I’m a huge fan of tagging as a means to organize data. It’s powerful and flexible– and it oftentimes has some pretty exciting social ramifications.

If you aren’t familiar with tagging (and you want to be), you could get up to speed fairly quickly by checking on the wikipedia entry on Folksonomy. If you’re more interested in insight rather than information, you should check out what Josh Porter has to say on the subject (Josh is hands-down one of the most insightful bloggers out there IMHO).

As a guy who built a web 2.0 resume posting doohickey (chock full of taggy goodness), I’ve put a ton of thought into tagging, specifically in the context of UI. So it was with great interest that I attended the SXSW panel entitled, “Tag, You’re It!”. The panelists consisted of a lot of impressive folks– George Oates from Flickr, Heath Row from DoubleClick, Ben Brown from Consumating.com, and Thomas Vander Wal (the guy who evidently coined the term “Folksonomy”).

The panel was interesting but like a lot of SXSW panels, the more you knew about the topic, the less interesting it was… But, I digress.

The most interesting moment (for me) was during the (very short) Q&A session. A person asked the question, “How do you deal with synonymous tags?” It was obvious that this was not an uncommon question– George Oates had a canned answer for that question…. “You don’t,” she said (yes, George is a girl). “It’s perfectly okay and wonderful that 3 people might tag a single data object in three different– but really similar– ways.” There it was, case closed.

The panel was wrapping up, but I wanted to shout, “Hey WAIT A MINUTE. That’s GOT to be wrong!”.

As I reflect on it, it turns out that there are (at least) two types of tagging– one of which is clearly a winner. The other (which is the type we applied at Jobby and currently are dabbling with at Jobster) is doomed to failure unless we get clever about how we pull it off.

Tagging where the Selfish Motivation is Organization

Flickr and Del.icio.us are the tagging poster-children. They are wonderfully simple– they provide a storage repository for big chunks of personal data (photos for Flickr and bookmarks for del.icio.us) and give you a powerful means to organize them. People oftentimes tag in radically different ways. Some people have dozens or hundreds of tags. Others have only a few. As George pointed out, people oftentimes tagged things with very similar tags. One person might tag a resource with “rockstar”, while another might tag it with “rock_star”, and a third might tag it with “rock-star”. This is fine with Flickr and Del.icio.us… With the service they offer, it’s most important to allow users to label their data in the way that makes sense to them.

The core functionality is organization, and the ability to search/browse/find similarly tagged objects is serendipitous. As a Flickr user or del.ico.us user, you really have no huge incentive to have your data be found by anyone else.

Tagging where the Selfish Motivation is Improved Findability

The only panelist whose userbase was largely concerned with findability was Ben Brown, of Consummating.org. Essentially, his site (recently sold to CNET) is a site where geeks come and tag themselves so they can get matched up with other geeks so they can fall in love and make lots of baby geeks, presumably. This is not unlike the tagging model that we used at Jobby (and currently use at Jobster). People labeling themselves to get found by other people.

This is where the tagging concept starts to break down a little bit. All of a sudden, it’s no longer important what tags you’d use to describe yourself– it’s a hell of a lot more important what tags people would use in a search to find someone like yourself. There are a few unfortunate byproducts of a system like this:

• You are incentivized to lie. The more desperate you are for a date/job interview, the more likely you are to lie. Why not tag yourself “slender”, even if you’re not? Might get your profile a few more views and one of the viewers might latch onto one of your other qualities. Why shouldn’t I tag myself “Ruby on Rails”, even if my experience with it is woefully limited? Maybe some manager out there will find my resume as a result of that tag and just maybe he’ll be wowed by something else he sees.
• You are incentivized to pour on the tags. The cost of adding a tag is very low, and every tag will increase the likelihood that my resume/profile will be viewed. This creates a lot of tag noise. Instead of having a nice and concise list/cloud of scannable tags, you can start seeing tag collections that are less than manageable. A smart coder will tag himself as “coder”, “programmer”, “developer”, “engineer”, “software developer”, “software engineer”, etc.
• Synonyms matter. A lot. Ben used an example of his perfect woman for a search… He did a search for a woman who was tagged “coffee” and “cigarettes” (ew). The problem with tagging in this model is that Ben’s search simply won’t find a woman who is tagged with “caffeine lover” or “smoker” or “chainsmoker” or “coffeelover”.
• People who are tagging data need to be smart about the psychology of searching, which they never will be. In the SEO biz, there are lots of tools (some of them very expensive) to allow a person to understand what people are searching for (to allow you to craft your content and meta-data to map to those search behaviors). People looking for a date or a job don’t have the time or money to figure out whether hiring managers are searching for “programmer” or “developer” more often. If the tagger is sophisticated enough to even consider this problem, he’ll probably (to be on the safe side) tag himself with both.

So, are you screwed if your service has tagging to enhance the findability of your users? I hope not. Here are a few strategies we’ve used at Jobby and Jobster to keep tags from getting “spammy”:

• Suggested tags. This is the best thing we did with Jobby and (unfortunately) it’s a feature that is really buried in the UI on Jobster. With Jobby, we essentially took what data we had on the user and presented tags that seemed likely to be of value to the user. So if a user tagged themselves “HTML”, we could fairly reliably say (given how often the two tags went hand in hand for other users) that they might want to tag themselves “CSS”. This also has a nice side-effect of giving your users the ability to tag themselves with a click rather than by typing all of their tags. And, of course, it encourages people to use the most popular iteration of a tag (using “CSS” instead of “cascading-style-sheets”, for example).
• Add scarcity to your model. At Jobby, we allowed user to assign a skill level to their tags (beginner, intermediate, advanced). There was functionally no limit to the number of tags, but it makes it a touch more difficult to lie. A lot of fledgling web geeks might be comfortable saying they know JavaScript, but a lot LESS comfortable saying they are JavaScript experts. At Jobster, I think we’ve provided a more elegant solution. Users are allowed to have up to 5 Superstar Tags– basically, these are the skills/areas that you feel you’re best at.
• Add REALLY SMART suggested tags within your search UI. We did this (in kind of a clunky way) with Jobby, and people LOVED it. With Jobster, we’re JUST starting to experiment with this, but I think it’s critical for tagging to work (when findability is the primary motivation for tagging). You’re search experience needs to be smart enough to present similar and/or related tags. Further, it needs to be smart enough to give the user a sense of the popularity of given tags.

I’m still waiting to see a site that really manages to nail the tagging/searching experience when the motivation isn’t just for personal organization. I’d love to hear more ideas on how this could be pulled off.