the challenges specific to product information.

Product information is unique in the variety of different data types associated with it. The information could include part number, price, and size (structured data); images, drawings, and merchandizing text (unstructured data); or inventory data (transactional data). To access, manipulate, and store each of these data types and maintain data integrity is a challenging task.

Besides managing different data types, product information systems must conform to unique and changing schemas. To be an effective long-term solution, a PIM system must accommodate unique data models for the enterprise as well as the product and allow for changes to either of those schemas as required. Frequent changes to product information in today's rapidly evolving marketplace make application maintenance a burden on most IT resources.

Product data must be organized and classified into taxonomies that consist of an arbitrary hierarchy of categories and subcategories. With product data, it is common to have multiple simultaneous hierarchies. A PIM system needs to support these without duplication of data.

Also, products will often share data such as warranty information. This becomes a daunting manual process, particularly when shared information is changed or multiple new products are created. It is critical for a PIM system to automatically share data changes in all places where reusable content appears, thereby avoiding costly duplication, data inconsistencies, and time-consuming maintenance cycles.

The organization and sharing of information between different products and product types creates unique relationships that must be maintained within a PIM system. As product needs change, the creation of kits, bundles, and accessories is common with new, even unknown, merchandising relationships consistently being created. Referential integrity of the complex data relations must be maintained even as management operations dictate change.

Other required functionality, such as search capabilities, data validation, and synchronization with trading partners and external data stores, is indicative of the additional requirements of an effective PIM solution.

Traditional PIM Solutions
Sometimes product information management is also called product content management. This usually suggests a predisposition toward enterprise content management (ECM). An ECM system was originally designed to manage large volumes of documents and Web pages. In its use as a PIM system, product data is maintained as a set of documents, which is excellent for unstructured data. In order to manage structured data or attributes, however, the system uses name-value pairs, which is common to all content management systems. While an ECM system enables the support of a single hierarchical view of product category data in the form of folders and documents, it does not handle multiple hierarchies well; unfortunately, the support for data sharing within an ECM system requires data duplication.

A more traditional PIM system is a relational database management system (RDBMS) in which product data is stored as a set of tables and records in a single central data repository. Structured components of product data are managed well with this type of system, but unstructured items must be stored in database fields as blobs, limiting some field sizes. Further, an RDBMS requires predefined database schemas in the form of tables, which impose limitations on the type of product schema that can be supported. Therefore, large sets of unused fields are included in the database to account for custom data requirements and changing schemas. Hierarchy organization and complex data relations are not handled in this type of system.

Adaptive XML Application as a PIM Solution
The extensive use of XML throughout a PIM system has only recently come to the forefront as a viable and competitive alternative to traditional PIM solutions. In the past, XML was simply used for inbound and outbound transformations. Utilized in the correct fashion, however, XML can be used as the foundation for a far more powerful adaptive product information management application. This new adaptive XML approach is quickly gaining ground as the most effective means of managing product information at some of the largest and most progressive companies in the world.

Until now, structured data has commonly been represented as records in RDBMS, and unstructured data has been stored as documents. However, XML provides an ideal means of representing both structured and unstructured data, as shown in Figure 1. XML is therefore quintessential in handling complex product information that consists of both structured and unstructured information.

While the management of different data types is an important advantage in a product information management application, it is now possible to go even further with XML to create a fundamentally new approach to application architecture - an adaptive application. In an adaptive application, the three tiers adjust to runtime changes of data schemas as shown in Figure 2. Such a system has no prebuilt notion of data schemas. At runtime the appropriate schema is located and interpreted for a given piece of product information. All the computations needed for a given schema are also applied at runtime.

All tiers of an adaptive application rely on the open standard W3C DOM as the only data model. The W3C DOM is a runtime data model that represents XML DOM as the common data model across all tiers of the application. This eliminates the need for data mapping and transformation commonly found in n-tier applications. Also, the runtime interpretation of schema means the system can work easily with custom data models, causing a significant reduction in product implementation time.

Technical Considerations and Concepts
One of the fundamental concepts that must be changed is the notion of XML as a document versus a fragment. This is an important distinction that has a profound impact on the flexibility and functionality of an adaptive XML application. Today, XML is commonly thought of as a document with tags. This impression is problematic when it comes to storing structured data in the form of records. If a single document represents a single record, too many documents would be created and become difficult to manage. Alternatively, if multiple records were represented in a single document, then an entire set of records would have to be manipulated each time a single document is sourced.

However, an alternative to this is the concept of XML as a fragment, with individual fragments accessed by value or reference. This enables read/write access to the fragments, with the fragments stored as either records in a database or as an XML document. This successfully manages both structured and unstructured product data whereas more traditional PIM systems falter.

Once XML is handled as a fragment, the management of these fragments must be considered. To effectively manipulate the XML fragments, W3C standard Xpointers can be used as XML fragment locators. Fragments can then be located within any system - file, content management, database, or legacy systems. The combination of XML fragments and Xpointers provides a complete uniform access layer to all enterprise data and business objects.

XML validation is another challenge that must be addressed with respect to XML fragments. In the XML world, XML schemas are used to validate XML documents. XML document validation relies on its ability to locate schemas. The schema location directives are specified at a document level. The use of XML fragments versus documents therefore creates a conundrum because there are no documents to validate. An alternative approach is to store XML schemas in a library and compute them as needed for a given XML fragment. To locate a schema for an XML fragment independent of its document, a schema locator service needs to be built. Locating the schemas for an XML fragment at runtime easily avoids hardwired data models in the application. This approach also enables schema changes to be detected at runtime, making applications resilient to data model changes. Given the constant, even unknown, changes to product information requirements, both internally and with trading partners, this ability to adapt to changing schemas is absolutely essential to providing a solution that meets the enterprise's current and future needs.

Because product information management inherently consists of taxonomies and complex data relations, XML must be able to manage these relationships effectively to be part of a successful PIM system. Xpointers can be thought of as providing the static relationship between two objects - a single pointer referencing another object. This technique can be used to manage multiple hierarchies without duplicating data. Referential integrity constraints can be easily enforced using simple link tables. Use of Xpointers in conjunction with Xlink provides considerable flexibility in managing object relationships.

The combination also provides a powerful tool for creating and managing multiple hierarchical views as well as complex data relationships involved in merchandising and data sharing. In addition, as these relationships are not expressed in relational tables, manipulating them becomes effortless. The links and pointers can be easily created, redirected, or removed. Sharing information is similarly straightforward, as links or references are directed to existing data, preventing time-consuming and expensive data duplication within the PIM system. A change in the shared data allows all information to be updated automatically.

An adaptive XML approach also has significant advantages in terms of the client layer. Any XML-capable authoring tool such as Excel, Word, PDF, and HTML can be used to easily interact with the application. Working with the product information management application via a spreadsheet is as simple as working through an XML-based client.

Adaptive XML vs Alternative Approaches
An adaptive application enables better management of structured and unstructured product information, as well as providing superior data model and data relationship capabilities. Table 1 summarizes how an adaptive XML approach provides superior product information management over alternative approaches across a variety of business requirements.

Bottom Line
Product information requires a very specific, flexible system to enable maximum functionality in an enterprise. An adaptive XML-based application is proving to be the future of any enterprise's successful PIM system. An adaptive XML-based product information management application is able to work with both structured and unstructured information, as well as quickly and easily incorporate changing product data models, product types and complex product data relationships. This flexibility will become a required feature of any PIM system as product information requirements continue to evolve. However, it is important to be wary of architectures that claim to utilize XML in their PIM system - often it is used only on the inbound or outbound transformations, limiting the full potential that XML has to offer in meeting requirements of an effective PIM system.

Resource