[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Fwd: Re: Linux end-user survey

From: Michael Eilers <eilers@rmi.de>
Date: Mon, 15 Jun 1998 00:26:43 +0200
To: Roger Dingledine <arma@MIT.EDU>
Subject: Re: Linux end-user survey

Roger Dingledine wrote:

> In message <3582748C.B54261F5@rmi.de>, eilers@rmi.de writes:
> >Don't expect too much from a questionnaire. Most end-users don't know what an
> >os
> >really is able to do. If you ask about their wishes they will answer: Give us
> >an
> >os that solve all our problems immediatly and complete.One word about your time
> It's true. I'm not expecting it to change the world, but I can always hope. :)
> I think that if we get anything at all out of it, then we will have gained
> something.
> >schedule. It is very hard to reach joe average user via internet. When we
> True again -- I think I can get many people to fill it out, especially if I
> put it in the right places on slashdot, lwn, linux-announce, freshmeat, etc.
> But if it takes us six months and we end up with something useful afterwards,
> that's good enough as well.
> >started the Linux questionnaire we announced it on several Linux mailing-lists
> >and newsgroups. During the first weeks the percentage of experienced (>1 year)
> >Linux users who participate where near to 90%. After the questionnaire was
> >mentioned in several places on the web (Linux gazette, Linux weekly, LDP) the
> >percentage drops to a more realistic 78%. We think that the "real" percentage
> >of
> >Linux users with >1 year experience should be around 70-75%. So I think it is
> >nearly impossible to gain good results in only one week.
> I don't know much about the demographics of linux users (yet). What led you
> to pick 70-75%?

Main hint to assume this is the development of subscription of Linux Journal

            1995: 10000
            1996: 25000 (factor 2.5)
            1997: 45000 (factor 1.8)
            1998:  ~63000 (factor 1.3-1.4) (estimated, with linear development
                                                                           of the
factor assumed)

If you assume that the growth of the number of subscriptions of Linux Journal is
closely correlated with the growth of the number of Linux users, then this number
is increased by
30-40%  during the last year. That means that ~(71-76)% use Linux longer than one
year .
(of course only a rough estimation)

> >> Ideally, I'd like to get a more quantitative approach to the survey, rather
> >> than the qualitative approach that you have used. (Also, people taking a
> >> survey would much rather click one of the the available options than
> >> actually write sentences, in my experience.)
> >
> >That's true. Here a few 0.02$ that we concluded from our experiences      -
> >short and concise questions (people tend to read web forms much
> >          less accurate than printed ones)
> Yes, this is a good idea. Indeed, the quantitative approach lends itself
> to short concise questions because the answers have to be simple.
> >      - no interdependence between questions. (Such kind of questionnaire need
> >s
> >         a  skilled interviewer)
> I think this will be more meaningful to me after we've tried to come up with
> the first draft. I suspect we will tend to have inter-connected questions at
> first. This is a good thing to keep in mind.
> >      - make the results online (best without delay).
> >            (That gives the participants a good feedback)
> This can be done with some parsing behind the cgi. Shouldn't be too
> difficult. (yay perl.)
> >Till now the number of filled forms of the Linux questionnaire has reached the
> >number 1000. As we only publish most actual 100 answers for each text question
> > I
> >can offer you to make a tarball with the complete stuff. After browsing through
> >this you may have a
> >good feeling of what Linux users sucks or pleases. Maybe this is helpful for
> >your project.
> This looks like it would be extremely useful.

look at ftp://aachen.heimat.de/pub/alug/lin_quest_ans.tar.gz

> Thanks!
> --Roger

good luck,