Thursday, October 14, 2004

Hit Me

Sometimes I'm utterly fascinated by data. (Re yesterday's post and its reference to vital statistics in step 7: Did you know that the 10th most common death of males aged 20-24 is getting shot by the cops?)

Of course I'm obsessed with my Site Meter stats. Not about who you are; that isn't important to me (I won't tell that you should be working). Rather, I wonder what drives the traffic numbers and their wild oscillations.

graph of site stats
Figure 1: This month's visits by day

Some things are predictable. If a high-prestige, high-traffic blog mentions something I've written, then my numbers go up for a day. Site Meter tells me that usually these new visitors view only one page -- the one referred to by the other blogger. Sometimes I'm lucky and I'll see someone with more page views. Two common explanations: either they're reading other stuff that I've written or they're checking to see what I've linked to (when they hit the back button my page reloads, recording another page view). Only a few of these become regular readers.

When someone leaves a comment, I'm more likely to get hits from people clicking Blogger's next blog button. And I'm more likely to get repeat traffic from people interested in the conversation. When I'm having a lot of comment action, Site Meter tells me that a higher proportion of visitors will view more than one page. I don't leave many comments elsewhere (I do most of my blogreading through Bloglines, so it takes a few extra clicks to leave a comment), so I'm not sure how or if that would change my traffic.

The effect of being on blogrolls doesn't show up in the daily traffic numbers. Over the long term, sure, but not on a daily basis. Some of the blogroll-generated traffic comes from people who read a bunch of blogs on a regular basis and who have a standard path that they follow through the directed graph of links. As her popularity (and blogroll) have surged, I've received fewer and fewer hits from Professor B., but the number coming from smaller sites, like New Kid on the Hallway, or Faculty Wife, or Yes, Yello Cello is fairly steady. If I was going to model the number of hits I get from being on someone's blogroll, I'd conjecture that was proportional to the log of their daily hits and inversely proportional to the size of their blogroll. And there'd have to be adjustments to account for the alphabetical listing (blogs starting with A would get more hits) and for Blogrolling's recently updated notations.

But the rest is such a mystery. Contrary to conventional wisdom, the weekend numbers are not appreciably lower than the weekday numbers. Some days I'll have bunches of hits before breakfast, and sometimes it feels like everyone on the internet has made a secret pact not to visit my blog.