Thursday, September 3, 2009
Philosophically, is an untrue compliment really a lie?
I obsess over statistics. We knew this already, right?
Last May, I decided to get a new gizmo to measure the relative popularity of the blog, a browser add-on called Sparky created by the well-respected internet traffic stats collecting company Alexa. I already have the Sitemeter stats, which are fun to look at and the source of the Flags Of Many Lands™ info, and then there's Google PageRank, which doesn't change much over time because it's on a logarithmic scale. A website at PageRank 5 is ten times more connected than a PageRank 4 site, and a 6 is ten times better than a 5 and a hundred times better than a 4.
By contrast Sparky changes a lot. It tells you how popular a website you are visiting is. For example, Google is #1, YouTube is #4, the Blogger homepage is #8, FaceBook is #3. Cnn.com is #52 and The Huffington Post is #306. This gives us confidence that software is measuring something real.
So what about this humble blog? When I started keeping track in May, the blog was ranked as about the five millionth most popular destination on the internet. Not that impressive, but there are billions of websites, so it's not that shabby, either. And clearly, there's room for improvement.
Well, improve the blog did. It climbed from a ranking of about 5,000,000 to 500,000 in the span of four months. It blew past almost all of the blogs I visit regularly in ranking.
Yow! Things are looking up!
Except... they're not, really.
Here's the visits in yellow and the page views in red from old reliable Sitemeter. There's been no significant change in popularity of this blog over the past four months. Unless about five million people just gave up having a presence on the Internets since May, this blog's popularity couldn't have jumped that much. Moreover, Sparky tells me this blog is WAY higher ranked than other blogs that have Sitemeter stats viewable by the public, and Sitemeter tells a different story from Sparky over and over again.
So, I'm going to keep the Sparky tool installed on my browser, but I'm not going to make it the stat I count on most. It's like buying a new scale and finding out you lost ten pounds since this morning. It's much more likely one of the measurements is incorrect, but human nature tells us which one of these measurements people are more likely to believe.
I'm trying not to be a sap, but the results so far of this effort are inconclusive.