What is success for Fedora?

[Christian Schaller]

----- Original Message -----
> From: "Stephen John Smoogen" <smooge at gmail.com>
> To: "Fedora community advisory board" <board-discuss at lists.fedoraproject.org>
> Sent: Wednesday, July 2, 2014 8:39:10 PM
> Subject: Re: What is success for Fedora?
> 
> 
> 
> 
> On 2 July 2014 03:01, Christian Schaller < cschalle at redhat.com > wrote:
> 
> 
> To me any definition of success needs to be tied to popular adoption of the
> products we make, any measure of success that doesn't take that into account
> becomes to me a version of congratulating yourself for having achieved
> freedom of speech by putting yourself in a situation where there is
> nobody around to listen to what you have to say.
> 
> So if we for instance define long term success as having a 50%+ marketshare
> among our target audiences, then I think a per release success criteria would
> be that each product release sees a significant userbase bump. how we define
> significant and how we measure the growth is of course another question, but
> I am sure that if we agree that this is a good way to do it then I am sure we
> can find some indicators to use for measurement and decide on ambitious but
> realistic goal for each release.
> 
> 
> My problem with this definition is that what do you 'git blame' for the
> reason that user levels either fall or don't grow. I mean its easy to blame
> whatever big change happened in that release.. but its a knee-jerk reaction.
> It has as much as basis as the person who made the change defending it. [We
> have had big dropoffs of users in the past which could be linked to us
> dropping Xen, moving to GNOME3, beefy miracle offended vegeterians, aliens
> invaded new york, etc but is it really that? Did other distributions see an
> increase in their usage or did they see drops at the same time? How do we
> get a reasonable guess and how do we fix it? [Uninvade New York? Add Xen
> back? etc]
> 
> Also when do you measure usage of a release and say 'yay we met our growth
> metric'. Because what happens is that a release doesn't go from 19 -> 20 on
> the day of the release. N usually gets to the number of users of N-1 just as
> N+1 is released. Our overall growth is in people keep using old old releases
> (we have a lot of Fedora 12, 14, and 17 users which don't seem to shrink
> which means our 'usage' curve goes up but not in the latest release).
> 
> Finally, how do we measure what the size of our target audience is? It is
> just as hard to count as counting who our 'contributors' and 'users' are. Is
> it a subset of the 1.9% of current non-MacOSx/Windows users? Is it the
> MacOSx/Windows market? We say everyone and we can basically fold up shop now
> because that doesn't look feasible, we say 20,000 people we can say have
> "Mission Accomplished" and go home.
> 
> We will hopefully be better able to measure usage of
> workstation/server/cloud/everything better in the next release if each
> release will give some indication to yum or dnf that the person is updating
> from that 'platform'.

Well I don't think we will ever find a perfect and 100% accurate measure,
so what we would need to do instead is pick a set of indicators to watch
like a combination of overall tracked Fedora product downloads during a release series, 
pick a few conferences to poll the attendees about operating system usage (JBoss 
and OpenStack conferences could possibly be good choices while GUADEC, Akademy 
or Flock would likely be bad choices :)) and Google trends for example. So none 
of them would give us a 100% accurate answer, but combined we should be able to 
at least detect trends from them.

Of course the exact indicator composition would wary a bit between the 3 products.

Christian