I’m having a spirited debate about tagging with Matthias Müller-Prove (@mprove) on OraTweet, and I thought it would be interesting to open it up to long format and other viewpoints.
Plus, we’re flooding OraTweet, which I’m not sure people appreciate.
Photo by moonlightbulb on Flickr used under Creative Commons
Setting the Stage
I like tags. They’re quite valuable and provide an easy way to group information for quick retrieval, and if you have a large community using them thoughtfully, you can reap huge benefits.
Too bad that’s rarely the case.
Tags are a Utopian concept, like electric cars or universal health care, good ideas that create a better overall experience for everyone. Too bad not everyone wants to play along and contribute value to the community.
Background
When we introduced the IdeaFactory in 2007, we included tags. At the time, tags were a necessary bullet point on all Web 2.0 pitch decks, after all Tim O’Reilly included tagging as an evolutionary pointer in his What is Web 2.0 primer. So, duh.
Tagging proponents even coined a new word “folksonomy” to contrast the new hotness of a social classification system with the old and busted rigors of a centralized, machine-managed taxonomy.
Web apps like Flickr and Delicious were shining beacons of tagging’s success, demonstrating how an informed and thoughtful community could use tags to create faster and better information retrieval.
These were the salad days of tagging.
Experience
Over the years managing IdeaFactory, which eventually became Connect, I’ve dealt with many tag-related user queries, and the sad fact is that tags are too much work for the average user.
Maybe it’s conditioning by dropdown-list or list-of-values, but when presented with a text input field labeled Tags, most people just skipped it. Many who tried to use tags were confused. What are they? Are they public? Are they case sensitive? Can anyone use them? How do I know what tags to use? What if I mistag something? How can I delete a tag? Can you combined these tags for me? How do I create a new tag? Can I make my tags private?
Basically, tags weren’t intuitive enough.
We tried to alleviate confusion by throwing technology at the problem, e.g. type-ahead to suggest tags, but it became clear that tags weren’t getting used very much.
Beyond the confusion factor, Connect has fast and good search. Search is a known behavior for every user, and there’s no educational learning curve. Thanks to internet search engines, people have become adept at searching and manipulating their keywords to get the best results.
Search might not be better than tags, but it’s a known quantity and good enough.
So, why should users invest effort in a new feature when good old faithful search does the job well enough?
Maybe it’s Connect’s users. After all, we’ve always fought an uphill battle to get people to use Connect; Enterprise 2.0 and its sibling the social enterprise are only now finding widespread adoption. So, being ahead of the Bell Curve could have hurt.
But guess what reference points people have for social today?
Facebook and Twitter, neither of which include classic tagging.
Fast Forward to Today
You used to see a tag cloud on nearly every blog and web app, not so much anymore. Paul (@ppedrazzi) and I always shared the same skeptical view of tag clouds, i.e. they only make sense if you understand and use tags, thereby limiting their value and taking up valuable space. I suspect many large properties agreed and have removed their tag clouds, even if they continue to tag their artifacts. See TechCrunch, GigaOm, Mashable, and most other big traffic spots.
Chris Messina‘s (@chrismessina) hashtags have become the new folksonomy, easier to use from a UI perspective, but equally difficult to explain. So much so, that there are actually primers on how to use hashtags properly.
I kid you not, search for them.
The point is tagging is hard, and search is easy.
That’s not to say that search is perfect, or that tagging is a failure. Neither is true.
Through Matthias and Peter Reiser (@peterreiser), I know that Sun had a vibrant internal community, Sunspace, where tags were used with much success. It seems that people had no other choice, but to use tags, which forced them to understand the concept and get the maximum value from tags, a great way to educate people.
Flickr and Delicious are models of the same, i.e. a dedicated community that collectively agrees to get value from tags.
Too bad all communities aren’t like this, and if you want to build a community at all, I’d avoid tags altogether because they’ll only cause you headaches.
So, until all communities are Utopian, I’ll stick with the reality of search.
Find the comments.
I think the people that came up w/ “folksonomy” missed the boat; should’ve been tagsonomy. Then again, I have an unhealthy affection for puns.
I’ve found tags to be useful on a personal level, mostly as a memory jog when the search terms I can think of are a bit too broad to be helpful (“Oh yeah, I put that in my ‘R12’ bucket last week, let me give that a quick scan”). I’ve never really given much thought to how well my tag choices meshed with a larger community (Delicious, etc). I figured that popular tags would emerge in a kind of Darwinian fashion; i.e. the most useful/”right” tags would just become self-evident.
I could certainly live *without* tags, though, so I could see where some/many people can’t be bothered.
The type-ahead thing didn’t do much for me, until I had enough commonly-used personal tags. Kept me from pluralizing tags accidentally.
You raise an important point. Tags are a great personal tool for organization; once you add an entire community, results vary wildly.
Even so, you can live without tags, which is kinda the point I’m making here. Sure they’re useful if used in a thoughtful and cooperative way, but ultimately, tags are a nice-to-have feature.
Search is a must-have.
Full agreement there, search needs to be much higher up than tagging on any requirements list.
Right, so investment-wise, I prefer to make search better vs. adding tagging.
Thanks for the thoughtful post Jake. You make a bunch of great points.
A core issue is training & learning. Even for controlled taxonomies with drop-down lists, many users find the lists unclear, lacking important options, or to be in service of the site’s structure rather than their own content uses.
Tags address that limitation by allowing users to select keywords that make sense in their own minds. But good tagging is tough and often not taught properly.
What’s the best way to get people to use tags? Have them ask the question “what terms would I use to search for this page in six months from now?” In this approach tagging both becomes a service for the individual (addresses the “what’s in it for me?” issue) and a service of search.
Often both structured taxonomies and tagging don’t appear to help the user at all and, therefore, are a hard sell.
On at least a couple of oracle fora, the largest tag is badgerbadgerbadger. In effect it is much like the icons here. I like that guy!
On flickr, my own tags are inconsistent. I wish I had been consistent. Part of the problem is the not-so-clear upload interface.
Exactly my point, training is the issue, bc tags only offer value if they’re used, and they offer incrementally more value if they’re used more. So, you’re constantly fighting uphill until you reach a point where a large percent of your users are tagging.
In my experience, it’s hard enough to convince people to use tools, let alone learn how to use tags effectively. So, they’re doomed from the jump.
This is why I’d rather invest in better search and other tools (stars and likes are a form of tagging, as Matthias pointed out) that require little to no training.
The icons here? Does not compute.
Personal tags are great for a while, until you reach a critical mass and can’t remember what you used for what. Again, it’s perfect world stuff. I equate tags to organizing email, yet another crutch for search that puts the work on the wrong shoulders.
I tend to treat personal tags as form of communication about the content object itself – like some kind of metadata of cryptic meaning or even emotion. But as a search enabler? Nah. Kinda makes sense that tags aren’t used all that widely or usefully – look at the average book and determine how useful the index is. The great thing about a really powerful search and crawl framework is that it can find your objects without being held hostage to degrees of articulation of the creator. Even better thing about search – it’s killing off the lunatic creation of nested folders and the the sanctity of their naming conventions – just throw it in there and search…
I tend to treat personal tags as form of communication about the content object itself – like some kind of metadata of cryptic meaning or even emotion. But as a search enabler? Nah. Kinda makes sense that tags aren’t used all that widely or usefully – look at the average book and determine how useful the index is. The great thing about a really powerful search and crawl framework is that it can find your objects without being held hostage to degrees of articulation of the creator. Even better thing about search – it’s killing off the lunatic creation of nested folders and the the sanctity of their naming conventions – just throw it in there and search…
I equate the index of a book to search, i.e. a standardized way to find terms. Tags are more like hightlighting or dog-eared pages, personally meaningful.
Again, we’re proving the power of personal tags, and search can be so much more powerful and helpful, just as you say, bc it’s independent of the content creator.
Algorithms rule.
I am also a big pun fan, do yo know the one pun in ten joke?
Tags in the enterprise, or more specifically in enterprise apps may be a bigger success. The users of enterprise apps are well trained and a dedicated community (dedicated to doing what they are told also works).
For as long as I can remember we used tags in the bug database, e.g. ‘10.6:10.7:FWDPRT’ added to the description, of course tagsonomy wasn’t invented in those days, but we used the same principles and now we have added tags to more of our apps and I do see them starting to be used more and more.
The problem I remember w bugs is knowing exactly which tag to use, which is a purely enterprise problem. That implementation is a taxonomy bc someone on high decided what to use and enforces it. It’s an interface problem that should have been solved w a dropdown, given the key reports that rely on proper metadata entry.
I see enterprise apps tagging being similar. What tag should I use for invoices? What tag should I use for this type of supplier? Tags have to be right for reports. So, lots of tag gardening for the apps sys admins. I guess we’ll see.
This is the use case for tags in the enterprise: A small group of people want to track something fast and for a short amount of time. I could go to IT and get them to add something to a drop down list somewhere or I could just agree on a tag with the interested group and then start using it right now. So if I want to track all my team’s expense reports for lunch at OpenWorld, I can get them to tag it with haimes-oow11-lunch and then I can pull a report on it without involving IT at all and without whatever list of values I might use growing full of out of date values.
I know the use cases for tags in enterprise apps. I was involved in the initial discussions. There’s a good chance I suggested your favorite use case.
After some experience, my main pause with tags is the learning curve. How do you connect the dots between a valid use case and tags? It’s not terribly obvious, and users have to learn.
So, I’m not debating this is a valid use; I’m just not sure how it will play out in practice, kinda like tags elsewhere. Some people will love them and find enormous value; they’ll be in the minority, as the rest ignore them or struggle w nomenclature, capitalization, training, gardening, etc.
I’m saying the use case is working out nicely in Oracle development. Tags are being used in a disciplined way, because one thing enterprise networks are better at than the ‘social network’ is enforcing discipline.
Ah, got it. Discipline and enforcement, the hallmarks of enterprise-grade.