Bracing for Impact

[Jeff Spaleta]

On Tue, Jun 2, 2009 at 6:21 AM, Paul W. Frields <stickster at gmail.com> wrote:
> I'd suggest you get with the Infrastructure guys and find out how we
> could cull referral data in a way that would be helpful.  Perhaps that
> data could tell us how effective various types of marketing work are,
> so we could concentrate on doing the most effective things for
> spreading news to *outside* the Fedora Project, and bringing those
> visitors in.


First of all, the alternative explanation for the spike is that people
showed up on the original schedule release day. in a massive wave.  It
could be we had nearly 2 million really disappointed people all
showing up on the originally scheduled release day.  If everyone is
prepared for possibly seeing that publicly stated I can examine daily
counts for the month of May and post a graph.

Second, in order to seriously track effectiveness I need to to have
some sort of implicit measure of aggregate marketing "output" that I
could trend along side with hit counts.  As aggregate marketting
output goes up..do counts go up?  I dont have any suggestions for a
trendable marketting output metric.

Or, if we want to look more fine grained than that, and look at the
effectiveness of individual marketting efforts I would need to be able
to pinpoint those efforts in time and maybe location and see if they
had a localized impact.  Localized meaning...in the scope of the
audience that saw the marketting. For example, should we expect to see
an increase in traffic from a certain region or country where Fedora
has a strong marketting presence at a physical event?  I could try to
parse the logs looking for that sort of uptick.  Another example, when
we have specific digital marketting like the focus on feature
interviews..can see how that drives traffic via refurls? Can we get
agreements from technical laypress sites about always including
linkages to a particular project page?

I'm pretty sure the pythonic log parser i hacked together can deal
with parsing referral urls I just haven't had a reason to do it yet.
If I had a list of specific urls I could probably give you some useful
relative stats about them.

-jef