pmouse wrote:
Ledi, I recognize that huge spike in the second half of '02 because it's when I joined. Yahoo mail users got a link in the 'weekly picks' newsletter that Yahoo used to send out. It's fascinating that that recommendation had such an enormous effect on traffic. Or was there some other outreach program going on then?
You can see that traffic dropped off by a bunch when neo incorporated the ad banners, but I have no clue why traffic dropped off by 20% beginning last January, even discounting the advent calendar spike in December. Anyone have a theory?
I don't think that's just due to that newsletter - it's more likely that's when there was enough popularity to spark a fad. For all the December spikes, that's probably winter vacation and its end. Same for the summer humps.
The down spikes (best visible without smoothing) are probably due to downtime. You're not going to get many page views if all of them are the red Pteri.
Here's a fairly wide graph of all the data with lots of smoothing applied:, which brings out the humps very well:
clickQuote:
If I had a better idea of how these, I could tell if this was normal weekly fluctuation, but from the looks of it there hasn't been any decline in pageviews since the redesign. People need time to look at it before they decide to stay or leave, so the real test will be the coming months's pageviews.
One can arrange the data into a 7xn matrix and run
principal components analysis on it to find the most significant few weekly trends.....or should be able to, anyway. (I hope I'm remembering this right, since it'll be on one of my finals two weeks from now!) One should be able to do it on 365xn for yearly trends, but then there's probably not enough years to do it with...
I'd do it - I'm very curious too - but I don't have access to the data.