[Foundation-l] Wikipedia growth since February 2004 greatly slowed - uh, no

Fri Jul 16 07:40:03 UTC 2004

--- Pete/Pcb21 <pete_pcb21_wpmail at pcbartlett.com> wrote on wikien-l:
> Christopher Mahan wrote:
> I've been meaning to find a peg to hang the following post on ever since 
> I lost the post with mav's original 90% growth projections on:
> 
> All the evidence (see below) suggests that Wikipedia has *not* grown by 
> a significant amount since the huge explosion in February/March 2004.

Huh? Check out the graph on page one here (again, this model assumes business
as usual):

http://meta.wikimedia.org/upload/0/07/Hardware_costs_-_year.pdf

The first big bump was in Q1 2004 (Jan-March) and the second, much larger spike
was in June during Q2 2004 (1.21 M hits/day on average). We are still about at
the same level (1.23 M hits/day average) as June, which is about the same as
the figure in the model I developed (1.29 M hits/day average for July  2004).
However that model averages each three month period so we won't know just how
close the estimate is until the end of this quarter.* 

But replacing the model figure with the current partial one and extrapolating
for the rest of the quarter still gives 75% more hits/day for Q3 vs Q2 (the
model predicts 85%). That assumes we keep the *same* average we have now for
each month this quarter (1.123 M hits/day). Our average may do down but more
likely it will increase to at least the modeled figure (I fear it may surpass
it). 

> ''However'' I do not think it is a coincidence that this four months of 
> little growth corresponds almost exactly with the period that Wikipedia 
> has "disappered" from Google.

I think we did take a two month hit from that and that may explain why our
traffic only grew by 73% from Q1 2004 to Q2 2003 but we also have to take into
account that our traffic grew 176% from the last quarter of 2003 to the first
quarter of 2004. So that *mere* 73% growth *alone* is in fact more traffic than
we served in all 2003 *combined* (that's right, add up the totals of each of
those 12 months and you are just shy of the *increase* in traffic from the
first three months this year to the second three months). 

Given our continued rapid growth in spite of our Google rank situation, I think
an exponentially growing number of people are starting to go to Wikipedia
*first* and then using Google in the increasingly rare cases that Wikipedia did
not have all the info they needed. 

So the Google effect, like the previous Slashdot effect, is starting to show
some signs of having somewhat less influence on us, IMO. I won't be surprised
if Google comes to us in a years time trying to push us into having Google ads.

NOTE: I just revised my January filler figure from 5000 K hits/day to 1600 K
hits/day since that is the carry over number from the last time we had data
(October 2003). Also, IIRC the server farm wasn't installed until late January
so a lower figure is probably closer to reality. 

> This time last year WP was very often top of typical searches. Now our 
> mirrors always come out ahead except for very recently breaking news 
> (where they don't have the new content yet). What I term the "second 
> generation" mirrors (such as thefreedictionary.com) are "gaming" the 
> Google algorithm much better than Wikipedia is, and in a way the older 
> mirrors (4reference.net etc) didn't.

This is indeed a problem, but a good one until we can firmly get a handle on
our finances. The server situation would be much worse if the articles we host
were still in the top ten like they were before Google changed their PageRank
criteria. Since we do not get ad revenue from hits and our traffic/popularity
is already growing exponentially, I don't see a problem with our mirrored
content getting higher PageRanks than the originals. At least for now. 

-- Daniel Mayer (aka mav)

__________________________________
Do you Yahoo!?
New and Improved Yahoo! Mail - Send 10MB messages!
http://promotions.yahoo.com/new_mail 