Don't Believe Every Stat You Read
Mobile OS analytics prove deceptive
There are lies, damned lies, and--well, you know the rest.
Statistics about operating systems are, by their nature, a bit hard to gather. Computers and handheld devices don't typically broadcast their status to the general Internet populace. If devices do phone home, it's usually to the hardware or software makers, who then dutifully report exaggerated numbers to make their bottom line look good.
Getting objective numbers, then, is more of a challenge.
One time-honored way of getting operating system statistics is to use web analytics to sample incoming traffic requests and figure out what browser and operating system the requests are coming from. This sorta-kinda works, but it relies on a lot of "ifs" for accuracy. Is the browser user-agent correctly interpreted? Which sites are being sampled?
This last is pretty important: if a web site sample included data from, say, Linux Today, Slashdot, and LWN, then one would expect the number of Linux users visiting these sites to be higher than those in a sample that had MSN, Windows.com, or IHatePenguins.com. A bit of an exaggeration, but you see the point.
That's why sampled data is so different. A comparison of browser data, for instance, shows a pretty wide variation of percentages on browsers, because there are different samples and different sampling methods used.
All of this, I should emphasize, is a long way of saying don't take these numbers as gospel. They are only meant to represent general trends. But sometimes the trends seem too interesting not to point out.
On Wednesday, I wrote about recent moves some open source projects are taking to replace Firefox with the Chromium browser. In that article, I emphasized that, based on the same chart I linked above, Firefox seemed to be in good shape, and ultimately the Chromium moves were just something that happens in the product-rich environment like ours.
But while I was noodling around in the throes of data, something else caught my eye. I was poking around the Wikimedia Visitor Stats for May, and noticed something peculiar: operating system stats for Linux were (as usual) depressingly low. The numbers for Android browsers, however, were shockingly bad: only 0.20 percent of reported page requests came from Android devices, compared to 1.60 percent of requests coming from iPhones. (Linux in general came in at 1.80 percent, in case you were wondering.)
Clearly, that seemed really off, considering the claims made by market researchers that put Android device numbers much closer to iPhone numbers. I know there's a disparity, but Android requests being eight times smaller than iPhone requests? Whoa.
Again, there could be other factors at play. If Android users are using Wapedia, Quickpedia, or some other app to get info, that may not register as a browser-based page request.
Other stat gatherers seem to hold up similar proportions: Clicky shows around 60 percent of incoming mobile OS traffic belonging to iPhones, and about 12 percent to Android.
Ah, hang on. A telling stat that makes me wonder what's going on: Blackberry numbers are about 13 percent, which definitely seems off because every market survey I have seen to date still puts Blackberry in the market lead position. So, what's going on? Given that Blackberry is the known market leader, why do iPhones get so much higher traffic?
This has been noted before: mobile web requests have usually not reflected the sales data, which SearchEngineWatch noted back in May 2009. But, also tellingly, iPhone web requests have historically been much higher than anyone should expect.
Clearly, web traffic as OS analyzer is a faulty method of gleaning data, because usage patterns are so broadly different. And it does make sense that iPhone users would be more likely to surf than Blackberry users. Android, however, puzzles me. All I seem to do on my Android is surf and check e-mail, so I am unclear how my usage pattern differs from that of an iPhone user. Are these folks just that much more connected?