How Accurate is Alexa?

Alexa is an internet service that by using a toolbar that is installed on a large number of people’s browsers, is able to collect a lot of information on internet sites. Since the claim is this statistical sample is large enough to make statistically significant statements, Alexa theoretically can be used to accurately measure traffic of most (if not all) internet websites.

 
The lower the traffic rank score, the more popular the site is and the more traffic it gets. For example, I just checked Google’s traffic rank and it is 1. I tend to believe this is accurate and Google is the most popular website in the world. Bing is ranked at 19. Of course, it can’t be 100% accurate since not all internet users in the world have Alexa installed, but the claim is that it’s accurate. And with a large enough base of users that accurately represent the collective behaviors of internet users (this is pretty important!) it should be true (this is the same problem faced by people who conduct polls – to get a sample that represents that the population demographics).

 
I noticed that with some of my sites, Alexa seems to estimate traffic pretty accurately However, with others, it is completely and utterly wrong. For example, my most popular site – quite a niche site admittedly – yet one that has been able to get a very decent daily number of visitors and two page 1 Google rankings, is assigned an extremely bad traffic rank of ~5,000,000! This is far from accurate. The mentioned site gets at least 3-5 times as much traffic as this blog, yet has a far, far worse rank. And this blog is barely a month and a half old.

 
The answer to this question is simple: the ‘average’ user does not visit my mentioned website as much. However, isn’t the claim that Alexa’s user base is large enough to be able to give a good indicator? And if that is not the case, how can we trust its rankings for anything except for the most popular/mainstream websites, really?

 
My goal in this post has not been to analyze the pros and cons of Alexa, but more to pose an open question to readers: How has been your experience with Alexa rankings? And if it’s been accurate (in your opinion), is your site big/small, on a popular subjects/niche topics, targeted to a specific demographic/worldwide? I am just curious in trying to get a better understanding for this.