Survey methods and implications

Mary Morris (marym@Finesse.COM)
Fri, 16 Dec 1994 11:57:08 -0800

> I tried several times to complete it, but finally gave up after
> completing about 3/4ths of it. I don't know if I was counted as one
> of the 3,299 respondents (someone else said 18,000 - where'd that
> come from?), but if so, my data was incomplete.

There were about 3500 respondents. There were several sections to
the survey. If you responded to 3 sections, you were counted as
3 responses. Since most of the people had experience in most of the
7 (I think) areas, 7 * 2500 =~ 18,000.

I didn't experience problems with my browser which was also 2.4. I'd
love to hear if other people had problems. The technical issues here
aren't necessarily limited to surveys. If you had problems with a form
that long, you may have shopping problems as well.

> It may well be impossible to create a Web-based a survey that is not wildly
> skewed. Perhaps other mechanisms need to be employed (phone, mail, email)
> to generate less biased results.

I think that at this point, it is impossible to create an unslanted web
based survey about the web, but your hit rate would be so low if
you did phone or mail surveys that they wouldn't be any more significant.

Surveys via email suffer from the same "junk mail" status as advertisments
to many people, so they wouldn't be the ideal alternative either. The
problem is that I believe that this survey was posted in several netnews
groups (related to the scope of the group of course) which is why the
survey is skewed. The Internet is a good place to get feedback on a
specific topic, but not to query a spectrum of people with differing
levels of interest/involvement.

This is the same issue that comes from getting people to visit your
store. You don't get the general public, you only get "interested"
parties. This is good for pre-qualifying things, but it is lousy
for generating new awareness.

Mary