[Foundation-l] clicktracking

[Nimish Gautam]

We developed clicktracking to see what buttons were being used on our 
toolbar.  It was enabled for all Wikimedia projects that had the 
usability toolbar and only gave us information of the type that "a user 
who has an edit count of 3 just clicked the 'bold' button". Collecting 
such data for internal research purposes is fully consistent with the 
letter and intent of the Wikimedia Foundation privacy policy: whenever 
we do so, we strive to minimize the amount of information collected to 
what's strictly necessary.

(Read 
http://techblog.wikimedia.org/2009/10/click-tracking-on-edit-toolbar-deployed/ 
for more details, or see the code in SVN at http://bit.ly/c7hg0J )  .

We wanted to know, roughly, who was a novice and who was an expert, and 
we had the edit count metric for that. We also wanted to know who is 
active and who isn't, so for that we created a metric of edit count vs 
timeframe. Specifically, when someone clicks, we record what event 
happened (i.e., they clicked 'bold', 'left navigation-logo' etc), what 
their edit count at the time was, what their edit count over the last 1, 
3, and 6 months was, and a randomly generated session ID that lasts 
until page refresh. This was done intentionally to keep the data anonymous.

We turned this feature on for left-navigation at a sampling rate of 
about 1 in 1000 to get an idea of what was being clicked and what 
wasn't, so that we'd know what to collapse and what not to.

We also used click-tracking on old editing toolbar to make sure that we 
didn't leave any buttons off that people really liked.

The Special:ClickTracking page referenced here:
http://en.wikipedia.org/wiki/Wikipedia_talk:Arbitration_Committee/Audit_Subcommittee/Archive_1#New_privacy_issue.3F 
was a prototype for visualizing the aforementioned data. It was never 
actually "turned on" to begin with, and it wasn't our highest priority 
to get it up and running, especially since *the visualization's* 
database operations aren't exactly lightning-fast and haven't been 
optimized yet. It also wasn't and won't ever be a wikipedia 
administrator tool.