Lou Rosenfeld was recently in London presenting his one-day seminar on enterprise information architecture. I have been doing a lot of work with intranets lately, so I thought it would be useful to attend.
No one dealing with Web site development in a large organization would have any trouble identifying with some of the issues Lou raised:
- Competing demands from many quarters
- Tension between centralized and distributed Web solutions
- Information organized by business function, leading to content "silos"
- User confusion
Against this raft of problems, enterprise information architecture offers several sets of strategies:
- Enterprise search
Not surprisingly, top-down strategies deal mostly with the taxonomy of a site, transforming it from organization-oriented to user-oriented. This won't happen overnight, so a number of different short-term strategies are suggested:
- Superficial changes to the taxonomy to make it more topic- or product-oriented
- Transforming site maps from "org chart" to topical
- Specialized site index
- Guides and topic pages
Mostly these short-term strategies are attempts to make a site appear to address real users' needs while trying to disguise the underlying Medusa-like structure. For many sites, especially those with products or services to sell, these will be good short-term solutions. At the same time, some usability issues should be aired. Site maps and indices will possibly give more political mileage within the organization than any real improvement in usability. Jakob Nielsen  discussed site maps in an Alertbox in January 2002. At that time, only about a quarter of the users in his study tried to use a site map when stuck (and all of the sites selected for testing had one). When the Web was viewed as a whole, only about half of sites had a site map and when asked to name a site that did, users were correct just under half the time.
I conducted a more recent study of my own to see what changes have taken place in the intervening two years. Of six randomly selected sites (Dell, Cisco, Ford, General Motors, Hewlett-Packard, and IBM), only half had site mapsnot much change there, then. So while it may be a good idea to persuade a department to move its yawn-inducing material from the home page to a prominent place in the site map, chances are that users will never see itwhich may be fine as long as the department concerned never finds out.
Specialized site indices and guides may fare better at actually improving site usability. In some cases, the limitations of the primary navigation may mean that a specialized index becomes a valuable user resource. That thought crosses my mind every time I try to find a product on Adobe's site. Although excellent in many respects, I find the choice of product categories bafflingIllustrator is a Web Publishing product, whereas Photoshop is a Digital Imaging product. And Adobe's new audio package, Audition, is listed under Digital Video! An alphabetical product index is what I really want, but unfortunately the page that looks the most promisingthe All link under Productsuses all the infuriating categories I just mentioned.
Specialized indices, guides, and topic pages allow better "cross-silo" access. Information that may have been hidden away deep within a departmental site is put into its proper context and made easily available. Topic pages allow links to be presented according to users' goals, and guides are slightly more elaborate. They have the added advantage of allowing a judicious amount of explanatory text to be added when the terminology of an index or topic page may be inadequate.
The bottom-up approaches deal less with presenting a site's current pages under a new light than with structure. The main activities are:
- Content modeling
- Metadata development
Content modeling involves finding useful relationships between different types of data that might currently be hidden in different silos. The example Lou used was a music site with a variety of CD-related information within the e-commerce silo, linking to group and event data currently held by different departments in other silos. Although this strategy may or may not improve the actual usability of the site, it does give much better scope for providing services that are truly useful compared with the static alternatives offered by printed media. The development of metadata-fields describing content in ways that make such links and related searches possible plays an important role in the bottom-up approaches. So too do the issues of semantic consistency and relationships. Semantic consistency is provided through controlled vocabularies, meaning that concepts are described in only one way within a site's content. Semantic relationships allow the controlled vocabulary to be mapped onto a variety of user-oriented terms. For example, the controlled vocabulary term "personal digital assistant" might be mapped by users entering the search terms "handheld," "PDA," "PocketPC," and so on.
"Search" is an important topic in its own right. Unlike site maps, "search" is present on the vast majority of sites and is the first tool out of the box for many users. However, a number of issues conspire to make "search" less useful than it might be:
- Too few results
- Too many results
- Irrelevant results
Some of these problems will be addressed by good metadata and controlled vocabularies, but sometimes more is needed:
- Helpful feedback on combinations of search terms
- Control over areas searched or context of terms
- Spell checking
The impact of this last issue cannot be underestimated, and search engines such as Google are setting the standard in this area. When users come to your site, they will expect to have spelling errors either corrected or drawn to their attention. More important still in many cases is the issue of spell checking on internal employee directories. Many, many proper names are simply not spelled the way they sound. Forcing users to make multiple guesses or to revert to the printed alphabetical is just not a good use of technology. This is true especially when it means standing in reception trying to guess the correct spelling of the person you are trying to visit, as happened to me recently.
I have not covered everything in the full day's seminar, but I hope I have given the flavor of Lou's approach to enterprise information architecture. It seems to me to be very pragmatic in dealing with real problems found in many large organizations. I would not suggest it as a replacement for user-centered design and usability techniques, but as a team they should work well together.
William Hudson >Principal Consultant >Syntagm Ltd >email@example.com
©2003 ACM 1072-5220/03/1100 $5.00
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2003 ACM, Inc.