Signed, sealed and delivered in closed envelopes

Summary: Can contradictory statistics be trusted when the recipe for calculating them (and data too) are a kept secret?

AS A SCIENTIST, I CAN attest from personal experience that one must never underestimate the importance of sharing not only methods but data as well. Without them, how can anything be trusted or verified, particularly when one prominent source of the information is partly led by a former Microsoft employee (and maybe existing investor), to whom Microsoft is currently a paying customer? Yes, we wrote a lot more about Net Applications and its 'magic' (but secret) numbers.

ITWire, a publication of growing influence in Australia, has repeatedly pressured Net Applications to share some facts about its numbers, but there is still no response. Here is the latest article:

I looked into where Net Applications gets data. They say “We use a unique methodology for collecting this data. We collect data from the browsers of site visitors to our exclusive on-demand network of live stats customers. The data is compiled from approximately 160 million visitors per month.”

That’s certainly a lot of data, although it would be nice to know who the “exclusive on-demand network” participants are. For instance, if Microsoft’s TechNet web site is included (which may or may not be the case, but a Microsoft logo is part of a rotating list of logos displayed) then you would expect it is likely visitors to that site tend towards using Microsoft operating systems.

This is how the author, David, summarises it in Slashdot:

"Q: What is the true market share for Linux? A: Depends who you ask! You can find desktop penetration of 1%, 2% and 4% — and a server share right up to 46%!"

David's colleague, Sam Varghese, is probably still waiting for answers from Net Applications, but he should not hold his breath. Net Applications just unleashes numbers and experience suggests that it rarely speaks to the press. Its numbers are taken for gospel and propagated further mostly by the pro-Microsoft press. We wrote about this before, e.g. in:

• Three New Articles Question Net Applications' Integrity
• Net Applications: the Big Lie, Boosted by IDC|IDG et al
• Is Netcraft Promoting Microsoft Just Like Net Applications?
• Summary: Lies, Damn Lies and Net Applications (Fake 'Statistics')

Earlier today, Fewa pointed out that "even Microsoft's numbers have Linux higher." He points to this presentation from Steve Ballmer as proof. Microsoft suggests that the desktop market share of GNU/Linux is higher than that of Apple, which also happens to be among the paying customers of Net Applications. ⬆