August 28, 2009
measuring global browser share
I've recently been bad-mouthing the browser usage share data from both StatCounter and Net Applications. There remain some really strange internal inconsistencies with very serious and unexplainable spikes and dips for some of the browsers they're monitoring and I'm discouraged by that.
But until we find some better sources of data, I think it might be useful to try to figure out what we can learn from these sources, if anything. My first pass at this task was just to do some rough comparisons between the two and what I found was both surprising and I think positive.
Here's a chart comparing the last year of data for Internet Explorer and Firefox.
The first thing of note is that the two sources come up with radically different shares for Firefox and IE. The difference between the two is nearly as large as the sum of all the other browsers they measure. I have no idea which of the two sources is closer to "the truth" and I don't think that just averaging them gives us any better picture of "the truth" so I don't think that's a productive avenue but maybe some statistician can show me otherwise.
The second issue worth remarking on is that there's some consistent seasonal changes. Firefox has considerably stronger growth in the winter than it does in the summer and that shows up in both data sources.
The third thing of note is the surprising and more positive bit. Over the course of a year, the two sources are in reasonably close agreement on the trends. Net Applications reports that for this time period, Firefox gained ~4.6 points of usage share and StatCounter reports a gain of ~5.4%. Looking at the IE losses during the period, Net Applications sees an 8.7% loss and StatCounter sees a 9.9% loss. That's really not too wide a gap in the trend, even if the month to month deltas are a little wackier and even if the two pretty firmly disagree on the total share for each browser.
So, my first conclusion, probably an obvious one, is that it's probably not a good idea to fret over the month to month variations but that the longer term changes are probably worth paying some attention to.
This doesn't address any of the geographic disparities, but I think that's a problem we're basically stuck with until we get more representative sources. Given the general agreement between these two sources on at least one front, are the longer term trends worth giving credence to? What do you all think?
reactions, thoughts, comments, etc.
Posted by: Stifu | August 28, 2009 12:19 PM
I get the feeling Net Applications is measuring corporate type websites to a higher degree than StatCounter, since they only have paid services, right?
That would quite logically explain the differences.
That to me feels quite positive, because that would mean that StatCounter to a higher degree is showing what people use at home, where they have *a choice*.
Posted by: David Naylor | August 29, 2009 12:03 AM
David Naylor, Net Applications has their services mostly in corporate websites but that doesn't mean people only browse those websites at work. People still browse corportate websites at home, like the official Sony, Nokia, Apple, Microsoft, etc. etc. websites.
I'd say Net Applications has a relatively better representation of the state of the Web than StatCounter, since Net Applications have big name sites like CNN, New York Times, Sony, Microsoft, Yahoo, etc. etc. so it has a better sampling rate than StatCounter. When analyzing internet traffic, one big name portal site can produce more meaningful data than a whole bunch of personal sites together.
As for the trend, IE's marketshare is decreasing and Firefox's marketshare is increasing, that's the fact, and we all know that. The only question is how steep/flat the curve is in the real world. By StatCounter's data, it looks like Firefox will surpass IE in 2011, but I doubt that will happen in real world.
Posted by: kaixin001 | August 30, 2009 5:02 PM
You may have a point there about Net Applications.
Posted by: David Naylor | August 31, 2009 2:13 AM
kaixin001: I agree with your comment about the slopes of the graphs. What is interesting is that although the absolute numbers differ, the slopes of the graph show the same general trends at almost the same rate.
Regarding the eventual intersection of the lines, only time will tell. Asa had a historical chart that shows that few events seem to affect the larger pattern. However, there does seem to be a lot of activity and competition, so the either of the slops could become flatter or steeper between now and 2011.
Posted by: Ben Chuang | September 8, 2009 12:34 AM
IE going down, Firefox going up, slowly but surely. And no ultimate stats source, but interesting trends. So yeah, in other words, nothing new really. :)