Thursday, June 23, 2005
Tags vs Search
Lemme take a min or two to outline some thoughts on tags and search, using Google as an example, because I think they illustrate some coming micromedia dynamics.
Let's imagine a zero search cost world. Whatever you want to find, you type it in, and it's found. In such a world, AdSense marginal revenues drop to zero.
The point I'm trying to make is that Google finds itself facing a similar tension to Macrovision. If Macrovision makes perfectly efficient DRM, it puts itself out of business. If Google makes perfectly efficient search, it loses revenues fast.
Now, let's think about tagging for a second. As the size of the tagosphere grows, the number of clicks it takes me to find the the thing I'm looking for drops.
In other words, search costs are falling rapidly as the tagosphere grows and gets interconnected between taggable apps across media.
At some point, tagging becomes more efficient than search. This point is reached when the number of Google clicks is greater than the number of related tag clicks (or similar proxy). When this happens, tagging becomes a perfect substitute for search.
Alternatively, you could consider a world where tagging picks up where search leaves off - you search for tennis racquet, click over to racquetworld, and then tag your way over to the perfect racquet. This is imperfect substitution, but still close enough to cause Google problems.
Now, in either case, I think we've explained a couple of fairly interesting things. First, why exactly it is that Google refuses to play with tags very much (despite the fact that even a not-so-ubergeek like me can code up a working tagging engine in less than a day); or, alternatively, why it is that Yahoo embraced tags. Second, why it is that Google's search gets more efficient across domains, but almost never within domains (ie, web search is barely more efficient than a couple of yrs ago).
Another interesting (geeky) point to note is that this line of thought helps clarify tag b-models. That's because it implies that the value of related tags rises nonlinearly. The first one isn't worth much, but later ones are more and more valuable. This is an example of the snowball effect: demand for tags increases at the margin.
1. Can you explain why Macrovision goes out of business if they make perfectly efficient DRM?
2. Can you explain why tagging is so effective at lowering search costs? This is not obvious, especially since large-scale tagging systems have yet to be hit by spammers (which they will when they get big enough). It may be true when taggers of are like mind, but presumably you want tagging to be for more than just the technorati (pun intended)?
This really is the year that you love to criticize Google for its strategic missteps, or however it was you put it in that post awhile ago. Yahoo! used to be your whipping boy, now...I guess it's all about the tags!
Brilliant post, IMHO, because it illuminated the issue so that even I could understand it.
"At some point, tagging becomes more efficient than search. "
I see that trend initially, but I do worry about tagging in the long-run.
How about a future where tagging becomes less efficient because of tag spam?
Or the increase in tags means less efficiency rather than more (I suppose it depends on the specificity of the tag itself. Right now, most tags are one word, whereas a good, focused search might have 3 or 4 keywords and other variables.)
Does this guy even read his comments?