Google Analytics designed for inaccuracy
We've recently been evaluating several web analytics packages for use on Cleartrip.com and took another look at Google Analytics; a product we used at the time we launched, but then dropped, because it was slowing down our site and diminishing the user experience.
The product has certainly come a long way from the time we first used it, but we're still not sure that it's the right way to go for us. It's undoubtedly the best "free" hosted web analytics service around right now, but it's got some glaring inaccuracies built right into its reporting.
There isn't a single piece of web analytics software that is 100% accurate, there's just too many nuances and variables to contend with. So, one can expect a certain level of inaccuracy in the data and reports from all of the available options, but Google seems to have specifically designed their product for increased inaccuracy.
The generally accepted standard for a "visit" to count as a visit is that the visitor must visit more than one page on your site. And if they leave after viewing just one page, it's not a "visit," it's a "bounce."
Why is this important? There's a massive number of key metrics that depend on your visits count--conversion rates, total visits, cost-per-visit, dropoffs etc. and if the manner in which visits are reported is inaccurate, then you can count on the rest of your reports being thrown off course as well.
Understanding the inaccuracy involves a little bit of Google history. Analytics was originally created by a company called Urchin that Google acquired. At the time of the acquisition, Urchin incorrectly counted bounces as 0-duration visits. Then, in July 2007, Google launched a new version of Analytics that corrected the manner in which visits and bounces were counted. One month later, they went back to the wrong way of doing it. Brett Crosby, Senior Manager at Google, explained that the new (correct) reports were out of sync with the old (incorrect) reports and users were complaining.
Let's go over that again--users complained because they wanted consistent reports with inaccurate data instead of inconsistent reports with accurate data. And Google just bent over?
Right, I guess we'll just continue evaluating other analytics options.
Hrush
11 Comments
View Printer Friendly Version
Email Article to Friend
Reader Comments (11)
"And if they leave after viewing just one page, the visit is marked as a bounce." Hence, bounce rate will never be more than 100%.
Ref: http://www.google.com/support/analytics/bin/answer.py?hl=en&answer=81986
Ref: http://www.google.com/support/analytics/bin/answer.py?hl=en&answer=57164
On a different note, I have been using Google Analytics for a while now, and find it to be quite accurate. Speed isn't that bad as well. If speed is really a concern, try Clicktracks for offline analytics (it is a great tool).
Analytics data is very different from themes, but Google may be easening the transition to the correct data somehow?
'Do no evil,' right?
N = 100n/(100-b)
where N= total ppl visiting
n=visits according to google nomenclature
b=bounce rate
Dhruv
"And if they leave after viewing just one page, the visit is marked as a bounce." Hence, bounce rate will never be more than 100%.
Ref: http://www.google.com/support/analytics/bin/answer.py?hl=en&answer=81986
Ref: http://www.google.com/support/analytics/bin/answer.py?hl=en&answer=57164
On a different note, I have been using Google Analytics for a while now, and find it to be quite accurate. Speed isn't that bad as well. If speed is really a concern, try Clicktracks for offline analytics (it is a great tool).
Analytics data is very different from themes, but Google may be easening the transition to the correct data somehow?
'Do no evil,' right?
We can handle some level of inaccuracy, but the GA method of counting visits is throwing too many of our key metrics.
BTW, I just tried to send a few articles from the blog using Email Article to Friend. The sender in the email which actually goes out is displayed as "Squarespace Services" which looks pretty odd. Should either be the "Your Name" variable or "Cleartrip Blog".
Completely agree with you, GA has weird tendencies and it's unclear sometimes what actually means what and how they derive things.
We actually started to use something called mixpanel to track user-behavior stuff for our startup: http://www.mixpanel.com, you should take a look, might help cleartip.