logo_text_trans.gif
Click to see the XML version of this web page.
Wednesday, December 11, 2002

Web metrics, part II (first installment): The goal was to shed light on the relationship between referrers and traffic. We all understand what happens when you get Slashdotted, but is anything else operating? It would make sense that more referrers would lead to more hits, but is there any helpful information that can be teased from the data? Is there some magic number that relates referrers and hits? How important are referrers to a blog?

This time around. I chose to look at only the data on the Userland Community Server's 'ranking by page reads' service because the same relatively simple process generates hits and referrers (distributed services might make the data noisier, requiring longer periods of observation). Weblogs using Userland's Radio authoring software can choose to include a one-pixel .gif file on their pages that is linked to Userland's server, which both counts the hits and stores a list of referrers.

Again, this is hardly a scientific study: the data were collected over only one day (which seemed to be typical to this observer). I encourage others to repeat the observations and see if the numbers hold up. My observations:

Hits and links have a linear relationship

Weblogs, and Internet sites in general show a power-law relationship, where a few have large metrics, and many have small numbers. Referrers and hits, while noisy, clearly have a linear relationship: more referrers make for more hits and the relationship is roughly linear: it converges to an average 'magic' number.

The referrer 'magic number' is 3

A referrer delivers, on average, roughly 3 hits (page reads, actually) per day to a blog. High-flow blogs, including the 'celebrity' sites, get relatively more hits per referrer (4.76) while low-flow sites get around 2.6. The average for all Weblogs, minus the top-10 [OE]celebrities[base '] (whose traffic is being driven, presumably, by other influences) is around 3.6.

This seems to say that low-flow sites have to work harder: their referrers are, on average, worth fewer hits. But the work is worthwhile: sites with more traffic attract more potent referrers. The more traffic you get, the more it drives new traffic.

[Interesting: this suggests that the underlying reason for the power law distribution of hits on the Web is that, as a site grows, it attracts inbound links that deliver more and more hits, which drive faster growth, which attracts more potent referrers. It's a classic positive-feedback loop.]

There are 2 kinds of referrers, and 2 kinds of Weblogs

The most common referrer by far sends one to a few hits a day to a Weblog. Low-hit Web sites tend to have mainly have low-flow referrers. The least common referrer is high-flow: a Slashdot-style site that sends a relatively large flow. The two types sit at opposite ends of the power law curve.

The most common Weblog has a relatively good-sized base of referrers who each send relatively little traffic; the other type, a small minority, has a small number of referrers, one or two of which send relatively high traffic.

This is interesting: there are [OE]diffuse[base '] blogs that are widely linked to, and a few [OE]standalones[base '] that are favored mainly by one or sometimes two high-flow site. One seesm to be embedded in and widely accepted by the community, while the other is highly dependent on some 'parent' site.

Installment 2 tomorrow: what the data suggests for growing a blog...
Comments 3:09:43 PM    




Top of page | Home | About gulker.com | About Chris Gulker

Updated 4/16/04; 1:50:30 PM

Chris Gulker's view from Silicon Valley - in words and pictures

Updated 4/16/04; 1:50:30 PM


December 2002
Sun Mon Tue Wed Thu Fri Sat
1 2 3 4 5 6 7
8 9 10 11 12 13 14
15 16 17 18 19 20 21
22 23 24 25 26 27 28
29 30 31        
Nov   Feb

Gulker Photo Archive Logo

Features & Categories:
Columns (soon)
Dotcom Garden
Lone Genius Hackers
Picture Weblog
Theory & Strategy
Weblogging

gulker.com Cam
gulker.com Cam

Interesting blogs et al.:

AlwaysOn Network
Natalie d'Arbeloff
Azeem Azhar
Ken Bereskin
Blogging Ecosysytem
Blogging Network
BlogStreet
Boing Boing
Tim Bray
Matt Croydon
DaveNet
Rael Dornfest
Esther Dyson
Dave Farber's IP
Dave Fitch
David Galbraith
John Getze
William Gibson
Dan Gillmor
James Gleick
Bernie Goldbach
Meg Hourihan
Joi Ito
Xeni Jardin
Jeff Jarvis
Linux Journal
Mitch Kapor
Kuro5hin
Gunnar Langemark
Joshua Levy
Scott Loftesness
Macintouch
Ross Mayfield
Hans Moravec
Rafe Needleman
Nonsense Verse
OS Opinion
Tim Porter
Recommended Reading
Reverse Cowgirl
Glenn Reynolds
Roger Ridey
Phil Ringnalda
John Robb
Scott Rosenberg
Anita Rowland
Brent Simmons
Robert Scoble
Doc Searls
Jessica Shea
Gavin Sheridan
Shifted Librarian
Stefan Smalla
Bruce Sterling
Scripting News
Slashdot
Dan Shafer
John Tringham
Jon Udell
Moicho Umeda
Philipp Weltentummler
Kevin Werbach
Amy Wohl

Click here to visit the Radio UserLand website.

Subscribe to "Weblogging" in Radio UserLand.






Google