Monday, April 17, 2006

Hiltzik's evidence needs more 'splainin'


I came close to migraine today over LA Times business columnist Michael Hiltzik's claim that Hugh Hewitt's Web traffic is declining. Bottom line after Excedrin relief, I still can't see the data his way. The numbers he provided me just don't jibe with the source he cited, Site Meter. Hiltzik and those who agree with him are welcome to point out where I am wrong. Send your observations to bradley (at) sandiego.com. (I apologize for not having comments, but this blog isn't even a week old, and I'm still learning the ropes).

Hiltzik wins a point right off -- he can't be faulted for using Site Meter for judging Hewitt's viewership, since that's what Hewitt uses. I was too quick to praise Independent Sources's analysis with Alexa.

There is a major flaw in measuring Web traffic through Alexa. Its measurements are performed through a tool bar that users have to install. Some don't install the Alexa toolbar because of fears it is spyware. And some can't install it, because they're not using Windows or won't use Microsoft's creaky, less effective Internet Explorer.

My own bias: I have long distrusted Alexa, perhaps unfairly. I prefer to err on the side of caution with rumored spyware. In any case, my viewing habits make it impossible to use Alexa. I am writing this at home on a Linux (XandrOS) computer. At work, I use a Windows XP computer running Mozilla Firefox, the browser that I also use at home. (A sister company to Alexa, A9, has a toolbar for Firefox, but you have to be an Amazon.com member).

According to Site Meter, less than 65 percent of Hewitt's traffic comes from versions of IE, and more than 25 percent come from Firefox 1.X. So 35 percent of Hewitt's viewership is beyond the reach of Alexa. Unanswered question: has that number changed significantly over time? And how much of Hewitt's viewership bothers to install Alexa in the first place?

But even dispensing with Alexa, I still can't find the decline Hiltzik wrote about.

I independently stumbled upon what Armed Liberal had found earlier, that the Site Meter stats for Hewitt show no evidence of a sustained decline. They fluctuate, but there's simply no evidence of any unusually large dip. Here is a view for the last month.

Try as I could, even with my MSM reality-distorting glasses, I simply could not get the Site Meter data for Hewitt to show a decline. The numbers Hiltzik emailed me simply didn't match up with what Site Meter showed me. I emailed him on this point Monday, and am awaiting a reply.

I see three possibilities:

(1) I am reading the Site Meter stats incorrectly

(2) Site Meter is showing incorrect information now and Hiltzik recorded the only accurate set of numbers

(3) Hiltzik goofed.

I am trying to be cautious, because I wrote too hastily in my first post. Interpreting Web stats is tricky. Some of the other commentators have pointed out that site traffic can be driven by events, so it falls and rises for reasons having nothing to do with the site's sustained popularity. So you check with proxies for Hewitt to see if they too have experienced declines. (And if you accept Alexa as valid, the liberal Web site Daily Kos has experienced a big loss in readership during the same time period.

But for that very reason, Hiltzik should have been more cautious in advancing his argument. He cited a 19.3 percent decrease in traffic over two months. That number seems absurdly precise, given how volatile the numbers are. And Hiltzik certainly blew it by not linking to Hewitt's Site Meter readings, as Armed Liberal did. It doesn't take a Pulitzer Prize winner to see that direct access to the underlying data should be provided on the Web, since it's so easy to do.

So there it stands for now. I'm waiting for Hiltzik to tell me why the data he collected is superior to what Site Meter shows for Hewitt, or how I have misinterpreted it.



Links to this post:

Create a Link



<< Home
|

This page is powered by Blogger. Isn't yours?