Ubuntu Popularity

[Eric Dunbar]

Really, this is sounder material!

On 12/22/05, Mike Bird <mgb-ubuntu at yosemite.net> wrote:
> On Sat, 2005-12-17 at 22:04, Eric Dunbar wrote:
> > I would suggest you stop and take a look at the following (crude)
> > break down on page views for the past three months at DistroWatch
> > <http://www.distrowatch.com/index.php?dataspan=13>:
> > Rank  Distribution    Hits Per Day
> > 1     Ubuntu  2739
> > 2     SUSE    1891
> > 3     Mandriva        1801
> > 4     Fedora  1030
>
> Those figures are certainly interesting.  However, they count
> views of the page describing Ubuntu, not people actually using
> Ubuntu.  Most of the people viewing that page are people who
> have heard about Ubuntu and are considering switching but are
> not actually running Ubuntu yet.

I believe I did not state usage, now did I ;-)?

> It's already been noted that Ubuntu was invisible in Netcraft's
> server rankings, but that measure is subject to the criticism
> that Ubuntu has not yet put out a proper server release.

Plus, popular perception is that Ubuntu is not targeted at a server
release so it wouldn't be at all accurate to extrapolate from server
to desktop.

> I therefore analysed the last million hits at a national-
> -brand website with largely teen and young adult demographics.
> (To avoid bias, I first filtered out the accesses from our
> office and customer systems - about 0.3% of the total hits.)
>
>   PERCENTAGE OF ALL HITS
>   Linux                0.13%
>   Macintosh            3.08%
>   SunOS                0.01%
>   Other               96.78%

How come no breakdown by Other? What % of Macintosh and Linux agents
report as Windows agents?

>   PERCENTAGE OF LINUX HITS
>   Debian                3.7%
>   Fedora                0.1%
>   Mandrake              0.0%
>   Mandriva              2.2%
>   Redhat                0.0%
>   SUSE                  0.2%
>   Ubuntu                5.7%
>   Anonymous            88.4%

Are you lumping Ubuntu and Kubuntu together here?

> It turns out that 88.4% of user agent hits which included the
> Linux string did not include a recognisable distro.  Although
> Ubuntu and Kubuntu include the distro name in their Firefox
> and Konqueror user-agent strings this is rare.
>
> I also analysed the 50.4% of Linux hits which were from either
> Firefox or Konqueror:
>
>   PERCENTAGE OF FIREFOX OR KONQUEROR LINUX HITS
>   Kubuntu/Konqueror     5.9%
>   Mandriva/Firefox      4.3%
>   SUSE/Firefox          0.3%
>   Ubuntu/Firefox        0.8%
>   Anonymous/Konqueror  14.4%
>   Anonymous/Firefox    75.0%

Seems a little suspicious that Kubuntu (arguably more crude than
Ubuntu) would out-number Ubuntu 7 to 1 (can you draw that
conclusion... do the user agents self-identify as Kubuntu and
Ubuntu?).

> >From this unbiased but limited data, it appears that Kubuntu
> is doing well in the KDE world but that Ubuntu is lagging.

You may consider yourself unbiased (and, I don't doubt you analysed
them in as fair a manner as possible), however, it is questionable
whether the underlying data was representative.

Your demographic was reportedly "young", can you be sure that all user
agents are recorded properly, and, that the user agents self-report
accurately (as you yourself partially recognise with the "Other" list
above... unless you're avoiding mentioning Windows on purpose, with I
don't think)?

I question your conclusion about the popularity of Kubuntu vs. Ubuntu.
In the identified Linux visits, you have a 20.3% to 75% Konq vs. FF
ratio. However, you report 5.9% Konq/Kubuntu vs 0.8% FF/Ubuntu.

Anyway, there are too many questions still about your analysis. The
one thing I do find informative is that Ubuntu is the most common
_identified_ Linux!

> As always, better statistics would be welcome if anyone has
> them.

Absolutely! A wider selection of web sites would be helpful, but, what
would be more interesting would be to have a non self-selecting survey
of what computer users run (given that browsers mask their identity,
estimates of platform usage can never be more than crude guesses (more
so than normal surveys)).

Your quick-n-dirty analysis raises more questions than it answers, but
that is a good thing.

Eric.