Tom Cotton’s 8109

Seasonally adjusted, however, it’s 92%. With web analytics - it’s even higher.

Currently I’m investigating the reason why there are discrepancies between AWStats and Google Analytics - specifically the large differences between a common statistic Page Views.

AWStats bases it’s reports on what happens on the webserver. The challenge is to correctly interpret server hits into behaviour.

Google Analytics bases it’s reports on what happens on a web page. It’s challenge is data completeness. The permutations of browser, connection, computer, operating system and user settings can undermine validity of statistics.

The big conclusion is web analytics is not an accounting exercise but a spot-the-trend exercise dependent on many assumptions.

My challenge is to recognise and account for valid and invalid assumptions:

AWStats might be over reporting

• What is AWStats reporting as a page load? See http://awstats.sourceforge.net/docs/awstats_glossary.html which details a OnlyFiles conifguration list
• Are spambots and screen scrapers creating false page views in AWStats? (check individual pages statistics)
• Site administration still generate visits whereas Google Analytics does not.

Google Analytics might be under reporting because:

• Site blocking (eg visitors behind education institution firewalls
• Page loads interrupted causing the script not to be activated
• Firefox ‘NoScript‘ add-on which can block the analytics script
• Web browser cookies turned off
• Web browser javascript turned off
• A dynamic webpage built using AJAX undermines the assumption of what a web page / visit actually is. For example the new Hotmail/Live service does not require separate page loads to check different mail folders to view webmail.

Sources:

• http://web.uwaterloo.ca/documentation/AnalyticsvLogReader.php
• http://aonach.com/chatter/lies-damned-lies-and-web-statistics/
• http://awstats.sourceforge.net/docs/awstats_glossary.html

Do you have any further suggestions or corrections?