[Air-l] 988 exabytes of info created in 2010

Bram Dov Abramson bda at bazu.org
Wed Mar 7 08:10:41 PST 2007


> The results of commercial research firms are always so nicely
> packaged and illustrated---and maybe that's one of the reasons that
> actual real world decisions are based on them.  (50 Rhode Islands?
> That sounds like a lot, I'd better buy more ;)
>
> Which, of course, makes them 'important' but for reasons other than
> their 'truth'.  Their importance seems self-reinforcing

Oh, I don't know.  The latter thesis -- that, faced with different answers
to a similar question, "actual real world decision" makers prefer the
truthiness of name-brand commercial research -- is no doubt possible. 
Perhaps there are domains of activity where it is often true.  But it's
certainly not my experience as both a producer and user of non-academic
indicators-based research, each in both commercial settings and in
government.

Rather, I've found that a less sneering version of the former thesis fits
a lot better.  But it's not just packaging and illustration: it's choice
of topic matter.  Any time I have seen non-academic research I produced
used "out there", it's been because there were very few alternative
sources for answers to the question being asked.  For that matter, when
there did exist alternative sources, they tended if not to converge, at
least to take each other into account.

In this instance, the few-alternatives thesis seems to be the case.  The
EMC-sponsored studies, first at Berkeley and now at IDC, have gotten a lot
of press.  I have not seen a lot of other comprehensive attempts to answer
the "how much information" question -- a rather different proposition,
obviously, then what one thinks of that question.

For what it's worth, an IDC narrative of how they counted things, as
opposed to a third-party nwes story about it, is at:

http://www.emc.com/about/destination/digital_universe/pdf/Expanding_Digital_Universe_IDC_WhitePaper_022507.pdf

>  research huh?  Does that qualify as an 'A Journal'? ;)

I wouldn't think so.  Academic and other forms of research each have their
own norms and control mechanisms.  They differ from each other, and they
differ internally.  I would wager that those who place all weight on how
they differ from each other, and no weight on how they differ internally,
will end up with trust mechanisms that disappoint them in the long run.  
I have seen very careful and properly-reviewed non-academic research --
even shepherded some through reviewing myself -- and, by the same token,
some very inaccurate and poorly-reviewed academic journal articles.

So perhaps the journal/not-journal binary is not a good totalising lens
through which to view the wide world of doing research and other words for
figuring stuff out.

> Is anyone aware of research that asks the question,
> 'how reliable are  the predictions of commercial
> research firms?'.

... or that asks that question of any form of institutional research.  I
wonder how useful such studies would really be.  In the telecom and IT
sectors there have, from time to time, been scorecards in industry
magazines, which would fairly up front about how they did their scoring. 
I suppose this could get you started.

(Parenthetically: one firm, eMarketer, does something like this.  More
specifically, and at least the last time I checked, their business model
was that, rather than undertaking primary research of their own, they
bought or borrowed, etc. other primary research and conducted
meta-research -- that is, showed what all the other research firms were
saying, then sold that as a competing product.  I suppose there are issues
of shooting one's golden goose here.  But that's another story.)

More helpful from a moving-things-forward standpoint, I would think, would
be to investigate the methodologies being used.  The standard by which
predictions should be judged is a bit of a moving target -- ensuring that
a prediction is based on sound reasoning and data is more do-able, and
probably more reasonable.

cheers
Bram



More information about the Air-L mailing list