Linked by Thom Holwerda on Tue 24th Jul 2007 15:11 UTC, submitted by flanque
In the News Break out the salt, boys and girls, it's time for some statistics again. "According to Net Applications, in June Windows Vista accounted for 4.52% of all systems that browsed the Web, up from January's 0.18%. Vista has grown its usage share each month since its release to consumers Jan. 30, hitting 0.93% in February, 2.04% in March, 3.02% in April and 3.74% in May. Apple Inc.'s Mac OS X, meanwhile, accounted for 6.22% in January and hit its high point of 6.46% in May, but it slipped back to 6% in June. If Vista's uptake trend continues, it should pass Mac OS X in Web usage share by the end of August." Do with it as you please.
Thread beginning with comment 257833
To view parent comment, click here.
To read all comments associated with this story, please click here.
RE[2]: US market or world market
by Googlesaurus on Tue 24th Jul 2007 17:46 UTC in reply to "RE: US market or world market"
Googlesaurus
Member since:
2005-10-19

Someone correct me if I am wrong......

It appears to me there is a serious problem in using web usage statistics to calculate the OS being used.

Some non-windows browsers emulate Internet Explorer, indicating the use of Windows even when it's not the case.

I can't prove these statistics are flawed, but I'm sceptical of the accuracy of such numbers.

Reply Parent Score: 3

MollyC Member since:
2006-07-04

"Some non-windows browsers emulate Internet Explorer, indicating the use of Windows even when it's not the case.

I can't prove these statistics are flawed, but I'm sceptical of the accuracy of such numbers."



You make a valid point regarding inherent flaw of using web usage stats as indicated by a browser's UserAgent string as being indicative of usershare of an OS.

But there are three things that argue against your scepticism.

First, the chart shows Vista's share increasing and XP's decreasing. So in order for your scepticism to hold water, non-windows browsers that were misidentifying themselves as XP are now misidentifying themselves as Vista. I'd guess that the percentage of browsers doing that is virtually zero (what would be the point?).

Second, the chart shows the monthly trends, rather than just the stats for one month. Meaning, that you could subract 5% of Vista's totals each month to account for "lying browsers" and the trend would still be upward (I picked 5% out of the air; I'd say the real percentage of browsers claiming to be Vista that aren't is virtually zero).

Third, the non-Windows OS with the largest userbase is Mac OSX by far (one order of magnitude larger than Linux, according to these same stats). There's no way that Safari is identifying itself as Windows, and I'm sure that Mac Firefox doens't either. So since Mac browsers aren't identifying themselves as Windows, that leaves only the non-Windows browsers of *nix, which have a small market share, and of those, how many are really set to identify as Windows browsers? We're talking about a very small percentage of browser share at this point.

Edited 2007-07-24 18:09

Reply Parent Score: 5

archiesteel Member since:
2005-07-02

First, the chart shows Vista's share increasing and XP's decreasing. So in order for your scepticism to hold water, non-windows browsers that were misidentifying themselves as XP are now misidentifying themselves as Vista.


Actually, no. Because Linux browsers represent such a small part compared to Windows' market share, you could have an increase of Vista share and a decrease of XP share even if all the Linux browsers continued being identified as XP. As in all statistics, one should expect smaller numbers to be *less* precise than bigger ones. So one could still see big trends (i.e. XP towards Vista) while smaller trends (Linux' market share increasing, for example) would be almost certainly swamped in statistical noise.

Some Linux browsers don't identify themselves as Linux by default, by the way - Konqueror is one example, IIRC, which would mean that a large proportion of KDE desktops would fall into the "others" category.

The fact that Mac browsers don't identifiy themselves as Windows would mean that the Mac numbers are more accurate, however they could in fact be a couple of points higher, because the proportion of broadband users is probably higher among Mac users (since they are in general more affluent). Broadband users tend to keep the same IP address for longer, which would fool these kind of studies. Also, some Windows client (such as AOL) increase the number of referents when visiting a page, which can also skew things.

The conclusion? Web stats are pretty much useless when it comes to giving any kind of precise numbers. It's all right when it comes to large percentages, or for Windows OSes (which don't represent themselves as other OSes when surfing), but for small numbers the margin of error is too big to be of any use.

Reply Parent Score: 2