NONE: Net and Web Use Baselines and Growth
Net and Web Use Baselines and Growth
Donna Hoffman (hoffman_at_colette.ogsm.Vanderbilt.Edu)
Sat, 20 Jul 1996 14:10:25 -0500 (CDT)
xerxes writes with some incomprehensible criticism of our papers:
> The Commercenet study, which was in the field last summer or fall, was the
> approach that measured users "with access to the Internet"....as opposed to
> actual users....the figure they came out with was 22 million, if memory
> serves, and was widely criticized as being wildly overstated. In fact, I
> recall Donna Hoffman making these criticisms herself. It seems the work
> had been funded by a hardware industry association with an eye toward
> exaggeration.
The study used a variety of measures of use. One was based on potential
access, but the others were based on actual use. The inflation was due to
inadequate weighting and logical inconsistencies in the measures, not, I
believe, a deliberate intent to inflate.
>
> Here we have a notice of a reanalysis (?) of data approaching a year in
> age. There are studies on the street right now, generalizable analysis,
> not web-based self-selected samples, based on data collected this spring,
> or later. I go for 1996 data at this point, since growth this year is
> slightly different from last.
The CommerceNet/Nielsen study is a population projectable study, not a
Web-based self-selected sample. Our corrections make it representative.
There is no argument that the data are now nearly a year old, but
calculating accurate baselines are important to track growth and evolution
of segments. We are now awaiting the data from the six-month followup and
once we get that, we will be able to project overall and segment growth.
You might also be interested to know that our revised estimates jive nicely
with the O'Reilly and Find/SVP estimates AND the just released Intelliquest
estimates - which use our categories and weighting variables for
comparability.
If we're going to have a legitimate discussion of Internet size and growth
estimates, then we need to be clear on the distinctions between online
samples which are not projectable and random samples, which may be.
The GVU studies are self-selected but good descriptions of the Web
population that responds. O'Reilly, Find/SVP, CommerceNet/Nielsen,
Intelliquest are RDD approaches and potentially generalizable. The
Yankelovich Cybercitizens Panel also appears to be projectable to certain
segments of the population.
Again, I urge all interested readers to examine the segments we have
derived. They provide some interesting insights into use - not all the news
is good news.
Best,
DLH
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Professor Donna L. Hoffman hoffman_at_colette.ogsm.vanderbilt.edu
Owen Graduate School of Management 615-343-6904 voice
Vanderbilt University 615-343-7177 fax
Nashville, TN 37203 129.59.210.109 CU-SeeMe
Project 2000: http://www2000.ogsm.vanderbilt.edu/
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
