[thelist] flash detection

[matthew garrett]

> From: jdowdell at macromedia.com (John Dowdell)
 > The MediaMetrix consumer audits are done on *viewability*, not on
> downloads. Of the representative consumer samples who are asked to view
> pages, X% can do so immediately.

Note: I'm politically neutral on Flash.

The thing that jumps out at me from the viewability methodology is that the
extrapolation from the sample is greater than 179,000 to 1. A sample the
size of the [thelist] is speaking for the entire world. While I don't have a
degree in statistics on me, this kind of math always makes me uncomfortable
with the professed accuracy.

For every 6 people that are not truly representative, the total is off by
over 1,000,000. The stated risk alone (5%) is 17,200,988. That's not a large
percentage, but it's a *lot of users*.

I'm *not* saying that the numbers are wrong, but the old "lies, damn lies,
and statistics" bit is never far from my mind when looking at these things.

matt