http://joelnagy.com/blog/2009/06/tracking-invalid-hits-accuracy-is-quite-important/ web technology is a way of life Wed, 08 Sep 2010 03:09:45 +0000 http://wordpress.org/?v=2.7.1 hourly 1 http://joelnagy.com/blog/2009/06/tracking-invalid-hits-accuracy-is-quite-important/comment-page-1/#comment-163 Sinan Ascioglu Fri, 17 Jul 2009 20:52:50 +0000 http://joelnagy.com/blog/?p=140#comment-163 Well, I had been spending a lot of hours in my last 3 years reading analytics, and I think that wouldn't be the biggest problem in tracking user stats. Besides if pages get being tracked accurately or not, I think the big problem that people couldn't still solve is how to provide that data. We know that, since tracking == raw data. It doesn't help much, so everyone wrote their own algorithms to analyze data and come up with terms like 'unique visitor', 'pageview vs. unique pageviews', 'time spend on site', etc.. I think the questionability of these algorithms are far more of an issue than if pages being tracked accurately or not. Take 'time spend on site' for example; every time I meet an analytics professional, I ask how it gets tracked, and every time the answer I get is different, and none of them can correctly define the 'time' spend on site, since no website starts a counter watch to count that time. Eg. Google analytics counts this as (last page hit time - first page hit time) in the exact same session. This is especially a bad calculation method for sites like NYTimes.com, since one would expect to find an article after 3 clicks, read the article for 10 minutes, and close the site. Because the calculations went very complicated, they don't show unique pageviews and visit for the same page for example... And yet, it become even harder to track since we started to track event calls, ajax calls, etc.. so, is an event call another hit for a visit? should I include that in a "we broke the record of our monthly visitors!" report... So since giving thought on these, I started looking at the data more referencing to itself than absolute number. Those number are good for comparing (with last month, with this section, etc..) rather than seeing them absolute numbers.. My 2 cents.. Well,
I had been spending a lot of hours in my last 3 years reading analytics, and I think that wouldn’t be the biggest problem in tracking user stats. Besides if pages get being tracked accurately or not, I think the big problem that people couldn’t still solve is how to provide that data. We know that, since tracking == raw data. It doesn’t help much, so everyone wrote their own algorithms to analyze data and come up with terms like ‘unique visitor’, ‘pageview vs. unique pageviews’, ‘time spend on site’, etc.. I think the questionability of these algorithms are far more of an issue than if pages being tracked accurately or not. Take ‘time spend on site’ for example; every time I meet an analytics professional, I ask how it gets tracked, and every time the answer I get is different, and none of them can correctly define the ‘time’ spend on site, since no website starts a counter watch to count that time. Eg. Google analytics counts this as (last page hit time - first page hit time) in the exact same session. This is especially a bad calculation method for sites like NYTimes.com, since one would expect to find an article after 3 clicks, read the article for 10 minutes, and close the site. Because the calculations went very complicated, they don’t show unique pageviews and visit for the same page for example…
And yet, it become even harder to track since we started to track event calls, ajax calls, etc.. so, is an event call another hit for a visit? should I include that in a “we broke the record of our monthly visitors!” report…
So since giving thought on these, I started looking at the data more referencing to itself than absolute number. Those number are good for comparing (with last month, with this section, etc..) rather than seeing them absolute numbers..
My 2 cents..

]]>