Thursday, September 04, 2003
Zipf volatility
There's so much work that's good but isn't popular. That's why my first and second posts to this weblog were about non-blockbuster movies.
Clay Shirky has been pounding Zipf's Law as if the world were made of nails. Zipf's law states' roughly, that, "the 100th most popular web site would get 1/100th of the traffic of the most popular site (Yahoo), and the 2,000,000th most popular web site would see 1/2,000,000th of Yahoo's traffic." He's right, of course -- hits on websites and blogs are distributed much the same as viewership of TV shows -- a few sites and blogs and shows get very high ratings, and most don't. And he's right that there are, to a first approximation:
But I think there's a key property of works -- especially on-line works -- that Clay doesn't cover. I call it "Zipf volatility." Yahoo and Google are permanently popular; they have low Zipf volatility. But my hypothesis is that there's a middle tier of blogs with high Zipf volatility, where a well expressed idea or a funny story or a new factoid can rapidly catapult a blog from #100,000 to #1000, or in rare cases even to #10, in a matter of hours.
I am not sure how you'd test this idea experimentally (comments appreciated), and I am afraid that if you take 100 blogs, say between #100 and #200, and look at their delta-rank over a one week period, they might not look any different than the blogs between #20,000 and #20,100. Despite this caution, I strongly suspect that blog rank (and web site rank, to a lesser extent) has a burstiness that is not characteristic of other media, that permits new ideas (and new sites and blogs) to bubble up and subside, to move more readily than other media along the x-axis of Zipf's Law.
I could be convinced otherwise, but it'd take a good experiment.
Clay Shirky has been pounding Zipf's Law as if the world were made of nails. Zipf's law states' roughly, that, "the 100th most popular web site would get 1/100th of the traffic of the most popular site (Yahoo), and the 2,000,000th most popular web site would see 1/2,000,000th of Yahoo's traffic." He's right, of course -- hits on websites and blogs are distributed much the same as viewership of TV shows -- a few sites and blogs and shows get very high ratings, and most don't. And he's right that there are, to a first approximation:
"three tiers [that] drive media strategy. The bottom tier -- the millions of sites with very little traffic -- will be composed almost entirely of labors of love, or expensive subscription-driven sites with very selective offerings. The middle tier -- say, the 1000th to 100,000th most popular sites -- will use its closer connection with its users' interests while banding together into networks, web rings and so on in order to aggregate their reach. The top tier, the Yahoos and Geocities of the world, may be victimized by its own success, managing increasing problems of supply and demand."
But I think there's a key property of works -- especially on-line works -- that Clay doesn't cover. I call it "Zipf volatility." Yahoo and Google are permanently popular; they have low Zipf volatility. But my hypothesis is that there's a middle tier of blogs with high Zipf volatility, where a well expressed idea or a funny story or a new factoid can rapidly catapult a blog from #100,000 to #1000, or in rare cases even to #10, in a matter of hours.
I am not sure how you'd test this idea experimentally (comments appreciated), and I am afraid that if you take 100 blogs, say between #100 and #200, and look at their delta-rank over a one week period, they might not look any different than the blogs between #20,000 and #20,100. Despite this caution, I strongly suspect that blog rank (and web site rank, to a lesser extent) has a burstiness that is not characteristic of other media, that permits new ideas (and new sites and blogs) to bubble up and subside, to move more readily than other media along the x-axis of Zipf's Law.
I could be convinced otherwise, but it'd take a good experiment.
Comments:
Post a Comment
