[Design] SotM11: BoF meeting

[Michal Migurski]

On Sep 13, 2011, at 4:03 PM, Tom Hughes wrote:

> On 13/09/11 22:05, Gregory wrote:
> 
>>   The stats that we have are here:
>> 
>>   http://stats.openstreetmap.__org/cgi-bin/awstats.pl?config=__www.openstreetmap.org
>>   <http://stats.openstreetmap.org/cgi-bin/awstats.pl?config=www.openstreetmap.org>
>> 
>>   Not sure that they help much though.
>> 
>> Yep, I didn't hear all the statistics people were talking about that
>> they wanted. Steve said something about Google Analytics wouldn't be
>> possible because it wasn't free/open enough and their were problems with
>> that for some reason.
> 
> Well there are things - piwik would be one example - which are basically open source versions of google analytics. So we could set something like that up.


Tom, what do you think about something a bit more ghetto? Most analytics packages are front-ends to big, dumb log files (basically) and the kinds of metrics I'm imagining might not be captured by a standard collection of server metrics.

The dream log file I have in mind would have five columns: tracking cookie, timestamp, OSM username (if any), URL path, and a message with some detail like "searched map", and it would be written by some ajaxy-javascript thing that would watch people's actions on the site, seeing things invisible to HTTP logs.

A few metrics I've got in mind that can help us evaluate whether a site design helped after the fact are:
	- time on site, from page to page (more is better)
	- frequency and number of repeat visits (more is better)
	- follow-on signups (more is better)
	- interactions with the slippy map (where do they pan/zoom to?)
	- search terms typed into the box (what kinds of things do they look for?)

Tom, would it be interesting to you to help instrument the site in this way, in a hacked-up way at first?

-mike.

----------------------------------------------------------------
michal migurski- mike at stamen.com
                415.558.1610