Blog Traffic Metrics

Warning: What follows is likely of no interest whatsoever to non-bloggers. And maybe not even to bloggers.


Recently, N.Z. Bear wrote a piece at TCS analyzing the decline in traffic that most blogs experienced after the election. A footnote at the end of his piece provides an important disclaimer:

[I]t should also be noted that not everyone in the weblog world believes that SiteMeter data are perfect, or even the best counter available. Some bloggers scoff at SiteMeter, claiming that the results it provides undercount the actual traffic they see by measuring traffic more directly from their server logs. These complaints may or may not be true: I frankly don’t know. But more to the point, neither does anybody else. Debates over SiteMeter reveal a more significant underlying problem in the blogosphere, which is that there are absolutely no established standards for measuring traffic, and (to my knowledge) there have been no studies done to rigorously measure the accuracy of SiteMeter or any other of the many popular weblog traffic counters.

This is a real problem, and one that will only grow in importance as weblogs continue to take their place alongside traditional media as a source of information and entertainment. Blogging is no longer exclusively a hobby done for the sheer pleasure of it: for some, it’s a business, with real money coming in from real advertisers — who want to know exactly what real traffic they’re paying for. And accurate measurements of traffic are important for other than financial reasons: a new blogger can bootstrap himself into a position of greater visibility and influence by demonstrating credibility based upon their audience levels — if the metrics are available, and more to the point, believable.

For now, though, we make do with what we have. And while questions may exist around whether SiteMeter provides perfectly accurate absolute measures of traffic, it seems reasonably safe to use SiteMeter data for the relative comparisons that I have provided here.

Michael J. Totten was surprised by this assertion:

This is news to me. I̢۪ve used SiteMeter ever since I started this blog and always assumed it was reasonably accurate. Maybe it̢۪s not. So I checked my traffic level using Webalizer (which is installed on my server), something I̢۪ve never done before. And I was shocked. SiteMeter says the average number of visits per day on my blog is 3,200. Webalizer says my daily average is almost 6,000 for the same time period.

What̢۪s the deal? How can these two traffic counters be off by so much? Does anyone have an informed opinion? (Both SiteMeter and Webalizer claim to count a single visit as all activity on my Web site from a single IP address with less than 30 minutes of idle time in a row between clicks. So it̢۪s not like they̢۪re comparing apples and oranges.)

A wonky discussion ensues in the comment section.

In my own experience–having used SiteMeter for just under two years now and having stats from my site host going back nearly as far–SiteMeter is very useful for a free hit counter but it has many serious drawbacks. It’s frequently down and often several hours behind. It crashes constantly. This past Saturday, it was down all day at my site and I wound up with something like 1500 uniques counted, less than a third of a typical Saturday.

SiteMeter isn’t even internally consistent, with its total count often well off of the cumulative count of the hours/days/months. During the Nick Berg traffic spike, several of us “lost” tens of thousands of visits on the total as compared to the daily figures. Indeed, it went down on my site and Jeff Quinton’s, that I know of, that morning and didn’t come back into normal swing for a couple of days.

Further, SiteMeter is on multiple servers so not really a good comparative device at all. I’ve had it be down all day at OTB and see it working fine on other sites–and vice versa.

Because of its ubiquity, SiteMeter is the best device we have for comparing Web traffic across the board. But that’s not saying much.

FILED UNDER: Blogosphere,
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. Chris Short says:

    I use Webalizer on my server. Direct log analysis. Nothing better in my opinion either. SiteMeter is something I’m testing out this month and so far I’m really not impressed with it AT ALL.

  2. Zed says:

    I’m not sure if I like webalizer or awstats better, I like having them both to compare with analog which is pretty cheesy, webalizer seems to be more consistant but awstats will give a complete search engine list for every search phrase which brought viewers to your site as opposed to webalizers top 20. I like having them both to compare to each other. I don’t even use sitemeter, I don’t feel I’m missing anything other than my readers having any idea what my traffic is. But I really don’t care about that.

  3. reliapundit says:

    i have noticed that sitemeter is ALWAYS off by 50% compared to extreme tracking!

    why is this so – and why is it so much like what totten avows?

    i think that the lefties bloggers have hacked a filter on to sitemeter which does this to rightie bloggers (and DOUBLES their traffic stats, too — how else could kjos – who’s a moron – have numbers twice as high as glenn!?!?). iot’s this, or sitemeter is doing it themselves.

    i’ve also noticed that sitemeter consitently misses some links that i know work because i have used them/tested them.

    i think that someone is deliberately effing with the stats because the leftie bloggers got caught trying to eff with the wizbang web awards.

  4. Jay says:

    I have seen clear indicators of Site Meter being off, sometimes hugely, and at its worst during particularly heavy traffic. I wrote about it once upon a time. Basically if you take your tally at the end of one day, subtract it from your tally at the end of the next day, and compare to Site Meter’s number of hits it says you got for the day, they only match some of the time. I saw over, I saw under, I saw matches, but on the whole it was under, by enough to add up over the course of time.

    It is worst during a lanche of whatever kind, and that was how I noticed it.

    So I take Site Meter to be an approximation, and to be perhaps internally consistent enough that comparing the relative traffic of two blogs that use it is close enough for CBS work. I think it’s pretty safe to say that if I get 200,000 hits per SM in 2 years, and you get 2 million hits per SM in 2 years, that you did get close enough to ten times the traffic I got.

    It also never helps when it goes down. It tends to do that *right before a milestone* in uncanny fashion. About to hit 500,000? Watching the meter and referrers to see who wins the mention in a post? So sorry! The meter will stop just before 500,000 and mess with your plans.

  5. jen says:

    I use StatCounter.com and find that it’s closer to my server stats than SiteMeter ever was. It’s easier to read the data as well.

  6. Jay says:

    How cool! The software interpreted my asterisks put around words for emphasis as a bold tag. That’s a nice touch.

  7. Rob says:

    Once a month or so I’ll log onto sitemeter mid-day and it will say zero hits for the last hour, so I know it is down because I get enough traffic to assume that someone is always stopping by. I figure it goes down from time to time, and that is part of the problem.

  8. Ian says:

    I would never, ever use Webalizer. I have my own web server and it’s not accurate at all. It produces 5x as much traffic as I get. What I think it does is count the references to your site and count it as a hit. For example if your blog link is mentioned in someones website, the number of people who view that entry on someone elses blog, it counts as one hit.

  9. Ian says:

    Also..I agree with Zed, if you’re just blogging for stats and have people view your blog, please stop blogging than. I check my stats now and then to see if there is any spike in it. If so I will see if I have any trackbacks, so I can thank the person or track back to them to return the favor.

  10. Bithead says:

    Trouble is if you wanna be listed on The Bear’s system ya gots to be on SiteMeter.

  11. I finally registered by blog URL (used to piggyback it onto my business site) and found immediately that Sitemeter undercounted by at least 50%, but it also appears to me that Sitemeter counts non-IE browser traffic at an even lower rate.

    I appreciate Sitemeter but I don’t understand Bear’s thinking that there’s no way to measure site traffic accurately. Of course there is! You can’t access my site(s) without my ISP knowing it, and Webalizer counts that traffic just fine.