Wednesday, February 08, 2006
does pixel based tracking work?
If you don't know already, pixel tracking refers to putting 1x1 sized IMG calls on other servers that record stats (such as hits or sales) to your web application. Lots of companies use them and I've been curious for a while how reliable they are and how often they “miss.”
Here is the experiment I've been conducting:
When a user hits a certain page three things happen:
The server adds 1 to a database count.
A Pixel calling the same domain name is loaded which, if called, adds 1 to its database count.
A Pixel calling a different domain name is loaded which, if called, adds 1 to its database count.
As you can see, if all is working then all three counts should match. They don't. Instead, with a little over 5000 visits counted, I've found that the same domain pixel is off by 11.54%. The different domain pixel is off by 5.6%. This is much worse than I expected.
I've attempted to do some research on the topic but haven't found much. Does anyone know what factors might cause pixels to not fire? Are there any good articles on the topic?