22 December 2005 - 3:16pm
On readers, page views and Daily Kos: What's in a number, anyway? [updated] [updated]
Via Garance Franke-Ruta's American Prospect blog post, I see that the Washington Monthly has a new article by Benjamin Wallace-Wells claiming to profile Daily Kos proprietor Markos Moulitsas Zuniga.
One passage, cited unquestioningly by Franke-Ruta, caught my eye:
The site, which has existed for only around three and a half years, now has 3.7 million readers each week. That's more than the top 10 opinion magazines—of both left and right—combined, more readers than any political publication has had, ever, in the history of the world.
That all sounds nice. But it's not true.
According to BlogAds, which has public statistics for all sites running BlogAds advertising (to inform potential advertisers), that 3.7 million/week figure (which BlogAds has this week as above 3.8 million) is not for readers, it's for page views.
Considering that each visitor is probably viewing many pages -- after all, you have to load new page views to read front-page stories, to read diaries, to read the comments to the stories and diaries, etc. -- and that many, if not most, people are returning to the site each day, the number of unique readers over a given week is but a fraction of that figure.
This is very basic web stats stuff. And something that should not have slipped by the Washington Monthly fact-checkers.
I note that Kos posted some other corrections to this Wallace-Wells piece, but not regarding the gross inflation of his site statistics.
It's great that Daily Kos is doing well -- we all benefit when alternatives to corporate-owned and -sponsored media succeed -- and Daily Kos is still by far the most active website in the political realm. But nobody is served by such false figures, or claims that Daily Kos has "more readers than any political publication has had, ever." It's just not supported by the facts.
And it makes the Washington Monthly, the American Prospect and the vaguely "progressive"/"liberal" blogosphere look bad.
----
Update 2005-12-23: Another source for web stats on Daily Kos is SiteMeter, which, as I write this, reports 5.5 million page views and 4.5 million "visits." In this context, a "visit" is a slippery term as a person going away and coming back 20 minutes later might be counted as a new visitor, depending on the settings of the counting software. (15-60 minutes is the typical setting.)
There's no doubt that it represents a ton of traffic.
So why does this differ so much from the BlogAds stats? One reason is that the BlogAds are on the content pages, while the SiteMeter stats measure all page loads, including composition screens -- which see much use, given the wildly active discussion threads there. The BlogAds figures are a more reliable indicator of actual pageviews of content being read, or at least viewed.
Does that translate into "probably" 500,000 actual readers each week, as Kos now claims? That's "probably" stretching it.
Update Two 2005-12-23 7:15pm EST: Washington Monthly writer Benjamin Wallace-Wells has posted corrections, which say, in part:
When I wrote that his site gets "3.7 million weekly readers," I should have used the technical term "unique visitors," which is the closest available approximation of a website's readership, but is certainly bigger than the actual number of weekly readers.
media girl also blogs at other places.
Similar entries
store
Buy stuff here.
Comments
Good evening, Media Girl.
You are right on the mark. Most "hit" meters measure pageloads. Moreover, most pageload meters do not make any effort to eliminate multiple hits by a single user in short sequence, as you explained happens on multi-page sites. Multiple hits are counted on blogs that have double-layer loads. You might notice that some blogs seem to "load" twice or even more before they come to rest and are navigable.
On my own blog, I provide both a pageload meter and a meter that measures the number of actual, "unique" visitors who have ever come to the site. Even my pageload meter I try to manage so it doesn't count one user too often, but this is a bit difficult to keep an eye on. My advertisers have their own ways of seeing traffic as they need to: they call their metric "impressions." Although I could impress some with hit meters that spin to high numbers more rapidly, it serves no purpose for me in understanding the composition of my audience and what is bringing readers to the site.
Apparently, however, a less disciplined metric presented visibly would impress some of the mainstream media sorts who decline the opportunity to understand basics of Web traffic and the various measurements of it.
The Dark Wraith has blustered enough about this subject, now.
most large sites like dkos tend to have a very low average number of "page views per visit". on dkos that number is about 1.2, which means the number of page views vs number of viewers is almost the same. on a smaller site its typical to see a view-per-visit ratio of 3-5 or more. in the case of dkos, the downside of this low ratio is that it implies a "lower quality" of visitor overall since the proportion of silent spectators is very high. also there is a lot of variability in the trends of web traffic depending on current events (see the recent Google Zeitgeist 2005 report for example), so its hard to say exactly what the average level of traffic is for dkos, probably it would be safest to look at the average over the last 4 months or so at least.
Are you making this up, or are you speaking out of school?
Using page view counts to announce website traffic is more akin to announcing magazine circulation based on how many magazines times how many pages are in the magazine.
Nearly all of the participants on the site must load the page dozens of times a day -- to read a diary, to read the comments, to read another diary, to read a front-page story, to see if there are more comments on the diary, and so on.
I loaded Daily Kos five times as I wrote the blog post just to see if he was going to mention any other "corrections."
And considering that most site visitors who read a blog will return several times during the week, it's clear that the notion that Daily Kos has "more readers than any political publication has had, ever, in the history of the world" is just based on a fundamental lack of understanding of the internet.
If you read a magazine, and put it down, and then pick it up later, does that count as another circulation number? Does that make you count as two readers of the magazine? Of course not. But it does in webstite traffic stats. And if you pass that magazine on to a friend to read, does that end up getting counted in circulation numbers? No. But it does in website traffic stats.
Even counting website "visits" (i.e., unique visitors who view one or several pages during one visit) cannot be compared with print circulation numbers.