Considering that Linux is not “sold”, how do you get accurate numbers on the number of users? And the laptop I’m running Linux on, came with Vista (a steaming pile if I ever saw one), that I’ve wiped. So technically, that counts as a Windows seat, even though it occupies no space on my HD.
Usually by tracking web activity. Your computer informs every web site you visit what type of operating system you are using, which web browser, etc. Of course, you have to statistically adjust for how likely someone is to defeat that; Fewer than 1% of Windows users do, but upwards of 30% of Linux users do. Are those numbers solid? No. But if Linux numbers should be adjusted up to 1.2%, instead of 0.9%, does that change the truthfulness of what I wrote?
Otherwise, you could compare a Linux download to a Windows sale. My guess is that greatly overestimates the portion of users using Linux, since a lot of people use it for limited purposes. (I’ve downloaded various Linux flavors a good dozen times, but bought a Windows-loaded computer only about four times.) But that’s still a whopping 2%.