The Art of Skywriting – The Demise of the Tag Cloud


Demise of Tag Clouds

This year I was honoured to be given the opportunity to speak at OZ-IA 2009.  I will also be speaking at the Edge of the Web 2009 in November for the first time, right after the keynote, talk about pressure!

The following is my initial transcript of my presentation at OZ-IA 2009, The Art of Skywriting  – The Demise of the Tag Cloud.

Now it’s not going to be word for word the same as the audio recording, but you’ll get the idea.  As usual it’s also available on Slideshare along with all my other presentations, distributed under creative commons license.  Soon as I have the audio I will sync it with the slides to make it easier to understand – promise.

Tag clouds have been a part of our community for a while now, we have come to love them, we have embraced them, welcomed them into our households. They have become part of our teams, an ever-present design or navigational component.

Or so we thought.  I don’t know if you have noticed but over the last few years Tag Clouds have been disappearing from web sites at an alarming rate.  In the dead of night they have been ripped from their home page, tossed aside, unwanted to die a lonely death on the curb side.  It’s as if no-one really cares.   I know I care!  What has been happening to our Tag Clouds!

I feel it’s time we got some real answers, answers from some fellow professionals, so I have called in a few favours the managed to get the SUXU team “Special UX Unit” on the case.

Now at last we can surely find out if some heinous crime has been committed and finally learn the true fate of the Tag Clouds.

The Investigation


To understand what has happen to Tag Clouds we really have to understand their history.

Tag Clouds first came onto the scene in a big way in 2006 when Stewart Butterfield (the designer at Flickr at the time) leveraged of the work of Jim Flanagan’s Search Referral Zeitgeist.   So for better or worst Flickr saw an opportunity to try something new and find a use for all those community tags.  Mind you Tag Clouds in theory have been around since 2002 in an online implementation and as theory way back to 1995.   It’s just in 2006 Flickr pushed them out into the limeslight.

Over the following year Technorati and Delicious would both implement Tag Clouds onto their home pages, shooting the status and popularity of Tag Clouds into that of a mega-rockstar.   Everyone wanted to have a Tag Cloud on their website, from blogs, to tech media sites.   It was even suggested that Tag clouds could replace the trusted menu navigation system.

By 2007, even the conservative end on town, in the form of the retail sector was looking at Tag Clouds. As the fringe of the retail sector began to implement Tag Clouds to allow customers an easier method to connect with their popular product lines.

Still what are Tag Clouds really like?   Just like we profile the suspects, so too the crack team at SUXU have profiled the major 3 types of  Tag Cloud

  1. Mr Popular (the Elmo)

    With this tag cloud the word size represents the number of times a tag is applied to the single item.   These Tag Clouds are usually generated of audience contributions, like in the case of

  2. The Streetwalker

    For this Tag Cloud the word size represents the number of items that to which the tag as been applied (or used).  As in the case of Deep Search, the large keywords have more search items applied to them.   There is no real freewill here, it’s just a case of take what you find.

  3. Mr All you Can Eat

    This final Tag Cloud is used for representation of content categorisation. The bigger the words, indicates the more items in that category.   This type of Tag Cloud is very popular in blogs.   It is rarely generated from data that is provided by a global community.

Was there a Crime Committed

Just a reminder that tag clouds in theory allow for a number of key visualisation techniques:

  • As we all know Tag Clouds allow for a visualisation of the presented word frequencies, we can see in this case that Swine-flu and flu are prominent when Japan is not.
  • They allow for a community or audience to contribute the tags and determine the direction of the Tag Cloud
  • Tag Clouds can present an alternative navigation aid, allowing for easy discovery of the required or new content.
  • And they can also present the visualisation of a semantic categorisaton, where the related items of information can be presented in close spatial approximation in the Tag Cloud. In this case Science, Politics and Religion are related, and so is Science Fiction, Sci-Fi, Sex and Zombies.   It gets stranger with Star Wars, Star Trek and Toilet – what are people saying here.

By the end of 2007, early 2008 Tag Clouds had really made it, even the Microformats community was now looking for a model to implement tag clouds. They where considering that there were enough instances “in the wild” as they like to say, to warrant a Tag Cloud Microformat.

Also the IA community stepped forward and deemed that Tag Clouds were now a useful technique for any designer’s toolbox and so they went and produced a design pattern for Tag Clouds.

However we did note and under-current, that like the Tech fringe of computer geeks, Tag Clouds seem too only appealing to a fringe community.   Like with Iggy Pop, the general public just didn’t embrace Tag Clouds.

Also like Rick Sanford, lead vocals, percussion and… flute, time had passed maybe, no longer was the white sexy neck tie, bare chest and mullet in vogue. Sorry Rick and prog rock, time to move aside.

It’s at this point in our investigations that we found unsolved cold case after cold case, hundreds of instances of the innocent Tag Clouds that have been brutally slain.   Clearly justice needs to be served.  We now had our suspects.

The Case for the Prosecution

Geeks and Geeklettes of the jury, we all know about Tag Clouds and their history now.  It’s is clear as has been presented by the members of the SUXU that their has been a heinous crime of murder committed here on a monumental level.

Suspect One

Suspect One - The Community

But did you know that by 2006 when the tag cloud population was at its peak that only 7% of the Internet community was in fact tagging [1].   Hence only 7% of the community was providing the tags for Tag Clouds.  This later increase to 12% as the Tag Cloud population waned.  Still with the beloved blogs being the home of the Tag Cloud. The number of bloggers on the web is only about 10-15%.   This means only 10-15% of the web community is supporting Tag Clouds.   Surely these people, the controlling fringe elements of the web community are a suspect.   Hence I give you suspect number one.

Suspect Two

We all know as professionals in the UX field that words do not always have the same meaning from society to society from one cultural group to another. For example I’m just looking for a wholesome “Family Guy” and I get this.   Err No. That’s not what I wanted.  Tag Clouds don’t help this matter.  But it’s not their fault.   Words and their differences are to blame.   So we have suspect two – our differences in meaning.

Suspect Three

The death of so may Tag Clouds can be lain at the feet of users.   They just don’t get them at all.   According to research people often just didn’t get that those items where links they could click on, or that they in fact lead to more information, or that they presented a visualisation of the information of the site.  Tag Clouds didn’t come in most cases with a manual.  But frankly realistically who would have read it if they had.  So the pesky users are suspect number three.

Suspect Four

You know what, people didn’t even realise that the big items in a Tag Cloud where big for a reason.   They considered that is was like that because the designers just wanted it that way.  If it was red or green in colour it was just a design element, and that was all.    Hence maybe the designers are a little to blame here in their presentation of Tag Clouds. So I give you the designers are suspect number four.

Suspect Five

People with accessibility issues are a pain in the arse.   They really made life hard for Tag Clouds.  Constantly complaining about, font resizing issues, colour-blindness, motor coordination issues on small sized words, lack of weighting and semantic spatial relationships for visually impaired users.   So much so that really we do have to consider these members of our community as suspect number five.

Suspect Six

The Usual Suspects - the Search Engines

Finally search engines are our last suspects.  The search engines and their bully gang have been lurking in the back alleys of the web. We know they have been secretly a little jealous of the Tag Clouds; especially the in-house search engine.   I put it to you that the in-house search engines with their brothers have been a contributing fact to the demise of the Tag Cloud.

However it does stop their ladies and gentlemen of the jury.  The 6 accused have marginalised Tag Clouds. To the point that they have become little more than noise or window dressing on a page.   And in the harsh commercial reality of the Internet anything that is noise on a commercial site of any nature is bound to be “remove”.  It just makes sense.

Designers started to list Tag Clouds in a warped mangled alphabetical order, but this was just another attempt to destroy them, as we will see.

Tag Clouds and User Research

Research Findings

Research has indicated that with Tag Clouds:

  • Large tags attract more attention from users
  • People don’t read a tag cloud, they scan for the word
  • The centre based tags of a cloud will attract attention
  • The upper left Quadrant of a tag cloud get more attention on average
  • And that Tag Clouds are just not suited to supporting the search for information

I put it to you that these 6 suspects are guilty of the crime of murder of the Tag Clouds by conspiring to present Tag Clouds in their worst light.

The Case for the Defence

There has been no murder here at all.  Let’s be fair, these suspects are guilty only of marginalising the tag clouds, not of wholesale murder.

The Tag Clouds are just alive and well, hidden in plain sight.

Tag Clouds Part of Our Community

Sure the Tag Clouds have disappeared from the home pages of Technorati and Delicious. Yes they are not easy to find on these sites, yes you have to login and be a member of the community to find them now.   But they are still around. Protected from the thugs of society.

You can still find Tag Clouds, alive and well in the lower reaches on articles on media sites.  Have a look; below the last adverts in the side bar, they will be living out their lives there.   Admittedly it’s not the best of places, but they are alive.

Tag Clouds are still in the mainstream; Governments are using them, as shown here. Also they are being used by MIT (not the Melbourne mob) and by the Creative Commons people.   Now I wouldn’t be calling these sites fringe, would you?   I would say more the conservative end of the community spectrum.

Anyone can now make or implement a Tag Cloud, there are now Tag Cloud Generators such as Wordle and also  a formula for the weighting on a Tag Cloud.   If anything we are promoting and cementing maturing Tag Clouds into our community.   So now Tag Clouds are here for everyone.

The Maturing of The Clouds

We can see that Tag Clouds have grown and matured, which the prosecution has failed to acknowledge in their haste to find a suspect.

Some Tag Clouds have become Index Clouds, where the information is presented in a alphabetic index like fashion.

Others have become Button Clouds, overcoming the issue of the user not knowing if it is link, by making it like a button with a little weighing for good measure.

Search Engines aren’t the enemy of the Tag Cloud – they are the Tag Cloud’s friends. They are not some back street gang of thugs to be avoided.

Many Eyes and Deeper Web have both been leveraging the power on the Search Engines to provide them with the best of both worlds.   With a situation that the results of the Search allow for an extended result set by using the semantic comparative nature of the tag cloud.

Text Clouds

This can be extend further with the use of Text Clouds, here the we feed the output of the search engine (the shark) to the goldfish (the tag cloud)  such that:

  • The popular keyword search terms become the items in the tag cloud
  • It still has all the display and usability issues of a traditional Tag Cloud.
  • However the information base is a lot larger than a Tag Cloud, allowing for a better distribution over the terms.
  • One issue with Text Clouds is that they are open to trending as the keyword search terms of a search engine are.

Topical Grouping

Luckily for us Tag Clouds have changed for the better via the use of Topical Grouping:

With Topical Grouping you take a topic of interest:

  • Then you display in a Tag Cloud for that topic all the related topics and sub topics.
  • This presents a drill down and sideways effect on topic examination.
  • Topical grouping does allow for an increase in findability for a moderately small population of tags
  • As you would expect using Topical Grouping with a Tag Cloud increases the rate of a serendipitous find.

Was there a murder? No good people there was not. It is just Tag Clouds have matured and the prosecution failed to see this. .

The Verdict

You have heard the case for the Prosecution and the Defence.   In this case we have 6 suspects

  • The Community
  • The Designers
  • The Users
  • The Search Engines
  • People with Disabilities
  • Or our Differences

Which of them is really guilty?

If you really consider it the UI designer, the IA is to blame, that’s us, and we are the guilty part here.  We have never give the Tag Cloud a chance, never allowed it to mature and grow.

Lucky for us they have not died but have just matured out of sight and are still in use on the fringes of the web communities.


[1] 28% of Online Americans Have Used the Internet to Tag Content (PDF)

[2] Improving Tag-Clouds as Visual Information Retrieval Interfaces (PDF)

Tags: , , , , , , , , , , ,

Looks like there is no conversation here yet, why not start one.