Josh Pigford, the creator of The Apple Blog, has posted an extremely well-written rant of sorts regarding the major flaw with Alexa website traffic statistics. Alexa is a serviced owned by Amazon that provides website traffic statistics based on information collected from web surfers with the Alexa toolbar installed. The problem is that the Alexa toolbar only runs on Internet Explorer on Windows machines, effectively voiding every single alternative browser and Mac user.
To give you an idea of how little the Alexa toolbar is used, IE6+/Windows users must physically go out onto the interweb, download and install the toolbar themselves. This is not some "oh it came bundled with Windows, how cool" deal. In addition, many pieces of spyware prevention software actually flag Alexa as spyware, as it tracks your browsing habits and reports back to the mothership (no not that mothership.. or that one). As a side note, people have tried to spread a Firefox extension that hooks up to Alexa, but Firefox users are all too smart for that ruse.
Why is this important? As Josh noted, many businesses are utilizing Alexa stats for things like advertisement services. This ends up hurting webmasters that own sites with a large chunk of Firefox, *nix, OS X, etcetera users.
Businesses like Text Link Ads and ReviewMe place a very large amount of weight on Alexa rankings. Obviously they need some way to rank sites, but the problem here is that the tool they are using is largely skewed towards Windows IE users when not every category of site has that demographic as the large majority.
This whole talk about Alexa started out when I posted something in the 9rules member forums regarding blogger John Chow's Alexa traffic stats compared to mine. John mentioned in a recent post that he received 266,641 page views from 127,614 unique visitors during the month of February. For February, I received 227,852 page views from 136,064 unique visitors. Does the Alexa graph below show that?
Alexa shows some wildly inaccurate information, but to Alexa's knowledge that is correct. John Chow's readers seem to be heavy PC users as John runs a PC-centric tech hardware site as well. It can only be assumed that he has more Windows and IE readers than I, a Steve Jobs-idolizing college student, do. Since Alexa bases everything off of data from IE users toting their hated toolbar, this chart leaves most of my readership out in the proverbial, unmonitored cold.
Moral of the story - take Alexa stats with a grain of salt if even that. Josh Pigford pointed out that there are alternatives to Alexa (Compete, comScore, Hitwise), but the problem is getting companies already utilizing Alexa information to stop and change what they're doing.
Before I go, I'll leave you with a Grade A rant on the subject written by the knowledgeable-in-all-things-random Kyle Neath:
We have far too tight a resolution of website statistics. As owner of a server, we can analyze down to the request. Unfortunately there's a lot of noise out there (spam bots, search engine spiders, people typing in the wrong domain, etc). However, we still feel we have a good grasp on how many "people" visit our site, even if that definition of a person is based on an IP address.
Realize that no other media has this. Magazines come next in terms of resolution, but even they can't count the number of companies that subscribe that then pass the magazine on to 50 other people.
Media like Television can't even try to grasp this resolution. How many people watched LOST last night? Well.. we don't have a hit for every person that watched it. We can't even count how many people downloaded it via bittorrent. Suddenly the game gets more complex. The reality is that we only have good resolution for very high profile TV -- The top 100 shows or so. Trying to measure how many people watched Sci-Fi's B-movie of the month is extremely difficult.
This is where we come back to Alexa. To me Alexa is a lot like measuring TV stats. We're taking the best possible source data we have, and trying to apply it to Websites. This data when compared to TV is excellent. We can get a really good handle on the top 100 sites -- hell, even top 1000, maybe more. But then people apply this Alexa information to sites in the 10,000's. Whoa... we're starting to introduce a LOT of assumptions, and a lot less good data. Extrapolations are frequent, and data can get skewed easily.
Now people take that alexa-view of their rank 20,000 site and compare it to their fine-grain stats by their webserver. Well, of course it's going to be off. It's not even comparable. When you apply that fine-grain resolution view to the extrapolated-view of alexa, it's going to look extremely inaccurate -- and it is.
Really, this whole web site "statistics" game is just a bunch of guessing and assumptions. Remember that it all starts off with the assumption of what a visitor is. If I changed my assumption of a visitor to a unique IP per 10 hours, I could more than double my traffic. Think about it.
As a webmaster, domain owner, blogger, etc., what's your stance on this issue? I believe that if companies can't be easily cajoled to stop using Alexa statistics, Alexa should at least hire someone to fix their problems.