Creating Frameworks for Organizing Information
Portal designers can learn much from architects and builders. Well-designed buildings are easy to use and structurally sound. We can find what we want, components like doors and windows appear in logical places, and, most importantly, the building stands up over time. We cannot go into a building and find the structural integrity the same way we can find the heating and ventilation units or the corner office. Structural integrity is a property of the way the building was designed and constructed; it is not a single feature added at some point in the construction process. The structural integrity of a portal is similar to that of a building. It is a fundamental property of the portal design, reflected in turn in visible characteristics, such as ease of use, functionality, and reliability.
In this chapter, we look into structural integrity from a user's perspective. The core question we address is “How will portal users find what they need?” Actually, we break this question into a number of more specific questions to which we can provide general but concrete answers.
First, we discuss how to organize information on a page. This may sound insignificant compared to other challenges that await us in portal development, but poorly designed pages hamper the portal's adoption. Next, we look at design patterns for logically grouping related content and applications to provide a sense of context for our users. We can all appreciate the sense of being in a particular section of a department store and knowing in general how to find other sections. We should provide something analogous for portal users. Without contexts users can easily become lost in an apparent jumble of hyperlinked pages. Finally, we look at specialized techniques (such as taxonomies, faceted content models, and visualization) that can aid navigation, especially in large and diverse portals. A case study shows how visualization and logical restructuring techniques improved customer care services for one organization.
Much has been written about usability and Web design techniques, and this book does not try to add to these well-discussed areas. The main concern here tends more toward architectural issues, which sometimes abut or even overlap with usability issues. For questions about usability and design layout, I defer to any of the well-written books on the subject such as Don't Make Me Think by Steve Krug and Roger Black  and Designing Web Usability by Jakob Nielsen . I will address the types of structural elements required in well-designed portals but won't try to describe the finer details of their layout, formatting, and other visual elements.
The Need for Structure in Portal Interface Design
When considering portal interface structure, it is useful to distinguish between the visible structures and the underlying structures. The visible structures provide the organization reflected in the designs of pages, groups of pages (known as subsites), and the entire portal itself. These structures are readily apparent to users.
The underlying structures are core services, such as authentication, access controls, and metadata management, as well as the policies and procedures that govern the evolution of the portal. These structures are not necessarily visible when they work well, but their absence is all too apparent. When users cannot work with essential applications because of access control problems or when navigation tools direct searchers to inappropriate content because of miscategorized metadata, users become all too aware of these underlying services.
Page-level structures include the distribution of content, applications, and navigation tools. Many pages use the basic three-panel structure shown in Figure 1.1. The top area contains global information about the site, the left side area contains navigation controls and links to commonly used objects, and the large central panel is home to the substantive content of the portal.
Figure 1.1. Many portal pages use a basic three-part layout.
The global area is consistent across the portal and often provides links to a home page, contact information, accessories, or other frequently used applications.
The navigation area provides a localized context for users. If you went to the human resources area of a portal you would expect to find navigation links to training, policies and procedures, benefits information, and related material; in a health and safety area of the portal you'd expect to find information on material safety, accident prevention, and reporting procedures. The role of the site navigation area is to provide an immediately visible and easily accessible path to related components in the portal while keeping the user from being overwhelmed by the full breadth of the portal.
There are several common approaches to organizing the navigation area. First, the area can be organized by subsite or neighborhood. The CNN Web site (http://www.cnn.com), for example, uses this approach by consistently listing subsites (such as Weather, Politics, Business, and Sports) in the navigation area. A variation on this model is to display subtopics when a topic is selected. A third approach focuses on tasks rather than content and is more appropriate for portals or subsites oriented toward content management. Yet another approach is a hybrid that combines content-oriented with task-oriented links. Care should be taken to clearly distinguish the two types of links, remembering that the purpose of the navigation section is to provide a sense of context. Intermixing content and task links could make it more difficult for users to perceive their location within the portal.
The main content area delivers the core information and application access that the users seek. By framing this information and the applications in navigational frameworks, you provide users immediate access to locally related topics as well as global landmarks, such as the portal home page.
Grouping Pages: More Than One “Right” Way to Do It
An organizational model describes how entities are related. In the case of a portal, it describes how content, applications, and other resources are made accessible to users. The simplest model offers hyperlinking without restrictions. In this case, any page or resource can provide links to any others. This is the general model of the Web and the de facto organizational scheme for ungoverned intranets as well as the Internet. The advantage of this model is that decision making is completely decentralized so anyone can add content at any time to any part of the intranet. The disadvantage, so clear from the World Wide Web, is that this organizational scheme provides no point of reference for users. For example, if you find yourself at a page in such an intranet, there is no absolute reference point such as a home page or directory. All pages are equally important with regard to navigation, as depicted in Figure 1.2.
Figure 1.2. Simple hyperlinking schemes provide no point of reference.
Fortunately, most sites are no longer so freewheeling that we are left to navigate without some fixed references. Hierarchical organizational schemes provide an organizational structure with a top-level starting point and one or more levels of content. The simplest form of hierarchical organization is a tree with a root and links to lower-level pages. For practical purposes, most portals and Web sites also link across the hierarchy, as shown in Figure 1.3.
Figure 1.3. Even hierarchical patterns need to cross-reference nodes to support the way users navigate a site.
This type of nonhierarchical linking is required because simple hierarchies do not adequately model the way we think about information and how to find it. Consider a general topic such as “wireless phones.” Where should this fit into a hierarchy? Some possibilities include:
Business > Services > Telecommunications > Mobile Services
Business > Office Products > Phone Systems > Wireless Phones
Consumer > Telecommunications > Wireless Phones
Each hierarchical categorization is reasonable; the most intuitive one depends on the context of the search. If the user is looking for a mobile phone service, the first categorization above appears logical; if he or she is interested in purchasing a wireless phone, the second makes the most sense; and if the user is looking for a wireless phone for home, the third categorization is the most likely to be followed. Clearly, a single hierarchical structure is not sufficient for even moderately complex portals. More importantly, there is no single correct answer about where to place a page in a hierarchy, so don't bother trying to find one.
Organizing Multiple Ways with Facets
Instead, use a more flexible approach to navigating between pages. Multifaceted organizational schemes avoid the problem of hierarchies by accounting for the fact that an entity such as a product, application, or Web page can be classified along multiple dimensions. For example, let's assume a user is searching for a wireless phone, for less than $150, in silver or black with multiline support. Ideally, the user could navigate to information about phones, find wireless phones as a subcategory, shift to navigating by price, select the $125–$175 category, and finally narrow the search by color and functionality. In this example we have four facets: product category, cost, color, and feature set. When designing an information model, it is best to consider several facets or dimensions along which content is organized.
Metadata about categorization and content classification constitute facets or dimensions for organizing content. For example, a document published by the Health, Environment, and Safety Department on the proper disposal of chemical waste may be an official policy, published on a particular date, constituting compliance with a government regulation and broadly categorized as a safety document. The document type (policy), publication date, category (safety), and regulation status are all facets or attributes useful for organization and retrieval. One way to think about facets is as dimensions in a multidimensional space. Figure 1.4, for example, depicts the location of documents in a multidimensional space.
Figure 1.4. When organizing documents by facets, documents can be considered points in multidimensional space where each dimension is a facet.
Facet-based information retrieval can help users target specific content more quickly than simple keyword searching or navigation through a directory. Facet-based searches should allow a combination of keyword searches and attribute searches, such as searching for all “policy” type documents that contain the phrase “toxic disposal” and were published between May 1, 2003, and July 1, 2003. This technique is especially powerful when working with product catalogs that list items described by several dimensions (e.g., cost, size, color, feature set).
Dimensions can use a list of values, a range of values, or hierarchical values and in this way is similar to many online analytic processing (OLAP) tools. Relational taxonomies can model hierarchical values, but continuous value attributes (e.g., cost, time) are best modeled with scalar variables such as strings, numbers, and Boolean values.
The dimensions should reflect the ways users describe or understand content since organizational structure, like a search tool, is a key method for facilitating information retrieval. A particularly large site, FirstGov (http://www.firstgov.gov), the official site of the U.S. government, uses multiple facets, including:
The main page provides links to citizen, business, and government audience channels. The target pages of those links then list options by service. The service page in turn organizes content by topic. FirstGov's organizational model uses multiple facets and mixes those facets within paths from the home page to the content pages. This approach works well when you have information about usage patterns and frequently accessed pages. Analyzing log files from Web servers can provide key information about the most frequently accessed pages. By analyzing and grouping those pages, you can develop a rough categorization scheme based on facets.
Understanding how users think about the content and other resources in a portal is essential to developing a logical organizational model. Neither free-form links across a site nor rigid adherence to a hierarchical structure will serve the user community. Multifaceted organizational models provide the organizing structure of hierarchical systems and some of the flexibility of free-form linking within a controlled framework. Later in this chapter, we examine how complex facets can be organized using taxonomies.
Flexible Organization with Navigation Sets and Other Design Patterns
Another approach to organizing links is to use a hybrid of the hierarchical and free-form hyperlinking approaches. With this technique, we make a decision that some dimension is more important than others, such as the organizational structure of a company or the categories of products. The hierarchy is based on this dimension. Within each branch of the hierarchy pages can be linked as needed to other pages in the same branch. The advantage of this approach is that it allows users to quickly find high-level topic areas (such as the human resources section of a company portal or the camping equipment offerings of an online store) while still allowing the site designer to customize links between related pages. These relatively closed-off areas of related pages are called navigation sets. Figure 1.5 shows this common navigation pattern.
Figure 1.5. Navigation sets are groups of highly linked pages with relatively few links outside the neighborhood. Exceptions include links to home pages, contact information, and other global content.
A number of other patterns have evolved along with the development of the Web. These patterns provide a sense of well-defined location within a portal and provide rapid access to other well-defined places. Some of the most useful patterns (in addition to navigation sets) are listed below [Rossi et al.]. These patterns can be used independently but are often found together.
Landmarks: Landmarks are links to entry points, large subsections of portals, or frequently used applications. CNN.com, for example, displays links to weather, politics, business, sports, and so on in the left side menu.
Nodes in context: A variation on navigation sets are nodes in context. With this navigation pattern, the same content is repurposed for multiple uses. Depending on the use, the links associated with content vary. For example, an online retailer could list a new tent in both the camping equipment area and the new products area. In the former case, links from the tent display would lead to other camping equipment while the latter category would include links to other new products. Again, the point is to provide navigation links to logically related content to create a sense of context and intuitive navigation paths out of one navigation set and into another. Both navigation sets and nodes in context provide a fine-grained sense of context. Users also need a sense of context relative to the portal or site as a whole. That is the job of active references.
Active references: Active references are indicators of one's position relative to the site as a whole or relative to a landmark. Directories such as Yahoo! and Google Directory use a list of nodes traversed in the directory. For example, see http://directory.google.com/Top/Computers/Internet/Web_Design_and_Development/, which uses an active link: Computers > Internet > Web Design and Development.
Visual active references are excellent methods for depicting location within a larger context. This is especially applicable when content corresponds to a physical location, such as a room in a building or a street address. As the New York Metropolitan Museum of Art timeline of art history shows (http://www.metmuseum.org/toah/splash.htm), two or more contexts, in this case time and location, can be depicted simultaneously.
News: Another common navigation pattern is the news section, which is used to prominently display important corporate information or new portal features. Many public news feeds are also available in XML formats, particularly the Rich Site Summary (RSS) scheme.
Together, these and other navigation patterns constitute an essential part of the overall information architecture of a portal. They provide a sense of context to the user and offer easily accessible links to significant or related sections of the portal. To ensure that the links are named logically and the user's experience is consistent across the portal, we must define labeling standards to identify the links that constitute these navigation patterns.
Labeling: Pointing Users in the Right Direction
Labeling standards dictate how content, links, and other objects are named within a portal. At first blush, this may seem like a trivial consideration compared to others you have to deal with in a portal implementation. However, users constantly see and use the labeling system in a portal. Well-designed systems aid navigation and should be almost unnoticed by end users. Typically, when end users notice the labeling scheme it is because of a problem. The scheme may be inconsistent or ambiguous, or it may have some other aspect that puts an additional burden on the user to determine an appropriate action or understand the meaning of a link. The issues you must contend with include the following:
Multiple terms that mean the same thing
Terms with multiple meanings
Not long after you begin work on labeling you realize there are many ways to describe an object. Choosing one term often pleases some users and leaves others disagreeing. Some objects, such as products, departments, and projects, have official names and so have an obvious label. Even in these cases, product names change over time and departments are reorganized; as a result, outdated terms can be found in older content. Nonetheless, standard terms should be used consistently throughout the portal. In many industries, controlled vocabularies or standard lists of terms have been developed by corporate librarians, information scientists, and others who have had to deal with information retrieval problems long before the advent of the Web. These industry standards can provide the basis for a labeling standard and minimize the time and effort required to develop your own.
When controlled vocabularies are not available, search log analysis can provide a starting point for labeling conventions. Search logs identify the terms used to query an intranet or portal and provide information about results as well. The frequency with which terms appear in the log can guide the selection of terms for the labeling standard.
Successful labeling schemes are built on two factors. First, the choice of labels should be based on either their use among the portal audience (as measured by search log analysis) or the terms used by other external sources (e.g., industry vocabularies). Second, the labels should be applied consistently through the portal. This also entails maintenance because labeling schemes change to reflect changes in the organization and general business environment.