By bringing together search, tagging and related links in a relatively automated process I really believe that active navigation provides a viable way of dealing with massive websites.
Just occasionally you come across a website that just doesn’t fit the normal pattern of things. Headscape was recently approach by a potential client who had literally hundreds of thousands of web pages which were almost impossible to organise into a traditional information architecture. They wanted us to suggest some alternatives and so I thought I would share with you my response.
There are lots of reasons why a normal information architecture might not work for your site. However, probably the most common is that your site has simply outgrown the constraints of a hierarchical tree and that your users are getting lost deep in the information architecture. Often the answer is simply to do some radical pruning to remove much of the deadwood, however occasionally a different approach needs to be found.
So what are the alternatives available to you?
The most obvious approach is to use search as your primary navigational method. Indeed search is the primary mechanism we use to navigate the entire web and so we know that it has the scalability required. However, although potentially the best solution it is often poorly implemented. I have written about how best to implement search before and so I am not going to repeat myself here. Nevertheless, there are additional things to take into account when search is used as the primary navigation for an extremely large site.
Firstly, position is important. A search box is a relatively small screen element and so can easily be lost within a design. When search is the main way that a user will find their way around your site you need to make this obvious. Move the search into a more central screen position like Google or Amazon does.
Secondly, on a large site advanced search is important. The problem is that users struggle to build complex search queries. They find concepts such as using quotation marks or Boolean joins difficult to grasp. As web designers we need to search for new ways to help the user build these complex searches. As it happens the answer might be right under our noses.
Dreamweaver comes with an interesting tool that helps users build complex search and replace commands without having to know about regular expressions. Although a similar approach has been used on websites in the past, the arrival of AJAX and DOM Scripting now means this kind of functionality can be implemented in a much more intuitive and responsive way.
Although not a replacement for traditional navigation, related links can work as an effective accompaniment to search. Related links allow you to establish a loose relationship between documents which is much more flexible than the static hierarchy of an information architecture. They can provide a context to a document which search alone does not offer. However related links are not without their problems. When publishing new pages it is a relatively simple process to link those pages back to previously published pages in order to provide context and additional information. However older documents are often neglected and when a new page is published these older pages aren’t updated with a link to the new page. In other words old pages atrophy. The problem is caused because the system is reliant on editors remembering that there are older pages which need updating. Ideally some automated system should identify related pages, but we will come on to that later.
Normally breadcrumbs are associated with a traditional information architecture where they show your current location in the hierachy. However, that is not the only use of breadcrumbs. A less used approach is that breadcrumbs show the path the user has taken through the site. The reason this approach is less used is because in theory the browser back button provides the same functionality. However, the majority of users are unaware that the back button allows them to skip back multiple stages. Historical breadcrumbs clearly show the user which pages they have previously visited and allows them to quickly jump back to anyone of those pages.
There is also an opportunity here to once again learn from desktop applications. Windows Vista has added an interesting new feature within its file explorer that might be adapted for our historical breadcrumbs. Basically the file explorer’s breadcrumbs allow you to view the children of any folder in the hierarchy:
This approach could be used with our breadcrumbs to show any links within a document allowing the user to quickly jump to related pages. The primary reason a user utilises breadcrumbs or the back button is to select another link so this approach could prove invaluable.
Next to search tagging is the most common way of organising non hierarchical information. Indeed this post has been tagged as shown at the bottom of the page and you can view a tag cloud for the entire site. Sites like delicious also allow for the bundling of groups of tags to create a basic information architecture for the site.
However, tagging does have some fundamental issues which need to be resolved. The greatest of these is who does the tagging? The most popular answer to this question is that the user does it. Indeed Russ Weakly of the Australian Museum put together a very compelling presentation at Webstock for a user driven information architecture based on tagging. There are also a growing number of tools that facilitate user tagging of a website. Of course the problem with this approach is that it requires the user to be motivated enough to do the job. When tagging my bookmarks on delicious I am motivated to do so in order to allow me quick access to them in the future. However, the same motivation does not exist to tag a page on a standard website like this one.
The alternative to user tagging is that the owner of an individual page tags it. The owner is certainly much more motivated to tag the document but that doesn’t mean they are the best qualified to do so. The problem is that an owner’s mental model can be radically different to that of a user and so their choice of tag can be inappropriate to those navigating the site. Web page owners tend to rely heavily on jargon and view the world from an institutional point of view. Of course the other problem of editorial based tagging is this can present a massive problem if there is a substantial backlog of web pages that need tagging.
Much of the problems that arise with both tagging and related links comes from the need for human intervention and in particular the degree of intervention needed to tag or link older documents. One solution might be to automate some of the process.
Back in 2000 I was fortunate enough to work with a ground breaking company called Active Navigation. These guys were working with similar algorithms to those found in a search engine like Google but applying them in a very different way. Like Google they were using linguistic routines to analysis a page and identify the keywords which best described the content. However instead of using the results to build a simple search engine they were also using it to build a navigational structure of related links and tags. This would ensure related links were always up to date and avoid much of the human intervention required for tagging. Instead of having to tag and manage individual documents they can simply edit and manage the tag cloud to ensure only the most relevant keywords are featured.
Because the system fundamentally uses the same approach as a search engine algorithm it is more than capable of also providing search functionality. By bringing together search, tagging and related links in a relatively automated process I really believe that active navigation provides a viable way of dealing with massive websites.