Blog Numbers Followup

Quick follow-up on yesterday’s piece, and I do mean quick, since it’s 2:30 and I have to be up in a few hours to take Athena to preschool:

* Several people wrote to ask why, when I was discussing my readership here vs. my readership elsewhere, I compared circulation numbers of newspapers and magazines with visitors to Web sites. The (quite accurate) point here is that circulation numbers showed the potential pool of readership for any one article in a paper or magazine, not the actual readership of that article (unless you assume that everyone reads papers and magazines cover-to-cover), whereas Web page visits actually register a visit (and, presumably, a read). Comparing the two is like comparing apples and oranges.

This is true enough, and I’ll grant it willingly, especially because I don’t particularly feel it invalidates my point. Even a fraction of the readership of a newspaper or magazine (or Web site like Slate or Salon) is larger than the total readership of most blog sites. To return to the example I gave of my DVD column in the Dayton Daily News, even if only 1 out of 10 readers that day glanced at it, that’s still a readership of 14,000, well outstripping my estimated 3,000 or so visitors over the course of a month (and coming close to my estimated daily readership for andrewsullivan.com and Instapundit). One-tenth of the circulation base of the New York Times is over 100,000, so any articles Sullivan gets in there would easily outdo his daily Web site visits. And of course, one assumes that more than 1/10th of the daily NYT audience would actually be interested in what Sullivan has to say — unlike me, people outside of his circle of friends know who he is.

(Also, any circulation manager will tell you that circulation numbers underreport total readership, since more than one person in a household will read a paper or magazine. When I worked at the Fresno Bee, the paper had a circulation of 150,000, but claimed readership of more than double that.)

Some people additionally looked askance at my estimation of 1,000,000 readers for my non-Web site work, which is fair enough. However, I’ll note the newsletters I write are opt-in, which means people have to sign up to get them, so the number there (500k) is pretty solid. I also feel pretty good about the DVD and CD reviews in Official PlayStation Magazine, since, if I may say so, the layout for these babies is pretty sweet; The DVD reviews go across two entire pages. You really can’t miss ‘em (buy a copy and see for yourself. Buy two! They’re small). Even throwing out the DDN numbers, my non-Web site readership outdoes my Web site readership by a multiple of at least a couple hundred. With multiples like that, the point still stands. Others ranging from Sullivan to Lileks to Marshall may have less dramatic multiples (they have much larger Web site readerships), but the multiples are still there.

* Getting back to Andrew Sullivan, his Web master got back to me with some updated numbers for the site. He writes:

“Andrew Sullivan’s website receives an average of 40K visits per weekday right now, or about, 25K daily uniques. This translates to about 200K monthly uniques, and slightly under a million visits.”

So, does this mean that the site receives 25,000 individual visitors daily, and 200,000 individuals over the course of a month? Well, no, not necessarily.

Here’s why: “Uniques” represent distinct IP addresses that visit a site; every computer on the Web has its own IP address. But there are two major types of IP addresses out there: Static IPs and dynamic IPs. Static IPs never change; these are for people who have their own servers. Dynamic IPs do change, every time you sign on. Most people have dynamic IPs because they go through an ISP, and most ISPs assign their users dynamic IP addresses when they go online (what, you think your ISP buys a new computer each time someone subscribes?).

So, for most people, if you visit this site, log off for lunch, then sign on again and come back, my site will log you as having two distinct IP addresses — and therefore as two unique visitors. If you had trouble following that, here’s the short version: Most people look like someone new every time they sign on to the Web. At least to a Web server.

(And actually, AOL uses “floating IP,” which means the IP address you use can change several times during the same Internet session!)

Therefore: Even the unique visits metric can overreport the actual number of visitors (it can also underreport if your company routes its Web traffic through a firewall, but I suspect that over the course of time, there’s more overreporting than underreporting, if for no other reason than AOL has 30 million subscribers and everyone else doesn’t). On a daily basis, this overreporting is relatively small (although probably larger on blogs than on other sites, since people often come back to blogs more than once in a day), but it compounds the longer the metric is used; use it to chart an entire month, and it’s probably not at all accurate.

You see where I’m going here. The 25K daily figure is probably not too far off; I’d trim about 5k from it to be safe, but, okay, 20K is still pretty damn good when you consider the Web. But the 200K number is very suspect. If I had to guess (and I don’t, but I will), I’d guess that the actual number of individual visitors is substantially lower: Say, about 75k on the outside, and of those, probably 50k-60k are regular visitors (which is to say, they stop by more than once).

Please note that I do not necessarily think Sullivan’s Web master was trying to pull a fast one here; he’s not to blame that the fundamental architecture of the Web makes it difficult to accurately gauge visitors over any large length of time (or, in dealing with AOL members, over any length of time at all). It’s simply a reminder that when you’re pulling numbers off the Web, they’re usually not what they seem; they’re usually a lot less.

* Also to be clear, I have no jihad against Andrew Sullivan in particular (or Glenn Reynolds, whose blog was also featured prominently in yesterday’s column). I used them primarily because a) Norah Vincent used their sites for the numbers in her LA Times opinion piece; b) As they are the best-known of the bloggers, they make convenient examples. For the record, I like both sites and both writers just fine (I don’t know either of them personally). Neither of them exactly has my politics, but neither does my mother, and I like her just fine, too. I’m aware that some folks have already started to use yesterday’s column as a cudgel against Sullivan, who has gained ill will in some circles. It’s not about Sullivan or Reynolds personally, it’s about numbers (or lack thereof), which, mostly as a matter of my convenience, happen to be theirs.

* At the request of irritated readers, I am dropping the apostrophe in front of the word “blog,” since the word has apparently entirely graduated into being its own thing and not just a shortened version of “Weblog.” One reader wrote: “A blog is a blog the way a phone is a phone.” He is, of course, referring to a ‘phone there.

One thought on “Blog Numbers Followup

Comments are closed.