Overview

The accuracy of Google Analytics can be manipulated, either by accident or by intention. The bounce rate and average time on site metrics are particularly sensitive to manipulation, with one effect of reducing the web site relevancy, resulting in traffic loss. This effect is more pronounced with organic search matches, but also influences PayPerClick matches as well. Interestingly enough, the manipulation works in the other direction as well, with only a moderate increase in effort.

The Mechanism

The manipulation is very simple. To increase the bounce rate, open a web page and then quickly exit the web site. To decrease the average time, connect to a web page and then quickly move to another web page in the same web site. To go the other way, merely do the opposite.

Manipulation By Accident

This is internal traffic that usually results from testing, and usually has the effect of reducing the relevancy metric. A change is made and deployed, and the web site is eyeballed by quickly moving from one page to another, with the sequence often repeated for different browsers. FutureWare has separate development, QA and production servers, but we still look at the production site after changes are deployed, just in case (and, more often than not, a good thing we do). Some web development tools purport to have a locally operable copy, but it virtually always differs from the production copy after deployment. Caveat Emptor.

Avoiding Accidental Manipulation (Maybe)

Google Analytics provides a way, through the FilterManager, to isolate local traffic by IP address, singly or in ranges. However, this does not help for dynamically assigned IP addresses, which are often changed for each browser session by a firewall or router somewhere.  A javascript snippet can also be added to web pages that isolates activity, but this makes deployment problematic, given the change that has to be done to remove the snippet.

Manipulating By Intent

The world being what it is, and the bad guys getting badder, manipulating the metrics programmatically is amazingly easy. It certainly does not help that the browser vulnerabilities are now increasingly being exploited, perhaps helped along by their open source, with roots going back to the original Mozilla. Caveat Emptor indeed.

Tags: analytics metrics, bounce rate, relevancy, web statistics manipulation

This entry was posted on Tuesday, June 3rd, 2008 at 14:16 and is filed under Software That Works. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.