Lies, damn lies and (web) statistics

Like most webmasters, we record website statistical information to help us determine who our regular readers are and what they’re reading. As of last week, this is a summary of the ratio of browsers used to access our website:

As you can see, Internet Explorer still leads the pack but it’s popularity is definitely dwindling fast, largely thanks to the mainstream adoption of a much better alternative in Firefox.

What’s interesting from the above figures though, is that the fourth most popular browser is “NONE”. We have tried a number of open source browser detection classes and eventually settled on a modified version of the PHPSniffer class, which works very well for us. However, the seemingly large percentage of unknown browsers is attributed to the swelling inconsistency in the user agent strings supplied by the ever increasing client browsers and web crawlers currently in use.

Microsoft are unfortunately the ones to blame for a lot of this confusion because it was they who originally made the decision to spoof the user agent string by pretending their Internet Explorer was actually Netscape Navigator.

As this Wikipedia article shows, user agent strings have become a real mess over the years and it’s got to the stage where it’s virtually impossible to continue to maintain true identification as new browsers and web crawlers get added.

To make things even worse, some browsers actually let you change the user agent string to whatever you like, sometimes resulting is names like “None of your business”, “Nutscrape” and (my personal favourite) “Internet Exploder”.

So, no matter what statistics package you or your webmaster use, the unfortunate fact is that your figures will never be entirely accurate.

“There are three kinds of lies: lies, damn lies, and statistics.”

I guess Benjamin Disraeli’s famous quote applies here too.

Posted in Microsoft, PHP, Technology, Web by Ivan