A tagcloud is a common element on web2.0 sites. This alternative navigational element not only gives users a view into the site contents, it offers a visual representation of the most commonly used terms to describe said contents. But just like other neat-o web2.0 effects, whenever I hear “we need {xyz feature}”, I find myself needing to ask why. Effects are fine, but is there a good rationale for inclusion?
First, I suppose I should define a few things. Tags are terms assigned to items. Generally, tags as assigned by users, as a way to engage with the site. It supports the development of a folksonomy, wherein individuals classify and categorize data, rather than referring to a formal taxonomy. I would argue that this makes more sense when the extent of the data is not known up front.
A tagcloud is a way to present this data to the user. Tags do not need to appear in a cloud-like fashion, but a tagcloud offers a viewer additional information as to the relative number of elements with that tag.
Navigation
Tags give site visitors a view into the information on a site. Rather than having to progress through the site in a linear fashion, they can access site information in relation to specific categories. As I mentioned before, this is particularly useful when the site contents are not known up front. An example is this blog: I did not know the different categories in which my posts would fall with the site was designed. Tagging gives visitors a quick way to access similar articles.
Search
Tags are sometimes considered an alternative (or supplement) to search, filtering the content the user accesses. Whereas a free form search may not return any results, a tagcloud guides the user by letting them know what is on the site. Again, this is an alternative type of navigation. The relative size of the elements in a cloud also aid the user in letting them know what filters are densely-populated.
User goals
In some cases, tagging may be useful without any such cloud; depending on the users’ goals. I can tag messages in gmail to facilitate retrieval, but their relative frequency of use is not important. If I need to find an email pertaining to an uncoming trip to Boston, I don’t care how many tags are tagged “Chicago”.
User Engagement
Tagging content is indicative of a certain level of user engagement. Tagging allows for content to be organized organically. On Shelfari I can add my own tags to every book in my collection, to help me develop my own view into the data. Initially as I tagged my items, I was thinking I could use my own tags to determine how the books could help me set up my bookshelf. Books could be displayed based on the categories I assigned.
There were a few reasons this wouldn’t work. For one thing, unlike that crazy Dewey Decimal system, a single book may have several tags associated with it. If I really want to continue with the metaphor of a bookshelf, a book should only appear in one place. This was the very departure that prompted gmail to adopt tagging: a document need not “live” in a single folder, rather, it could be associated with multiple tags (I believe their original analogy was that a document could have several sticky notes assigned to it).
Even if I only assigned a single tag to each book, how would they be displayed? By default, tag clouds list their tags alphabetically. This has some merit, but if I truly was looking to organize my bookshelf into discrete buckets, perhaps tagging was not the answer..
User Engagement, Take Two…
In drafting this post, I was initially thinking that tagging was exclusively a user behaviour. If a system is categorizing their own data, traditional navigational structure and information architecture techniques should be used. But then I thought of this blog: I am willing to assign tags to my own posts, and I do not allow visitors to do so. How is that different than a brand site assigning their own tags? I think the difference is that this is a blog, so it is still the creation of an individual. A blog develops organically, and there was really no way for me to foresee the tags I would be adding as it developed. In contrast, a site such as CNN would fall over if all the content were organized via tags: there is still a need for a traditional navigation and categorization structure. It makes sense for tags to describe articles by a given journalist, but the site still needs to have some sort of imposed navigational structure.
Results
Another key characteristic is that a single tag refers to a collection of items, but can only have a single destination.
In gmail, all items are emails, so just like search, clicking on a tag can bring me to a results set of emails. When we start looking at different types of items, returning results can become more difficult.
Only searching on a single “tags” field may simplify search rues, but it doesn’t help in displaying results. How you do return hits from different types of items? There is a reason why google continues to separate search results of text, news and images. For each type to be returned, a standard display template must be defined. For del.icio.us, which allows users to tag online documents, this template involves the link itself, a description and all the tags. When viewing results, a visitor may also see other tags that users have used to describe the document, as well as the number of other users that have tagged it. Once a template is defined for each discrete item type, a template for how the different types will be displayed together must be created. Ask.com has been working on returning results of different types.
Whose tags are these, anyway?
Another consideration around tagging is related to the level of surfacing these tags. Do my categories only relate to my own view of the data, that is, I can see a collection of “my tags”, or are they an attribute of the system? With del.icio.us, I may have a “to read” category, but viewing all links everyone has categorized in that way may not be useful. In gmail, my tags are tied to my specific user information, and they are completely private. In Shelfari, I may see my own tags, or I may access the tags that others have associated with items. This returns back to the notion of user goals. In gmail, I am trying to facilitate retrieval of known items. On amazon, I may use tagging for discovery (finding similar items). These differing goals will have an impact on how tagging is implemented.
So before I’m told that a tagcloud is a requirement, I’m always careful to ask what purpose it serves, what user goals it satisfies, and to what the tags will be applied. A tagcloud really is a means to an end, not an element to include “just because”. If you simply want a site to “look” 2.0, get a reflective logo or something 
Related posts:
- Subscribe to the RSS feed
- Get Email updates
- Tweet This
- Stumble It
- Add to del.icio.us
- Share on Facebook
- Email to a friend
