[ExI] Click path length

[Bryan Bishop]

There was once a paper published that explored the probability of a
group of users following a link on a page, and then the likelihood of
continuing on a particular chain of links. Anyway, for a website that
does not have many users, it's unlikely that you are going to be able
to get the users to where you want to go. However, when you have a
ridiculously high volume of traffic, it's like liquid- even the small
percentage that does happen to get to where you wanted them to get to,
it's still very significant.

So, let's look at Google's home page today- eternally restricted to,
what, 24 words? There's one particular link in a prominent location to
CADIE, this year's April experiment. On the following page there are a
total of six links. One leads to youtube- so we're three levels down
at the moment. I examined over this page and I immediately saw this
user named Tyler Emerson in the subscriptions box, but with the face
of Eliezer Yudkowsky (pulled from one of the uploaded videos).

Wait, what? Let's see some estimates for the number of page views that
the number one site on the web is likely to get in a 24 hour period-
200 million queries per day sometime in 2007; so let's say that 1% of
those 200 million queries are typed into the main Google page, and
let's also say that of that 1% of the total, only 1% bothers to click
on the youtube link. That's still 20,000 exposures, and that's a lower
bound on the estimate. How about 10% and then 10%? Those are some
pretty big numbers.

Though honestly, I do more than 200 queries per day, so if this was
the same across the entire population, that's really only a million
different people querying (maybe), and only a small percentage is
going to be looking at the links, and even less at the resulting
links- heck, most people don't even look past the first or second
search result (apparently) on the search pages (nobody knows how to
use a search engine, as it turns out).

Anyway, those are some pretty big numbers when it comes to audiences.

- Bryan
http://heybryan.org/
1 512 203 0507