Social Network For Analytic Professionals

How does a Alexa or Quantcast monitor web sites?

Looking at the Quantcast top 100 provided few surprises as to who was on it. I guess we take for granted that a search engine is on the top of the list, but I didn't know 140 Million people go to it each month. That's about 10 million hits a day. I was surprised that the NIH was on the top 100 list and that it's slightly more popular than FoxNews this month. So the ordering of the webs sites is interesting. What gets me is where its says Pandora is skewed toward African Americans or MTV towards women or that Amazon appeals towards educated people. There's no reason a priori these should be true.

How do sites like Quancast measuring not just the number of people who visit the web site, but also the gender, the age, income bracket, frequency of visits and correlated web sites? It's kind of neat, actually. I've also heard cookies can be analyzed.

▶ Reply to This

Replies to This Discussion

Permalink Reply by Amanda Rose on August 11, 2009 at 1:07pm: With the advent of Broadband and DSL, the user can be traced back the actual physical addres from whence they are connecting to the web. I haven't seen any weird append between static IP address and the demograhics available by physcal address.

I bet a trace poplation based on volunteered demographics, say at a survey site. When they can get a cookie, and login information then you can get demographically segmented tracking. Alexa is known to be phenominally off. I haven't tested Quantcast. Anyoe know how good it is?; ▶ Reply to This

Permalink Reply by Vincent Granville on August 11, 2009 at 3:57pm: Alexa monitors traffic from users who downloaded the Alexa toolbar. The data is biased, because Internet users who download toolbars are different from those who do not.

Quantcast has partnerships with large ISP's (Comcast, etc.) and access to the ISP logfiles, as well as client information gathered by the ISP's and from other external databases (demographic data by zip code). That's how their can tell the gender, education level and income of web site visitors.

In both cases, the reported traffic statistics have an additional significant bias: traffic from sophisticated anonymous robots is not filtered out. This traffic introduces a bias in number of users, number of page views, pages per user per month and visit duration, and even in the top keywords associated with a particular domain.; ▶ Reply to This