Saw this great tag cloud cartoon on ReadWriteWeb (RWW) today and it prompted this blog post. I've got a mini-rant on this topic due to my constant advocacy for tag clouds and resistance from everyone I recommend them to.
Why? OK, it's because it requires a custom build, I admit that. But I truly believe the reason tag clouds have hung in the margins (literally) is because everyone wants an "off-the-shelf" solution to tag clouds. I just haven't found one already built that both works for SEO and makes sense to users.I see a huge SEO value in them if implemented in a completely different way than everyone seems to be doing it.
That cartoon above seems to be suggesting tag clouds have died. I really don't care if they have or not - because they've never been done right for SEO purposes. I argue for tag clouds on a relevance model - not popularity models as they are most frequently used.
Let me back up a little for the uninitiated, and we'll go through the basic concepts here. The intent as designed in the popularity model seems to have been to display the most frequently used keywords that people apply as tags on a given site or blog or photo-sharing site. The more popular or frequent a word or phrase is on that site or for a particular photograph or article, the larger the font those keyword phrases are displayed in within the cloud. Larger words mean they are more popular and are more often used as tags on the site.
While the above is a fun graphical representation, it's not what "real" tag clouds look like, so to be thorough, here is an example from Wikipedia page on tag clouds.
Those individual tags are then linked to a page bearing a list of articles or photos tagged with the word or phrase you click on. While this blog is not using tag "clouds", it is using tags. If you click on the "yahoo" tag, you'll end up on a page which includes all posts I've tagged with "Yahoo" from among my blog posts. (Go on Try It!).
Most custom build tag cloud solutions fail to use the tag microformat which would be favored by search engines and instead - they simply link the tags to an internal search function which does a site-level search for the word linked in the cloud. This is horrible for search relevance, user experience (due to poor results), and for SEO.
I want tag clouds that group only topically relevant words. This requires a different mind-set than most apply to tagging. I want all RELATED words and phrases in a cloud, then I want each word or phrase hyperlinked the way the tags used on this blog are linked - by going to a page which includes all articles tagged with that particular phrase.
The pages grouping tagged articles listed could take excerpts of the articles to avoid duplicate content penalties. The headlines from articles could be linked to the full page versions.
If anyone knows of a script or utility which could generate effective, relevancy model tag clouds in a plug-n-play way that I can hand to clients - I'd be happy to become a reseller - since I'm not a developer. Sometimes I wish I had the skills to do SEO engineering, but oh well - I'm an idea guy who does SEO strategy.