concerns about statcounter

StatCounter is a relative newcomer to the public Web share stats game and I've been following them pretty closely for about 8 or 9 months now.

I really like their quick and dynamic graphs (though I'd obviously prefer they use Open Web tech rather than Flash) and their 4 billion page views across 3 million websites sounds like a pretty good sample if it's even close to evenly distributed globally.

The problem I have with their numbers is that they have a lot of unexplained spikes that I just can't make heads or tails of. For example, they're reporting that IE 6 gained almost six and a half points in usage from yesterday to today. Looking at a longer-term trend, it just doesn't fit that IE6 would have such an unexplained resurgence.

The clear trend for IE 6, that I can see in the number of accesses to the Mozilla websites, the number of IE 6 users downloading Firefox, the usage percentages from other stats packages, and the obvious movement from older versions of IE to newer versions, is that it's dropping at a steady rate.

Now, in IE 6 does vary more than any other browser between weekend and weekday usage. The remaining ~20 share that IE 6 holds has a much stronger enterprise skew so the "work versus home" difference is more pronounced than with any other browser or browser version. You can see this in the daily stats at both StatCounter and Net Applications so I'm not surprised to see the difference cause a pretty big zig zag up and down as the weeks progress, but this just doesn't look right to me:

I'm really excited to see another stats package publishing its metrics to the public. It's a really helpful service and I hope that more programs (like omniture and sitemeter) will start publishing their aggregate data for public consumption. But. As important as it is to have more data, it's also important that the data is consistent and trendable. I don't know what's up with these random spikes at StatCounter, but I hope that with some time they get things ironed out.

When they do get the bumps (literally) ironed out, I will start to include them more in my regular browser landscape updates and analysis.