Analyzing client software as a function of referring URL

The rationale here is that different types of people tend to view different types of sites, and that these different types of people also use different types of software. This because the general common wisdom of what percentage of people use Linux, or Internet Explorer, does not agree with my web statistics. With 169,000 referrals total, I have enough data to break it down into groups and still get meaningful statistics.

Analysis of client OS and browser vs subset of page fetches

Refereed by:	Google.com	Google (European)	Non-Google Search	Slashdot on July 22	Non Slash- dot, July 22	Slashdot Japan	Hardocp .com	Unmen- tionables	Fetches of Jhead page
Total hits	16131	2168	16959	49138	109476	15353	7288	11806	53494
Browser Technology:
Gecko based browser	10.9%	13.8%	2.5%	38.8%	27.3%	27.8%	10.3%	4.7%	24.0%
Internet Explorer	85.9%	82.7%	95.7%	57.1%	59.6%	65.7%	88.1%	94.2%	63.4%
Client OS:
Linux	4.1%	9.3%	0.3%	14.5%	11.4%	6.5%	1.3%	0.3%	13.8%
Mac	6.7%	2.5%	1.5%	5.4%	4.6%	7.9%	0.8%	2.7%	5.2%
Windows XP	34.9%	33.4%	36.7%	33.0%	31.1%	24.5%	60.6%	50.8%	35.6%
Windows 2000	24.8%	28.6%	21.2%	34.8%	32.2%	45.2%	25.5%	13.0%	25.0%
Windows NT 4.0	2.5%	3.6%	2.8%	2.4%	1.8%	1.0%	2.1%	0.7%	1.1%
Windows 9x	23.8%	18.3%	34.6%	4.3%	6.6%	7.0%	7.2%	29.9%	7.8%

More on this table and how it was generated

So what conclusions can we draw about certain types of users?

Slashdot readers

The people who came directly thru Slashdot have the highest portion of Linux usage (14.5%), and the lowest portion of Internet Explorer usage. Of course, Slashdot plugs Linux all the time (that was its primary original purpose when it started). Still, Slashdot users are very numerous, and reasonably representative of the tech community. I was quite surprised to see Linux at 14.5% for the client OS - maybe Linux for the desktop really is happening.

Interestingly enough, Linux usage drops off substantially for referrals not directly thru Slashdot. At least 95% of the page fetches on July 22 were as a result of Slashdot, but a lot of these pages were not referred directly from Slashdot. I suspect the people who click on every article off Slashdot, but don't read as far are more likely to be Linux users. That, or people who bookmarked it for reading it later in the day are less likely to be Linux users. Not quite sure.

These people found my site using www.google.com. As www.google.com is fairly mainstream, this set of users is probably fairly representative of users who are fairly savvy. Still, these users did search for something on my site, or at least were curious enough from the text in Google to click on it. I consider these users to be fairly representative of savvy and interesting users. Certainly, they weren't searching for Britney Spears or n-sync.

These people are 85% likely to use Explorer - not quite as bad as the 95% figure one hears sometimes, but of course, these people are still fairly with it. The 4.1 % Linux usage is also quite nice to see.

People who use anything other than Google to search the web are probably not as technically with it as those who do. The column for Non-Google search referrals seems to confirm this.

This set of users has the lowest portion of Linux users (0.3%), as well as the lowest portion of Gecko based browsers. These people use the PC with whatever it came with. Not surprisingly, this group uses mostly Windows XP, or Windows 98, and Explorer - the default install. Sadly, this group is as large as the Google.com users.

These are users of Google from primarily European countries.

These people are more likely to use Linux than the American Google users. My theory is that the further you get from Seattle, the more the world is Anti-microsoft, so this comes at no surprise to me. Interesting that the Europeans use the Mac so much less. They have a higher portion of Windows NT 4.0 and Windows 2000 users - strikes me as more serious people that way. Probably more browsing from work, and they aren't as quick to upgrade to the latest and greatest as Americans. Of course, this is a small sample size. The European mac users number just 25, so the ratios have a bit of uncertainty to them in that column.

These are users of Google.jp

Interesting that these people use Windows 2000 so much, more than any other group. I suspect this is because they work so much, and browse from work. Windows 2000, unlike XP, is well established in businesses. I'm surprised at the relatively high (6.2%) use of Linux. But then again, it is quite far from Seattle.

Generally users of the latest technology - the highest rate of Windows XP users. Few Linux or Mac users - Most games only run under Windows. About the same rate of Gecko based browser users as the Google referrals - this is good.

These all are from sites that are fairly sex oriented. No idea why they linked my page.

These people are as unlikely to use Linux as the non Google searchers. I'm actually pretty happy they don't use Linux!

These people are the least likely to use Windows 2000 or Windows NT. This is probably because more browsing of that nature is done from home. Interestingly, these people have much more up to date hardware than the non Google searchers. I guess they need it for all the 'multimedia' content!

These people came to get my Jhead program.

These people are generally quite sophisticated power users, who own digital cameras. As such, they are very unlikely to use Widows NT (too old for that, doesn't do USB). Some of them have websites, and thus are quite internet aware, running Gecko based browsers and Linux (for their digital picture gallery web servers).