If asked to define ‘Wikipedia mobile traffic’ you might answer ‘Wikipedia page requests originating from mobile devices’ or perhaps ‘page requests serviced by Wikipedia’s mobile site’. You might think both definitions are more or less identical. This is not so. At least one fourth of the requests from mobile devices is serviced by our non-mobile (main) site. Vice versa part of the traffic to our mobile site may originate from desktop computers.
The red line in the chart above shows the percentage of Wikipedia page requests that were issued from mobile devices (tablet, smartphone). We detect these requests by scanning for keywords in our traffic log (in the ‘agent string’), like iPhone, android, Palm.
Users of mobile devices may opt to access the main site, either for one page, or for all time. These options are shown on every mobile page. Maybe some traffic is never redirected to our mobile site anyway.
The blue line in the chart shows the percentage of Wikipedia page requests that were serviced by our mobile site. Did these requests all originate from mobile devices? Not necessarily. A user with a slow internet connection could follow this path in order to shorten page load time. The difference in page size is significant.
Here is a breakdown of data sent for the article about Rembrandt on the English Wikipedia.
Non-mobile: http://en.wikipedia.org/wiki/Rembrandt (939K) on the left.
Mobile: http://en.m.wikipedia.org/wiki/Rembrandt (472K) on the right.
The non-mobile page is almost twice as large.
Data collected with YSlow.
We should be able to check if desktop usage really contributes significantly to the fast growth of our mobile traffic. We just haven’t done it yet.
13 Aug update: a follow-up blog post deals with a comparison of above metrics, and how this effects the share of requests which come from mobile devices.