Brand Tags: the really effective tag cloud

In my book I write about some of the trade-offs involved in making tag clouds. Whenever you have a lot of tags, tag frequencies usually follow a power-law distribution. A few tags are used lots of times; most tags are used just a few tags. To make a tag cloud aesthetically interesting, not to mention useful, you have to compress that distribution--you don't want your largest tag 500 times bigger than the smallest one.

This is the example I use in the book (from Flickr):

The compromise is that you lose information; the tag cloud doesn't reflect the real distribution of tags. Instead, it just gives you a sense of what's popular--and most of the time that's good enough.

But Brand Tags, a new-ish site where you can play a tagging word-association game with brands, flouts this aesthetics-over-information convention to good effect. Darren Barefoot calls the site ad hoc market research, and I think that's a pretty accurate description.

Take a look at the full page of tags for Starbucks:

Because the tag cloud isn't downscaled, you can see that "expensive" and "overpriced" are the overwhelmingly top-of-mind tags. (I'm not saying this accurately reflects public opinion of Starbucks. Brand Tags users may be cynical under-caffeinated spendthrifts. But Brand Tags really works as way to capture and display those top-of-mind brand associations.)

The important thing here is that this tag cloud doesn't have to be used as navigation. It's strictly informational, and showing real distributions (or so I'm guessing) increases its impact.

Just for fun, let's compare Starbucks to Apple:

Who wouldn't want that as brand tag cloud?