Blog Traffic Numbers Inflated?

Patrick Ruffini argues that the traffic numbers for Daily Kos are “fishy” owing to the oddity in how SiteMeter works. The technical details are laid out in some detail and are interesting, if a bit confusing. Essentially, as I understand it, SiteMeter can’t keep up with massive amounts of traffic flooding a site and thus inflates the extreme high traffic blogs. Additionally, the disparity between “unique vistors” and “pageviews” is an issue.

There’s no doubt that SiteMeter’s numbers are skewed. The only reason bloggers with their own domains, who have substantially more useful statistical data at their disposal, even bother with SiteMeter is that its ubiquity allows cross-blog comparison.

There are, however, other metrics. Perhaps the most useful publicly available one is the BlogAds order page. It’s not as widespread as SiteMeter, since those who aren’t BlogAds network members aren’t included, but it’s all measured by a common server. So, for sites using both measures, we can compare their BlogAds weekly numbers and their weekly unique visitors (daily average multiplied by 7):

Site BlogAds SM Uniques *7 Ratio
DailyKos 4,276,718 3,186,680 1.34
Crooks and Liars 1,780,917 1,206,947 1.48
FireDogLake 406,926 310,961 1.31
The Smirking Chimp 284,194 156,359 1.81
myDD 268,097 123,683 2.17
Political Animal 259,991 218,911 1.19

Granting that this is a fairly small sample, there doesn’t seem to be any obvious pattern here. The ratio of BlogAds impressions to SiteMeter unique visits ranges from 1.19 to 2.17 with some clustering in the 1.3 range. If anything, SiteMeter seems to be slightly undercounting the extreme high traffic blogs compared to those with more manageable numbers.

UPDATE: Meryl Yourish explains more about how SiteMeter works and why Ruffini is off base. She also explains how to increase traffic by writing about sex with Democrat superheroes.

Please follow and like us:
FILED UNDER: General, ,
James Joyner
About James Joyner
James Joyner is Professor and Department Head of Security Studies at Marine Corps University's Command and Staff College and a nonresident senior fellow at the Scowcroft Center for Strategy and Security at the Atlantic Council. He's a former Army officer and Desert Storm vet. Views expressed here are his own. Follow James on Twitter @DrJJoyner.

Comments

  1. Cernig says:

    I prefer StatCounter to Sitemeter, in any case.

    Regards, C

  2. Adam Graham says:

    Site meter tends to undercount from my experience when I compare it to internal blog stats.

  3. Moonage says:

    There is an important distinction made that I think you’re missing out on. I use Sitemeter as well. It tends to be fairly accurate I think. However, I am definitely a low-end user. You probably are too. Probably 95% of the blogs out there fall into this category. What happens is when you start getting more than 100 page hits per 100 users. Sitemeter counts the last 100. If they’re still there, when they hit that second page, they become a new unique user ( the previous user then has a 0 read time since they never “left” ). So, looking at other blogs and trying to compare them to Kos, Malkin, and possibly OTB, is meaningless. I like Sitemeter fine, however, I have had high-volume days ( 58,000+ ) where I did notice something just didn’t seem right and it did not reflect my Google Analytics result at all. Now that volume is back to normal, it’s fine again. I think Ruffini nailed it. This effect isn’t necessarily true all the time either. So, if Site A gets a steady stream of hits all day, and Site B gets a sudden surge of hits, the results will be different. So, using BlogAds or some other norm to compare them at some point in time doesn’t strike me as working too well either. If both sites had the same flow patterns at exactly the same time, then a norm would work. Kos using Sitemeter hits to state he has that many readers isn’t appropriate either. There is definitely some skew in how Sitemeter records viewers and page hits. But, for the price, ( I pay nothing ), you can’t expect the top of the line service.

  4. Bithead says:

    Adam’s experiences, and my own are pretty much the same.

    My own internal site meters are recording five and 600 HPD, on days when Sitemeter only records 100 or so.

    now, normally, I would attribute the other hits that my internal counters are recording to such things RSS feeds for places like bloglines, feedburner etc, plus people hitting the RSS feeds directly from my site.

    Thing is, I would expect these automated feeds to remain relatively constant. I have noticed, however , that the automated stuff that identifies itself as such goes up in concert with the hits that SiteMeter records. So, I’m not quite sure what to make of all this, but I know darn well that site meter isn’t giving us accurate readings.

  5. Mark Jaquith says:

    There is absolutely no way to convert between visitors, hits, and page views. There’s no way to convert between visitors measured by one service and visitors measured by another service, because the definition changes. The best, and most meaningful metric, is page views. And page views are best measured with JavaScript, to prevent robots from counting. So SiteMeter’s page views metric, assuming their service stays up, is a pretty good metric.

    Yourish’s post is spot-on. Ruffini is making the rather silly assumption that if (free) SiteMeter is only showing you 100 visitors of data, it’s only tracking 100 visitors. That would be an incredibly stupid limitation of the service. How would they expect people to upgrade if their free version has such an arbitrary limitation? The difference between the free version and the paid version isn’t what (or how many people) they track, but how you can view that data.

  6. Bithead says:

    No, I think you’re misreading Patrick. One of the first things that anyone gets out of sitemeter, is the site summary, which gives you Total hits since starting the service AHPD, AVL, and HLits within the last hour.

    The best, and most meaningful metric, is page views

    Which, given that the TTLB status is wrapped around HPD would seem to be a problem.

    the problem here, as I see it, is that most of this stuff was thrown together in an ad hoc fashion, and now goes back so far into blogdom antiquity , that nobody really understands what’s happening.

    Which, in itself, may explain the inflated numbers that change points to.

  7. Bithead says:

    Tom McGuire has been doing some of his own testing. He basically confirms Patrick’s charges charges.