Lies, damn lies, and Google Analytics
Aug 29th, 2008 by handolio
There are probably two maths teachers out there somewhere, and perhaps an A-Level examiner, who can attest to a past lack of interest in statistics on my part. But only 95.7% of statistics are bad.
The good ones, I am reminded, live in Google Analytics and tell you fascinating things about how few people find your blog interesting, and what it was that the other people were looking for when they visited it by mistake.
I’ve noticed one particularly amusing trend. There are people out there, according to Analytics, who are searching for very specific things. Let’s take the delightfully worded how to turn off the fucking autoformat of word 2007. I’m reassured that at the time of writing we’re topping Google for that particular search.
As the box above suggests, quite a few people are finding Hackbash through a search for how to disable Word’s annoying formatting foibles. What’s odd is that Analytics says they’re all leaving immediately without reading the post they land on - which tells them exactly how to do that.
Hmmm. The post opens with a lot of waffle, so I’ve just updated it to let people skip to the relevant bit. But looking through the terms people find us by, there’s no end to the very specific searches for which we’ve got very specific posts that people seemingly aren’t reading: define piecemail, iconic bt home hub (and iconic means) and mnemonic plague, for example.
No doubt most first-time visitors to Hackbash don’t find what they’re looking for. Perhaps a proportion are slumping inert on their keyboards as I embark on another tirade about BBC News. But maybe Google Analytics lies, too.
This is really interesting - I’d love someone who knows about this stuff to comment. In previous jobs when I worked with a webmaster to look at site traffic using other web stat packages (in the olden days before Google Analytics) I was invariably told by them that stats about time on site were completely spurious and unreliable because the way that user sessions were cached and measured in different situations varied so wildly. It’s only since encountering Google Analytics and working at iCrossing that I’ve seen so much attention paid to this data. Obviously, if it’s accurate it’s incredibly useful, but I’ve yet to have a good explanation as to why Google Analytics is able to measure this any more effectively than previous web stats packages.
If anyone has any insight on this I’d be really interested to find out more.
In general, I believe, Analytics tools can only measure time on a particular page, if that visitor then goes on to visit another page on the site - as then there is a start and an end point to measure. This is why you have zero seconds for time on these pages.
Which means these people probably read the article and you helped them defeat Word. Feel good about that.