Sunday May 14, 2006

'lanches

To grow, blogs depend on links from other sites.  If nobody else ever adds you to their blogroll or links to you posts, it's very hard to increase your readership—there's always a steady trickle of search-engine hits, but recommendations and links from other bloggers seem to result in more visitors, more of whom eventually become regular readers.  Over the last month, I've been fortunate to receive links from four big bloggers, and I thought it might be interesting to analyze the traffic that resulted from each link.

[Fair warning: this post will be mostly inside baseball—bug out now if you're not interested in the gory details of blog stats.]

Getting a link from one of the big dogs is a momentous enough event that there are some special terms used to refer to the sudden influx of traffic.  Getting linked by Slashdot or FARK is called getting Slashdotted or Farked, respectively.  A link from Instapundit produces an Instalanche.  All of these are a mixed blessing—web servers have been known to buckle under the flood of incoming connections.  None of my links produced such a flood, fortunately, and TypePad probably would have been able to handle one in any case.

Here's a snapshot of my Site Meter statistics for the last month.  The green graph is the number of unique visitors; the purple graph is the number of page views:

See the four peaks?  Those are, respectively, the traffic bumps corresponding to an Instalanche, a link from Language Log, a link from Jerry Pournelle, and a link from Language Hat.  You might expect that the number of visitors resulting from each link would be closely related to the amount of traffic the originating blog gets, but that doesn't seem to be the case.  The Instalanche (to this post) was the largest by a wide margin, it's true, but the margin wasn't nearly as wide as you might expect from the number of daily visitors Reynolds gets—about 120,000 according to his Site Meter stats.  The Logalanche (to this post), the Hatalanche (to this post), and the Pournellealanche (to this post again) were all of roughly equal size, in spite of the varying numbers of visitors those sites get.

Here are the numbers in more detail:

-alanchesource's
daily
visitors
one-day
traffic
increase
total
excess
visitors
Inst- 121,000 +280% 1300
Log- 6,100 +44% 285
Pournelle- ? +68% 415
Hat- 600 +52% 380

These values were derived by examining both my Site Meter and eXTReMe Tracking stats and rounding the averages off to the nearest attractive number.  The number of daily visitors to Instapundit and Language Log come from their Site Meter statistics (here and here), while Language Hat's daily traffic was derived from his eXTReMe Tracking stats (which for some reason tend to be about 5-10% higher than Site Meter stats, I've noticed).  Visitor numbers aren't available for Pournelle's site, but he's been at this for a long time, and it's likely he's receiving a lot of daily visitors.

The value in the third column, the one-day traffic increase—the percentage by which this blog's traffic increased on the day with the largest bump—is clearly somewhat correlated with the traffic of the linking site, but the relationship isn't very linear.  The same is true of the total number of excess visitors (which also includes visitors on subsequent days and visitors from other bloggers who linked after reading the main 'lanche).  Note, in particular, that Language Hat's link produced both a higher one-day bump in traffic and a larger total number of excess visitors than the link from Language Log, which has over ten times Hat's daily traffic.  Why is there so much variation?

I think it has to do with the sorts of posts, the sorts of blogs, and sorts of bloggers where the links originated.  The link from Language Hat was in his only post of the day, and that post stayed at the top of his blog for a day and a half.  What's more, his post was just a paragraph long and only teased the subject of my post rather than revealing the punchline, and that gave his readers a lot of incentive to visit here.  The Instapundit link, in contrast, was one of nineteen posts Reynolds made that day, and it excerpted the conclusion of the post here that it linked to.  That means the link was only one of many his readers saw that day, and that it wasn't necessary to follow it to get the gist of his post.  The Language Log link was different from both of those: it occurred in the last of five posts that day, but it wasn't featured too prominently, appearing only in an update at the end of the post.  The link from Pournelle was in his daily list of links suggested by his readers, which I suspect isn't as well-read as his main column, but it still produced a big bump in traffic (which would probably have been a bit larger, by the way, if TypePad hadn't had a service disruption the following day).

Every link, then, is a unique occurrence.  The number of readers that follow a link is affected by the placement and context of the link as well as the extent to which the linked post is quoted.  It must also depend on hard-to-measure factors like how used a blog's readers are to having to follow links to get the sense of a post—if a blogger is a "linker" rather than a "thinker", readers are presumably used to short posts with interesting links, producing a higher percentage of click-throughs (clicks-through?).  Similarity in subject matter also clearly has some effect—notice how the link from Language Hat also produced a large number of page views as his language-inclined readers read other posts here they found interesting.

So, what have we learned?  First, getting linked by other bloggers is good for your readership.  Second, links from high-traffic blogs are nice, but the quality of a link seems to matter more than the mere quantity of the linker's readers—prominent links are better than ones that are buried, links accompanied by quoted content from the linked post are less likely to be followed, and links from prolific bloggers are less likely to produce traffic than links from bloggers who post less often.  How can you apply these lessons to increase your traffic?  Well, you can't, directly—in the end, the only way to improve your traffic is to keep writing, hope other people notice, and so incrementally increase your readership.

[Now playing: "Ball of Confusion" by Love and Rockets]

I am The Tensor, and I approve this post.
07:23 AM in Weblogs | Submit: | Links:

TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d8341c88ad53ef00d834bfa4ca69e2

Listed below are links to weblogs that reference 'lanches:

Comments

Yeah, I deliberately withheld the punchline in an effort to get people to actually visit your site; I'm glad it worked!

Posted by: language hat at May 14, 2006 12:45:34 PM

You need to make the frequency which your logo jiggles at a function of how many visitors per minute the site is getting.

Posted by: includedmiddle at May 22, 2006 8:47:23 PM

I considered making the purpleness of the site start off bluer and fade towards pink in a five week cycle synced to my hair. It was too much trouble for something no one would ever notice, though.

Posted by: The Tensor at May 22, 2006 9:35:19 PM