2007-01-30

Peak Users Online = Growth?

In the graph below i compared and superposed the shapes of the curve of users online of May 15, 2005 and January 17, 2007. Let’s discuss these graphs and the changes in their shapes …

The strange and brutal “dip” just before 16h GMT on January 15, 2007 is some kind of flaw in the counting system of concurrent users online, and has happened several times in the past. Several people noticed it for the first time the day we reached 7 million concurrent users online (see my post "frozen Skype Cloud" on August 29, 2006). Until now, no “official” explanation from the side of Skype*. But it is an erratic phenomenon, because it doesn’t happen every day, although it happens always when a new record of concurrent users online is reached. So let’s discard this phenomenon.

In the “old” pattern of 2005 there are 3 distinctive phenomenons:
  • Around 12 GMT there is a dip in the concurrent users online (green curve). The last months this has almost completely faded away (red curve). This could be explained by the strong growth of the American users.
  • Peak time was always occurring around 16h GMT. Now, it happens at any moment in the time span 15h to 20h GMT, moments were both Europeans and Americans are using their computers. In the past, Europe was the strongest Skype user continent. Perhaps now it are the Americas?
  • The “night dip” was in the past around 02h GMT. Now this dip seems to have risen (proportionally), what again could be an indication that more Americans are still awake and using Skype before that moment.
So what? Is Peak Time Concurrent Users the right way to measure the success of Skype? No, certainly not, but it is the easiest number AND to measure AND to understand!

A better measure would be to calculate the surface below the curve of concurrent users online (see the blue surface in the image on the left). This surface has as unit “users x time” and gives a better indication of the number of users AND the time they stay online.

Therefore i also colored in blue the “gains of relative surface” between the two curves of the first graph. This shows where the Growth of Skype Users is taking place:
  • Or indeed American growth
  • Or people stay longer online
  • Or … both
If it is American growth, and if this growth continues, we should see the peak of concurrent users online shift to the right side in the graph in the next months. This will be enough proof for me that Hudson Barton was 100% right, unless Skype explains it officially in the meantime.

* there was however yesterday a strange comment of Jaanus Kase on the official Skype Blog, that seems to confirm there is indeed a bug in the counting system! His words after commenting the 9 million concurrent users online milestone: "Next up: double digits in the millions department. I wonder if our testers have actually tested that it will display fine, or will it crash and burn as soon as it hits the tenth million…"

3 comments:

Claudius said...

"I wonder if our testers have actually tested that it will display fine (...)"

I think Jaanus was reffering to the online user count in the skype client software and not the RSS feed or the stats aggregator in general.

Jean Mercier said...

Claudius,
When the Online Users behaviour goes down unexpectedly, the data coming from the RSS Feed shows the same weird behaviour!

I don’t know how Skype makes the estimate of Users Online, but I guess the system used in the Skype Client and “for” the RSS feed are the same!

Hudson said...

Thanks for your kind words, Jean. However, the shift of the weekday peak from 16 GMT to 18 GMT, assuming that this is a permanent shift, could have several explanations other than "America is growing faster than Europe". I prefer this explanation: that Europe or probably both regions are becoming increasing oriented toward evening and weekend usage. Note that the weekend peak (Saturday) has for a very long time been 2-3 hours later than the weekday peak. Folks get up later (and go to bed later) on weekends, so the shift is hardly surprising.