Skype hasn't told us important information about their numbers. This hurts our ability to interpret them (not that we'll stop trying). Continue and see our letter to Skype.
Dear Skype,
Have you always measured "minutes served" the same way? Or has the data collection changed?
Does a 5 minute conversation among 3 friends = 15 minutes served? Or 5? Are they "person-minutes" or "conversation-minutes"? If one of the friends is only on for 2 of the 5 minutes in the call, do you report 12 minutes served (5+5+2)?
Are you including fractional minutes? Your contract says you bill 64 seconds as one minute and 65 seconds as two minutes. Are you counting "minutes served" this way too?
How do you get your data? Are Skype clients pinging you as they enter and exit a call? Or do they ping you during a call, a keep-alive ping? Do all pings get through; are you scoring 100% of the calls made p2p in the network?
About the number of people on the network at the same time:
Why do I see a different number than someone else in the cloud?
People log in or log out all the time. How long does it take for you to learn of this action?
What happens when someone just kills their Skype client, not politely changing their presence. How long does it take for my absence from the network to reach you?
It takes time for the cloud to propagate data. Are you getting this data equally from all regions in the world?
Dial-up users come and go more frequently than midband users. Do you have any way to distinguish dial-up presence from others?
If my presence is set to "invisible" do I still show up in your numbers? If my presence is set to "away" or "do not disturb"?
If I have multiple identities / instances of Skype running on my computer, do you count each of them, from the same IP address? Are you counting Skype IDs or people?
If you spell out your methodology, it will help the news and analyst ecology to get our stories right.