I was discussing on a private mailing list whether or not this number is
for real. Someone complained it wasn't. Here's my response:

-------------

Your utter certainty regarding the flawed methodology of their results
is somewhat ironic given that you've done no apparent research into how
they came to those results, nor any suggestion of what a more accurate
number might be.

A real critique would go to the source of their data and show why it's
bad. I'll help you with that. I found this year's report by simply
updating the year in the link from last year:

http://www.ifpi.org/content/section_resources/dmr2010.html

The actual quote from there is:

> Estimates on the impact of internet piracy vary but are consistently huge in scale. IFPI, collating separate studies in 16 countries over a four-year period, estimated unauthorised file-sharing at over 40 billion files in 2008. This means that globally around 95 per cent of music tracks are downloaded without payment to the artist or the music company that produced them.

When read in the actual context, they seem to admit it's a hard thing,
and acknowledge they're estimating. Furthermore, contrary to your claim
that it was pulled out of thin air, they kindly cite a variety of other
sources. I haven't dug into them all, but one is:

http://www.ipoque.com/resources/internet-studies/internet-study-2008_2009

The paper is 14 pages long with lots of interesting data, as well as a
completely description of its methodology. A weakness of their data is
they didn't cover North America or Western Europe as a whole (maybe due
to data privacy laws?). But focusing on Germany alone (which seems
middle of the road in terms of its data compared to other regions), it
says 53% of traffic was P2P downloads (separately from VoIP), to 26% web.

So right off the bat, P2P downloads account for double the traffic of
*all of HTTP* across their not insignificant sample set.

It doesn't seem unreasonable to assume that, say, 90% of that P2P
traffic was pirate. (It's what it was designed to do, after all.) So
let's say 47% of all traffic was pirate content downloads.

As for legitimate content downloads, we could probably back into that by
estimating bandwidth consumed by iTunes using public sales numbers and
average file sizes. But let's say iTunes accounts for 10% of *all HTTP*
(which seems astronomically high). That would mean 2.6% of the internet
is legitimate content downloads (separate from streaming).

47 + 2.6 = 49.6% of the internet devoted to content downloads
47/49.6 = 94.75% content downloads pirated.

Huh, I didn't even plan that out.

Anyway, there are obviously problems with all that data, how it's
collected, how it's analyzed, etc. But to categorically assert that the
data and results are flawed and everyone involved in the process is
either actively lying or allowing themselves to be mislead -- without
providing any evidence or analysis to the contrary... let's just say
your own methodology could use a bit more scientific rigor.

-david